sys/vfs/gnu/ext2fs/ext2_readwrite.c

   1 /*
   2  *  modified for Lites 1.1
   3  *
   4  *  Aug 1995, Godmar Back (gback@cs.utah.edu)
   5  *  University of Utah, Department of Computer Science
   6  */
   7 /*
   8  * Copyright (c) 1993
   9  *      The Regents of the University of California.  All rights reserved.
  10  *
  11  * Redistribution and use in source and binary forms, with or without
  12  * modification, are permitted provided that the following conditions
  13  * are met:
  14  * 1. Redistributions of source code must retain the above copyright
  15  *    notice, this list of conditions and the following disclaimer.
  16  * 2. Redistributions in binary form must reproduce the above copyright
  17  *    notice, this list of conditions and the following disclaimer in the
  18  *    documentation and/or other materials provided with the distribution.
  19  * 3. All advertising materials mentioning features or use of this software
  20  *    must display the following acknowledgement:
  21  *      This product includes software developed by the University of
  22  *      California, Berkeley and its contributors.
  23  * 4. Neither the name of the University nor the names of its contributors
  24  *    may be used to endorse or promote products derived from this software
  25  *    without specific prior written permission.
  26  *
  27  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  29  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  30  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  31  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  32  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  33  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  34  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  35  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  36  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  37  * SUCH DAMAGE.
  38  *
  39  *      @(#)ufs_readwrite.c     8.7 (Berkeley) 1/21/94
  40  * $FreeBSD: src/sys/gnu/ext2fs/ext2_readwrite.c,v 1.18.2.2 2000/12/22 18:44:33 dillon Exp $
  41  * $DragonFly: src/sys/vfs/gnu/ext2fs/ext2_readwrite.c,v 1.15 2008/06/19 23:27:39 dillon Exp $
  42  */
  43
  44 #define BLKSIZE(a, b, c)        blksize(a, b, c)
  45 #define FS                      struct ext2_sb_info
  46 #define I_FS                    i_e2fs
  47
  48 /*
  49  * Vnode op for reading.
  50  *
  51  * ext2_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
  52  *           struct ucred *a_cred)
  53  */
  54 /* ARGSUSED */
  55 static int
  56 ext2_read(struct vop_read_args *ap)
  57 {
  58         struct vnode *vp;
  59         struct inode *ip;
  60         struct uio *uio;
  61         FS *fs;
  62         struct buf *bp;
  63         daddr_t lbn, nextlbn;
  64         off_t nextloffset;
  65         off_t bytesinfile;
  66         long size, xfersize, blkoffset;
  67         int error, orig_resid;
  68         int seqcount = ap->a_ioflag >> 16;
  69         u_short mode;
  70
  71         vp = ap->a_vp;
  72         ip = VTOI(vp);
  73         mode = ip->i_mode;
  74         uio = ap->a_uio;
  75
  76 #ifdef DIAGNOSTIC
  77         if (uio->uio_rw != UIO_READ)
  78                 panic("ext2_read: mode");
  79
  80         if (vp->v_type == VLNK) {
  81                 if ((int)ip->i_size < vp->v_mount->mnt_maxsymlinklen)
  82                         panic("ext2_read: short symlink");
  83         } else if (vp->v_type != VREG && vp->v_type != VDIR)
  84                 panic("ext2_read: type %d", vp->v_type);
  85 #endif
  86         fs = ip->I_FS;
  87 #if 0
  88         if ((u_quad_t)uio->uio_offset > fs->fs_maxfilesize)
  89                 return (EFBIG);
  90 #endif
  91
  92         orig_resid = uio->uio_resid;
  93         for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
  94                 if ((bytesinfile = ip->i_size - uio->uio_offset) <= 0)
  95                         break;
  96                 lbn = lblkno(fs, uio->uio_offset);
  97                 nextlbn = lbn + 1;
  98                 nextloffset = lblktodoff(fs, nextlbn);
  99                 size = BLKSIZE(fs, ip, lbn);
 100                 blkoffset = blkoff(fs, uio->uio_offset);
 101
 102                 xfersize = fs->s_frag_size - blkoffset;
 103                 if (uio->uio_resid < xfersize)
 104                         xfersize = uio->uio_resid;
 105                 if (bytesinfile < xfersize)
 106                         xfersize = bytesinfile;
 107
 108                 if (nextloffset >= ip->i_size) {
 109                         error = bread(vp, lblktodoff(fs, lbn), size, &bp);
 110                 } else if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERR) == 0) {
 111                         error = cluster_read(vp, (off_t)ip->i_size,
 112                                              lblktodoff(fs, lbn), size,
 113                                              uio->uio_resid,
 114                                              (ap->a_ioflag >> 16), &bp);
 115                 } else if (seqcount > 1) {
 116                         int nextsize = BLKSIZE(fs, ip, nextlbn);
 117                         error = breadn(vp, lblktodoff(fs, lbn),
 118                                         size, &nextloffset, &nextsize, 1, &bp);
 119                 } else {
 120                         error = bread(vp, lblktodoff(fs, lbn), size, &bp);
 121                 }
 122                 if (error) {
 123                         brelse(bp);
 124                         bp = NULL;
 125                         break;
 126                 }
 127
 128                 /*
 129                  * We should only get non-zero b_resid when an I/O error
 130                  * has occurred, which should cause us to break above.
 131                  * However, if the short read did not cause an error,
 132                  * then we want to ensure that we do not uiomove bad
 133                  * or uninitialized data.
 134                  */
 135                 size -= bp->b_resid;
 136                 if (size < xfersize) {
 137                         if (size == 0)
 138                                 break;
 139                         xfersize = size;
 140                 }
 141                 error =
 142                     uiomove((char *)bp->b_data + blkoffset, (int)xfersize, uio);
 143                 if (error)
 144                         break;
 145
 146                 bqrelse(bp);
 147         }
 148         if (bp != NULL)
 149                 bqrelse(bp);
 150         if (orig_resid > 0 && (error == 0 || uio->uio_resid != orig_resid) &&
 151             (vp->v_mount->mnt_flag & MNT_NOATIME) == 0)
 152                 ip->i_flag |= IN_ACCESS;
 153         return (error);
 154 }
 155
 156 /*
 157  * Vnode op for writing.
 158  *
 159  * ext2_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
 160  *            struct ucred *a_cred)
 161  */
 162 static int
 163 ext2_write(struct vop_write_args *ap)
 164 {
 165         struct vnode *vp;
 166         struct uio *uio;
 167         struct inode *ip;
 168         FS *fs;
 169         struct buf *bp;
 170         struct thread *td;
 171         daddr_t lbn;
 172         off_t osize;
 173         int seqcount;
 174         int blkoffset, error, flags, ioflag, resid, size, xfersize;
 175
 176         ioflag = ap->a_ioflag;
 177         seqcount = ap->a_ioflag >> 16;
 178         uio = ap->a_uio;
 179         vp = ap->a_vp;
 180         ip = VTOI(vp);
 181
 182 #ifdef DIAGNOSTIC
 183         if (uio->uio_rw != UIO_WRITE)
 184                 panic("ext2_write: mode");
 185 #endif
 186
 187         switch (vp->v_type) {
 188         case VREG:
 189                 if (ioflag & IO_APPEND)
 190                         uio->uio_offset = ip->i_size;
 191                 if ((ip->i_flags & APPEND) && uio->uio_offset != ip->i_size)
 192                         return (EPERM);
 193                 /* FALLTHROUGH */
 194         case VLNK:
 195                 break;
 196         case VDIR:
 197                 if ((ioflag & IO_SYNC) == 0)
 198                         panic("ext2_write: nonsync dir write");
 199                 break;
 200         default:
 201                 panic("ext2_write: type");
 202         }
 203
 204         fs = ip->I_FS;
 205 #if 0
 206         if (uio->uio_offset < 0 ||
 207             (u_quad_t)uio->uio_offset + uio->uio_resid > fs->fs_maxfilesize)
 208                 return (EFBIG);
 209 #endif
 210         /*
 211          * Maybe this should be above the vnode op call, but so long as
 212          * file servers have no limits, I don't think it matters.
 213          */
 214         td = uio->uio_td;
 215         if (vp->v_type == VREG && td && td->td_proc &&
 216             uio->uio_offset + uio->uio_resid >
 217             td->td_proc->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
 218                 lwpsignal(td->td_proc, td->td_lwp, SIGXFSZ);
 219                 return (EFBIG);
 220         }
 221
 222         resid = uio->uio_resid;
 223         osize = ip->i_size;
 224         flags = ioflag & IO_SYNC ? B_SYNC : 0;
 225
 226         for (error = 0; uio->uio_resid > 0;) {
 227                 lbn = lblkno(fs, uio->uio_offset);
 228                 blkoffset = blkoff(fs, uio->uio_offset);
 229                 xfersize = fs->s_frag_size - blkoffset;
 230                 if (uio->uio_resid < xfersize)
 231                         xfersize = uio->uio_resid;
 232
 233                 if (uio->uio_offset + xfersize > ip->i_size)
 234                         vnode_pager_setsize(vp, uio->uio_offset + xfersize);
 235
 236                 /*
 237                  * Avoid a data-consistency race between write() and mmap()
 238                  * by ensuring that newly allocated blocks are zerod.  The
 239                  * race can occur even in the case where the write covers
 240                  * the entire block.
 241                  *
 242                  * Just set B_CLRBUF unconditionally, even if the write
 243                  * covers the whole block.  This also handles the UIO_NOCOPY
 244                  * case.  See ffs_write() in ufs for a more sophisticated
 245                  * version.
 246                  */
 247                 flags |= B_CLRBUF;
 248                 error = ext2_balloc(ip,
 249                     lbn, blkoffset + xfersize, ap->a_cred, &bp, flags);
 250                 if (error)
 251                         break;
 252
 253                 if (uio->uio_offset + xfersize > ip->i_size) {
 254                         ip->i_size = uio->uio_offset + xfersize;
 255                 }
 256
 257                 size = BLKSIZE(fs, ip, lbn) - bp->b_resid;
 258                 if (size < xfersize)
 259                         xfersize = size;
 260
 261                 error =
 262                     uiomove((char *)bp->b_data + blkoffset, (int)xfersize, uio);
 263                 if ((ioflag & IO_VMIO) &&
 264                    (LIST_FIRST(&bp->b_dep) == NULL)) /* in ext2fs? */
 265                         bp->b_flags |= B_RELBUF;
 266
 267                 if (ioflag & IO_SYNC) {
 268                         bwrite(bp);
 269                 } else if (xfersize + blkoffset == fs->s_frag_size) {
 270                         if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERW) == 0) {
 271                                 bp->b_flags |= B_CLUSTEROK;
 272                                 cluster_write(bp, (off_t)ip->i_size, vp->v_mount->mnt_stat.f_iosize, seqcount);
 273                         } else {
 274                                 bawrite(bp);
 275                         }
 276                 } else {
 277                         bp->b_flags |= B_CLUSTEROK;
 278                         bdwrite(bp);
 279                 }
 280                 if (error || xfersize == 0)
 281                         break;
 282                 ip->i_flag |= IN_CHANGE | IN_UPDATE;
 283         }
 284         /*
 285          * If we successfully wrote any data, and we are not the superuser
 286          * we clear the setuid and setgid bits as a precaution against
 287          * tampering.
 288          */
 289         if (resid > uio->uio_resid && ap->a_cred && ap->a_cred->cr_uid != 0)
 290                 ip->i_mode &= ~(ISUID | ISGID);
 291         if (error) {
 292                 if (ioflag & IO_UNIT) {
 293                         EXT2_TRUNCATE(vp, osize, ioflag & IO_SYNC, ap->a_cred);
 294                         uio->uio_offset -= resid - uio->uio_resid;
 295                         uio->uio_resid = resid;
 296                 }
 297         } else if (resid > uio->uio_resid && (ioflag & IO_SYNC))
 298                 error = EXT2_UPDATE(vp, 1);
 299         return (error);
 300 }