sys/gnu/vfs/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  */
  42
  43 #define BLKSIZE(a, b, c)        blksize(a, b, c)
  44 #define FS                      struct ext2_sb_info
  45 #define I_FS                    i_e2fs
  46
  47 /*
  48  * Vnode op for reading.
  49  *
  50  * ext2_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
  51  *           struct ucred *a_cred)
  52  */
  53 /* ARGSUSED */
  54 static int
  55 ext2_read(struct vop_read_args *ap)
  56 {
  57         struct vnode *vp;
  58         struct inode *ip;
  59         struct uio *uio;
  60         FS *fs;
  61         struct buf *bp;
  62         daddr_t lbn, nextlbn;
  63         off_t nextloffset;
  64         off_t bytesinfile;
  65         long size, xfersize, blkoffset;
  66         int error, orig_resid;
  67         int seqcount = ap->a_ioflag >> 16;
  68
  69         vp = ap->a_vp;
  70         ip = VTOI(vp);
  71         uio = ap->a_uio;
  72
  73 #ifdef DIAGNOSTIC
  74         if (uio->uio_rw != UIO_READ)
  75                 panic("ext2_read: mode");
  76
  77         if (vp->v_type == VLNK) {
  78                 if ((int)ip->i_size < vp->v_mount->mnt_maxsymlinklen)
  79                         panic("ext2_read: short symlink");
  80         } else if (vp->v_type != VREG && vp->v_type != VDIR)
  81                 panic("ext2_read: type %d", vp->v_type);
  82 #endif
  83         fs = ip->I_FS;
  84 #if 0
  85         if ((u_quad_t)uio->uio_offset > fs->fs_maxfilesize)
  86                 return (EFBIG);
  87 #endif
  88
  89         orig_resid = uio->uio_resid;
  90         for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
  91                 if ((bytesinfile = ip->i_size - uio->uio_offset) <= 0)
  92                         break;
  93                 lbn = lblkno(fs, uio->uio_offset);
  94                 nextlbn = lbn + 1;
  95                 nextloffset = lblktodoff(fs, nextlbn);
  96                 size = BLKSIZE(fs, ip, lbn);
  97                 blkoffset = blkoff(fs, uio->uio_offset);
  98
  99                 xfersize = fs->s_frag_size - blkoffset;
 100                 if (uio->uio_resid < xfersize)
 101                         xfersize = uio->uio_resid;
 102                 if (bytesinfile < xfersize)
 103                         xfersize = bytesinfile;
 104
 105                 if (nextloffset >= ip->i_size) {
 106                         error = bread(vp, lblktodoff(fs, lbn), size, &bp);
 107                 } else if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERR) == 0) {
 108                         error = cluster_read(vp, (off_t)ip->i_size,
 109                                              lblktodoff(fs, lbn), size,
 110                                              uio->uio_resid,
 111                                              (ap->a_ioflag >> 16) * BKVASIZE,
 112                                              &bp);
 113                 } else if (seqcount > 1) {
 114                         int nextsize = BLKSIZE(fs, ip, nextlbn);
 115                         error = breadn(vp, lblktodoff(fs, lbn),
 116                                         size, &nextloffset, &nextsize, 1, &bp);
 117                 } else {
 118                         error = bread(vp, lblktodoff(fs, lbn), size, &bp);
 119                 }
 120                 if (error) {
 121                         brelse(bp);
 122                         bp = NULL;
 123                         break;
 124                 }
 125
 126                 /*
 127                  * We should only get non-zero b_resid when an I/O error
 128                  * has occurred, which should cause us to break above.
 129                  * However, if the short read did not cause an error,
 130                  * then we want to ensure that we do not uiomove bad
 131                  * or uninitialized data.
 132                  */
 133                 size -= bp->b_resid;
 134                 if (size < xfersize) {
 135                         if (size == 0)
 136                                 break;
 137                         xfersize = size;
 138                 }
 139                 error = uiomove((char *)bp->b_data + blkoffset,
 140                                 (int)xfersize, uio);
 141                 if (error)
 142                         break;
 143
 144                 bqrelse(bp);
 145         }
 146         if (bp != NULL)
 147                 bqrelse(bp);
 148         if (orig_resid > 0 && (error == 0 || uio->uio_resid != orig_resid) &&
 149             (vp->v_mount->mnt_flag & MNT_NOATIME) == 0)
 150                 ip->i_flag |= IN_ACCESS;
 151         return (error);
 152 }
 153
 154 /*
 155  * Vnode op for writing.
 156  *
 157  * ext2_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
 158  *            struct ucred *a_cred)
 159  */
 160 static int
 161 ext2_write(struct vop_write_args *ap)
 162 {
 163         struct vnode *vp;
 164         struct uio *uio;
 165         struct inode *ip;
 166         FS *fs;
 167         struct buf *bp;
 168         struct thread *td;
 169         daddr_t lbn;
 170         off_t osize;
 171         int seqcount;
 172         int blkoffset, error, flags, ioflag, resid, size, xfersize;
 173
 174         ioflag = ap->a_ioflag;
 175         seqcount = ap->a_ioflag >> 16;
 176         uio = ap->a_uio;
 177         vp = ap->a_vp;
 178         ip = VTOI(vp);
 179
 180 #ifdef DIAGNOSTIC
 181         if (uio->uio_rw != UIO_WRITE)
 182                 panic("ext2_write: mode");
 183 #endif
 184
 185         switch (vp->v_type) {
 186         case VREG:
 187                 if (ioflag & IO_APPEND)
 188                         uio->uio_offset = ip->i_size;
 189                 if ((ip->i_flags & APPEND) && uio->uio_offset != ip->i_size)
 190                         return (EPERM);
 191                 /* FALLTHROUGH */
 192         case VLNK:
 193                 break;
 194         case VDIR:
 195                 if ((ioflag & IO_SYNC) == 0)
 196                         panic("ext2_write: nonsync dir write");
 197                 break;
 198         default:
 199                 panic("ext2_write: type");
 200         }
 201
 202         fs = ip->I_FS;
 203 #if 0
 204         if (uio->uio_offset < 0 ||
 205             (u_quad_t)uio->uio_offset + uio->uio_resid > fs->fs_maxfilesize)
 206                 return (EFBIG);
 207 #endif
 208         /*
 209          * Maybe this should be above the vnode op call, but so long as
 210          * file servers have no limits, I don't think it matters.
 211          */
 212         td = uio->uio_td;
 213         if (vp->v_type == VREG && td && td->td_proc &&
 214             uio->uio_offset + uio->uio_resid >
 215             td->td_proc->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
 216                 lwpsignal(td->td_proc, td->td_lwp, SIGXFSZ);
 217                 return (EFBIG);
 218         }
 219
 220         resid = uio->uio_resid;
 221         osize = ip->i_size;
 222         flags = ioflag & IO_SYNC ? B_SYNC : 0;
 223
 224         for (error = 0; uio->uio_resid > 0;) {
 225                 lbn = lblkno(fs, uio->uio_offset);
 226                 blkoffset = blkoff(fs, uio->uio_offset);
 227                 xfersize = fs->s_frag_size - blkoffset;
 228                 if (uio->uio_resid < xfersize)
 229                         xfersize = uio->uio_resid;
 230
 231                 if (uio->uio_offset + xfersize > ip->i_size)
 232                         vnode_pager_setsize(vp, uio->uio_offset + xfersize);
 233
 234                 /*
 235                  * Avoid a data-consistency race between write() and mmap()
 236                  * by ensuring that newly allocated blocks are zerod.  The
 237                  * race can occur even in the case where the write covers
 238                  * the entire block.
 239                  *
 240                  * Just set B_CLRBUF unconditionally, even if the write
 241                  * covers the whole block.  This also handles the UIO_NOCOPY
 242                  * case.  See ffs_write() in ufs for a more sophisticated
 243                  * version.
 244                  */
 245                 flags |= B_CLRBUF;
 246                 error = ext2_balloc(ip, lbn, blkoffset + xfersize,
 247                                     ap->a_cred, &bp, flags);
 248                 if (error)
 249                         break;
 250
 251                 if (uio->uio_offset + xfersize > ip->i_size) {
 252                         ip->i_size = uio->uio_offset + xfersize;
 253                 }
 254
 255                 size = BLKSIZE(fs, ip, lbn) - bp->b_resid;
 256                 if (size < xfersize)
 257                         xfersize = size;
 258
 259                 error = uiomove((char *)bp->b_data + blkoffset,
 260                                 (int)xfersize, uio);
 261                 if ((ioflag & IO_VMIO) &&
 262                     LIST_FIRST(&bp->b_dep) == NULL) /* in ext2fs? */
 263                         bp->b_flags |= B_RELBUF;
 264
 265                 if (ioflag & IO_SYNC) {
 266                         bwrite(bp);
 267                 } else if (xfersize + blkoffset == fs->s_frag_size) {
 268                         if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERW) == 0) {
 269                                 bp->b_flags |= B_CLUSTEROK;
 270                                 cluster_write(bp, (off_t)ip->i_size, vp->v_mount->mnt_stat.f_iosize, seqcount);
 271                         } else {
 272                                 bawrite(bp);
 273                         }
 274                 } else {
 275                         bp->b_flags |= B_CLUSTEROK;
 276                         bdwrite(bp);
 277                 }
 278                 if (error || xfersize == 0)
 279                         break;
 280                 ip->i_flag |= IN_CHANGE | IN_UPDATE;
 281         }
 282         /*
 283          * If we successfully wrote any data, and we are not the superuser
 284          * we clear the setuid and setgid bits as a precaution against
 285          * tampering.
 286          */
 287         if (resid > uio->uio_resid && ap->a_cred && ap->a_cred->cr_uid != 0)
 288                 ip->i_mode &= ~(ISUID | ISGID);
 289         if (error) {
 290                 if (ioflag & IO_UNIT) {
 291                         EXT2_TRUNCATE(vp, osize, ioflag & IO_SYNC, ap->a_cred);
 292                         uio->uio_offset -= resid - uio->uio_resid;
 293                         uio->uio_resid = resid;
 294                 }
 295         } else if (resid > uio->uio_resid && (ioflag & IO_SYNC))
 296                 error = EXT2_UPDATE(vp, 1);
 297         return (error);
 298 }