Home | History | Annotate | Line # | Download | only in ext2fs
ext2fs_readwrite.c revision 1.38.2.5
      1 /*	$NetBSD: ext2fs_readwrite.c,v 1.38.2.5 2005/11/19 17:37:00 yamt Exp $	*/
      2 
      3 /*-
      4  * Copyright (c) 1993
      5  *	The Regents of the University of California.  All rights reserved.
      6  *
      7  * Redistribution and use in source and binary forms, with or without
      8  * modification, are permitted provided that the following conditions
      9  * are met:
     10  * 1. Redistributions of source code must retain the above copyright
     11  *    notice, this list of conditions and the following disclaimer.
     12  * 2. Redistributions in binary form must reproduce the above copyright
     13  *    notice, this list of conditions and the following disclaimer in the
     14  *    documentation and/or other materials provided with the distribution.
     15  * 3. Neither the name of the University nor the names of its contributors
     16  *    may be used to endorse or promote products derived from this software
     17  *    without specific prior written permission.
     18  *
     19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     29  * SUCH DAMAGE.
     30  *
     31  *	@(#)ufs_readwrite.c	8.8 (Berkeley) 8/4/94
     32  * Modified for ext2fs by Manuel Bouyer.
     33  */
     34 
     35 /*-
     36  * Copyright (c) 1997 Manuel Bouyer.
     37  *
     38  * Redistribution and use in source and binary forms, with or without
     39  * modification, are permitted provided that the following conditions
     40  * are met:
     41  * 1. Redistributions of source code must retain the above copyright
     42  *    notice, this list of conditions and the following disclaimer.
     43  * 2. Redistributions in binary form must reproduce the above copyright
     44  *    notice, this list of conditions and the following disclaimer in the
     45  *    documentation and/or other materials provided with the distribution.
     46  * 3. All advertising materials mentioning features or use of this software
     47  *    must display the following acknowledgement:
     48  *	This product includes software developed by Manuel Bouyer.
     49  * 4. The name of the author may not be used to endorse or promote products
     50  *    derived from this software without specific prior written permission.
     51  *
     52  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
     53  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
     54  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
     55  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
     56  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
     57  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
     58  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
     59  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
     60  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
     61  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
     62  *
     63  *	@(#)ufs_readwrite.c	8.8 (Berkeley) 8/4/94
     64  * Modified for ext2fs by Manuel Bouyer.
     65  */
     66 
     67 #include <sys/cdefs.h>
     68 __KERNEL_RCSID(0, "$NetBSD: ext2fs_readwrite.c,v 1.38.2.5 2005/11/19 17:37:00 yamt Exp $");
     69 
     70 #include <sys/param.h>
     71 #include <sys/systm.h>
     72 #include <sys/resourcevar.h>
     73 #include <sys/kernel.h>
     74 #include <sys/file.h>
     75 #include <sys/stat.h>
     76 #include <sys/buf.h>
     77 #include <sys/proc.h>
     78 #include <sys/mount.h>
     79 #include <sys/vnode.h>
     80 #include <sys/malloc.h>
     81 #include <sys/signalvar.h>
     82 
     83 #include <ufs/ufs/inode.h>
     84 #include <ufs/ufs/ufsmount.h>
     85 #include <ufs/ufs/ufs_extern.h>
     86 #include <ufs/ext2fs/ext2fs.h>
     87 #include <ufs/ext2fs/ext2fs_extern.h>
     88 
     89 #define doclusterread 0 /* XXX underway */
     90 #define doclusterwrite 0
     91 
     92 /*
     93  * Vnode op for reading.
     94  */
     95 /* ARGSUSED */
     96 int
     97 ext2fs_read(void *v)
     98 {
     99 	struct vop_read_args /* {
    100 		struct vnode *a_vp;
    101 		struct uio *a_uio;
    102 		int a_ioflag;
    103 		struct ucred *a_cred;
    104 	} */ *ap = v;
    105 	struct vnode *vp;
    106 	struct inode *ip;
    107 	struct uio *uio;
    108 	struct m_ext2fs *fs;
    109 	struct buf *bp;
    110 	struct ufsmount *ump;
    111 	void *win;
    112 	vsize_t bytelen;
    113 	daddr_t lbn, nextlbn;
    114 	off_t bytesinfile;
    115 	long size, xfersize, blkoffset;
    116 	int error, flags;
    117 
    118 	vp = ap->a_vp;
    119 	ip = VTOI(vp);
    120 	ump = ip->i_ump;
    121 	uio = ap->a_uio;
    122 	error = 0;
    123 
    124 #ifdef DIAGNOSTIC
    125 	if (uio->uio_rw != UIO_READ)
    126 		panic("%s: mode", "ext2fs_read");
    127 
    128 	if (vp->v_type == VLNK) {
    129 		if (ext2fs_size(ip) < ump->um_maxsymlinklen ||
    130 		    (ump->um_maxsymlinklen == 0 && ip->i_e2fs_nblock == 0))
    131 			panic("%s: short symlink", "ext2fs_read");
    132 	} else if (vp->v_type != VREG && vp->v_type != VDIR)
    133 		panic("%s: type %d", "ext2fs_read", vp->v_type);
    134 #endif
    135 	fs = ip->i_e2fs;
    136 	if ((u_int64_t)uio->uio_offset > ump->um_maxfilesize)
    137 		return (EFBIG);
    138 	if (uio->uio_resid == 0)
    139 		return (0);
    140 	if (uio->uio_offset >= ext2fs_size(ip))
    141 		goto out;
    142 
    143 	if (vp->v_type == VREG) {
    144 		const int advice = IO_ADV_DECODE(ap->a_ioflag);
    145 
    146 		while (uio->uio_resid > 0) {
    147 			bytelen = MIN(ext2fs_size(ip) - uio->uio_offset,
    148 			    uio->uio_resid);
    149 			if (bytelen == 0)
    150 				break;
    151 
    152 			win = ubc_alloc(&vp->v_uobj, uio->uio_offset,
    153 			    &bytelen, advice, UBC_READ);
    154 			error = uiomove(win, bytelen, uio);
    155 			flags = UBC_WANT_UNMAP(vp) ? UBC_UNMAP : 0;
    156 			ubc_release(win, flags);
    157 			if (error)
    158 				break;
    159 		}
    160 		goto out;
    161 	}
    162 
    163 	for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
    164 		bytesinfile = ext2fs_size(ip) - uio->uio_offset;
    165 		if (bytesinfile <= 0)
    166 			break;
    167 		lbn = lblkno(fs, uio->uio_offset);
    168 		nextlbn = lbn + 1;
    169 		size = fs->e2fs_bsize;
    170 		blkoffset = blkoff(fs, uio->uio_offset);
    171 		xfersize = fs->e2fs_bsize - blkoffset;
    172 		if (uio->uio_resid < xfersize)
    173 			xfersize = uio->uio_resid;
    174 		if (bytesinfile < xfersize)
    175 			xfersize = bytesinfile;
    176 
    177 		if (lblktosize(fs, nextlbn) >= ext2fs_size(ip))
    178 			error = bread(vp, lbn, size, NOCRED, &bp);
    179 		else {
    180 			int nextsize = fs->e2fs_bsize;
    181 			error = breadn(vp, lbn,
    182 				size, &nextlbn, &nextsize, 1, NOCRED, &bp);
    183 		}
    184 		if (error)
    185 			break;
    186 
    187 		/*
    188 		 * We should only get non-zero b_resid when an I/O error
    189 		 * has occurred, which should cause us to break above.
    190 		 * However, if the short read did not cause an error,
    191 		 * then we want to ensure that we do not uiomove bad
    192 		 * or uninitialized data.
    193 		 */
    194 		size -= bp->b_resid;
    195 		if (size < xfersize) {
    196 			if (size == 0)
    197 				break;
    198 			xfersize = size;
    199 		}
    200 		error = uiomove((char *)bp->b_data + blkoffset, xfersize, uio);
    201 		if (error)
    202 			break;
    203 		brelse(bp);
    204 	}
    205 	if (bp != NULL)
    206 		brelse(bp);
    207 
    208 out:
    209 	if (!(vp->v_mount->mnt_flag & MNT_NOATIME)) {
    210 		ip->i_flag |= IN_ACCESS;
    211 		if ((ap->a_ioflag & IO_SYNC) == IO_SYNC)
    212 			error = ext2fs_update(vp, NULL, NULL, UPDATE_WAIT);
    213 	}
    214 	return (error);
    215 }
    216 
    217 /*
    218  * Vnode op for writing.
    219  */
    220 int
    221 ext2fs_write(void *v)
    222 {
    223 	struct vop_write_args /* {
    224 		struct vnode *a_vp;
    225 		struct uio *a_uio;
    226 		int a_ioflag;
    227 		struct ucred *a_cred;
    228 	} */ *ap = v;
    229 	struct vnode *vp;
    230 	struct uio *uio;
    231 	struct inode *ip;
    232 	struct m_ext2fs *fs;
    233 	struct buf *bp;
    234 	struct proc *p;
    235 	struct ufsmount *ump;
    236 	daddr_t lbn;
    237 	off_t osize;
    238 	int blkoffset, error, flags, ioflag, resid, xfersize;
    239 	vsize_t bytelen;
    240 	void *win;
    241 	off_t oldoff = 0;					/* XXX */
    242 	boolean_t async;
    243 	int extended = 0;
    244 
    245 	ioflag = ap->a_ioflag;
    246 	uio = ap->a_uio;
    247 	vp = ap->a_vp;
    248 	ip = VTOI(vp);
    249 	ump = ip->i_ump;
    250 	error = 0;
    251 
    252 #ifdef DIAGNOSTIC
    253 	if (uio->uio_rw != UIO_WRITE)
    254 		panic("%s: mode", "ext2fs_write");
    255 #endif
    256 
    257 	switch (vp->v_type) {
    258 	case VREG:
    259 		if (ioflag & IO_APPEND)
    260 			uio->uio_offset = ext2fs_size(ip);
    261 		if ((ip->i_e2fs_flags & EXT2_APPEND) &&
    262 		    uio->uio_offset != ext2fs_size(ip))
    263 			return (EPERM);
    264 		/* FALLTHROUGH */
    265 	case VLNK:
    266 		break;
    267 	case VDIR:
    268 		if ((ioflag & IO_SYNC) == 0)
    269 			panic("%s: nonsync dir write", "ext2fs_write");
    270 		break;
    271 	default:
    272 		panic("%s: type", "ext2fs_write");
    273 	}
    274 
    275 	fs = ip->i_e2fs;
    276 	if (uio->uio_offset < 0 ||
    277 	    (u_int64_t)uio->uio_offset + uio->uio_resid > ump->um_maxfilesize)
    278 		return (EFBIG);
    279 	/*
    280 	 * Maybe this should be above the vnode op call, but so long as
    281 	 * file servers have no limits, I don't think it matters.
    282 	 */
    283 	p = uio->uio_procp;
    284 	if (vp->v_type == VREG && p &&
    285 	    uio->uio_offset + uio->uio_resid >
    286 	    p->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
    287 		psignal(p, SIGXFSZ);
    288 		return (EFBIG);
    289 	}
    290 	if (uio->uio_resid == 0)
    291 		return (0);
    292 
    293 	async = vp->v_mount->mnt_flag & MNT_ASYNC;
    294 	resid = uio->uio_resid;
    295 	osize = ext2fs_size(ip);
    296 
    297 	if (vp->v_type == VREG) {
    298 		while (uio->uio_resid > 0) {
    299 			oldoff = uio->uio_offset;
    300 			blkoffset = blkoff(fs, uio->uio_offset);
    301 			bytelen = MIN(fs->e2fs_bsize - blkoffset,
    302 			    uio->uio_resid);
    303 
    304 			error = ufs_balloc_range(vp, uio->uio_offset,
    305 			    bytelen, ap->a_cred, 0);
    306 			if (error)
    307 				break;
    308 			win = ubc_alloc(&vp->v_uobj, uio->uio_offset,
    309 			    &bytelen, UVM_ADV_NORMAL, UBC_WRITE);
    310 			error = uiomove(win, bytelen, uio);
    311 			flags = UBC_WANT_UNMAP(vp) ? UBC_UNMAP : 0;
    312 			ubc_release(win, flags);
    313 			if (error)
    314 				break;
    315 
    316 			/*
    317 			 * update UVM's notion of the size now that we've
    318 			 * copied the data into the vnode's pages.
    319 			 */
    320 
    321 			if (vp->v_size < uio->uio_offset) {
    322 				uvm_vnp_setsize(vp, uio->uio_offset);
    323 				extended = 1;
    324 			}
    325 
    326 			/*
    327 			 * flush what we just wrote if necessary.
    328 			 * XXXUBC simplistic async flushing.
    329 			 */
    330 
    331 			if (!async && oldoff >> 16 != uio->uio_offset >> 16) {
    332 				simple_lock(&vp->v_interlock);
    333 				error = VOP_PUTPAGES(vp, (oldoff >> 16) << 16,
    334 				    (uio->uio_offset >> 16) << 16, PGO_CLEANIT);
    335 			}
    336 		}
    337 		if (error == 0 && ioflag & IO_SYNC) {
    338 			simple_lock(&vp->v_interlock);
    339 			error = VOP_PUTPAGES(vp, trunc_page(oldoff),
    340 			    round_page(blkroundup(fs, uio->uio_offset)),
    341 			    PGO_CLEANIT | PGO_SYNCIO);
    342 		}
    343 
    344 		goto out;
    345 	}
    346 
    347 	flags = ioflag & IO_SYNC ? B_SYNC : 0;
    348 	for (error = 0; uio->uio_resid > 0;) {
    349 		lbn = lblkno(fs, uio->uio_offset);
    350 		blkoffset = blkoff(fs, uio->uio_offset);
    351 		xfersize = MIN(fs->e2fs_bsize - blkoffset, uio->uio_resid);
    352 		if (xfersize < fs->e2fs_bsize)
    353 			flags |= B_CLRBUF;
    354 		else
    355 			flags &= ~B_CLRBUF;
    356 		error = ext2fs_balloc(ip,
    357 		    lbn, blkoffset + xfersize, ap->a_cred, &bp, flags);
    358 		if (error)
    359 			break;
    360 		if (ext2fs_size(ip) < uio->uio_offset + xfersize) {
    361 			error = ext2fs_setsize(ip, uio->uio_offset + xfersize);
    362 			if (error)
    363 				break;
    364 		}
    365 		error = uiomove((char *)bp->b_data + blkoffset, xfersize, uio);
    366 
    367 		/*
    368 		 * update UVM's notion of the size now that we've
    369 		 * copied the data into the vnode's pages.
    370 		 */
    371 
    372 		if (vp->v_size < uio->uio_offset) {
    373 			uvm_vnp_setsize(vp, uio->uio_offset);
    374 			extended = 1;
    375 		}
    376 
    377 		if (ioflag & IO_SYNC)
    378 			(void)bwrite(bp);
    379 		else if (xfersize + blkoffset == fs->e2fs_bsize)
    380 			bawrite(bp);
    381 		else
    382 			bdwrite(bp);
    383 		if (error || xfersize == 0)
    384 			break;
    385 	}
    386 
    387 	/*
    388 	 * If we successfully wrote any data, and we are not the superuser
    389 	 * we clear the setuid and setgid bits as a precaution against
    390 	 * tampering.
    391 	 */
    392 
    393 out:
    394 	ip->i_flag |= IN_CHANGE | IN_UPDATE;
    395 	if (resid > uio->uio_resid && ap->a_cred && ap->a_cred->cr_uid != 0)
    396 		ip->i_e2fs_mode &= ~(ISUID | ISGID);
    397 	if (resid > uio->uio_resid)
    398 		VN_KNOTE(vp, NOTE_WRITE | (extended ? NOTE_EXTEND : 0));
    399 	if (error) {
    400 		(void) ext2fs_truncate(vp, osize, ioflag & IO_SYNC, ap->a_cred,
    401 		    uio->uio_procp);
    402 		uio->uio_offset -= resid - uio->uio_resid;
    403 		uio->uio_resid = resid;
    404 	} else if (resid > uio->uio_resid && (ioflag & IO_SYNC) == IO_SYNC)
    405 		error = ext2fs_update(vp, NULL, NULL, UPDATE_WAIT);
    406 	KASSERT(vp->v_size == ext2fs_size(ip));
    407 	return (error);
    408 }
    409