Home | History | Annotate | Line # | Download | only in ext2fs
ext2fs_readwrite.c revision 1.38.2.3
      1 /*	$NetBSD: ext2fs_readwrite.c,v 1.38.2.3 2005/11/18 08:44:55 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 1.38.2.3 2005/11/18 08:44:55 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 #include <uvm/uvm_readahead.h>
     90 
     91 #define doclusterread 0 /* XXX underway */
     92 #define doclusterwrite 0
     93 
     94 /*
     95  * Vnode op for reading.
     96  */
     97 /* ARGSUSED */
     98 int
     99 ext2fs_read(void *v)
    100 {
    101 	struct vop_read_args /* {
    102 		struct vnode *a_vp;
    103 		struct uio *a_uio;
    104 		int a_ioflag;
    105 		struct ucred *a_cred;
    106 	} */ *ap = v;
    107 	struct vnode *vp;
    108 	struct inode *ip;
    109 	struct uio *uio;
    110 	struct m_ext2fs *fs;
    111 	struct buf *bp;
    112 	struct ufsmount *ump;
    113 	void *win;
    114 	vsize_t bytelen;
    115 	daddr_t lbn, nextlbn;
    116 	off_t bytesinfile;
    117 	long size, xfersize, blkoffset;
    118 	int error, flags;
    119 	struct uvm_ractx *ra;
    120 
    121 	vp = ap->a_vp;
    122 	ip = VTOI(vp);
    123 	ump = ip->i_ump;
    124 	uio = ap->a_uio;
    125 	error = 0;
    126 
    127 #ifdef DIAGNOSTIC
    128 	if (uio->uio_rw != UIO_READ)
    129 		panic("%s: mode", "ext2fs_read");
    130 
    131 	if (vp->v_type == VLNK) {
    132 		if (ext2fs_size(ip) < ump->um_maxsymlinklen ||
    133 		    (ump->um_maxsymlinklen == 0 && ip->i_e2fs_nblock == 0))
    134 			panic("%s: short symlink", "ext2fs_read");
    135 	} else if (vp->v_type != VREG && vp->v_type != VDIR)
    136 		panic("%s: type %d", "ext2fs_read", vp->v_type);
    137 #endif
    138 	fs = ip->i_e2fs;
    139 	if ((u_int64_t)uio->uio_offset > ump->um_maxfilesize)
    140 		return (EFBIG);
    141 	if (uio->uio_resid == 0)
    142 		return (0);
    143 	if (uio->uio_offset >= ext2fs_size(ip))
    144 		goto out;
    145 
    146 	if (vp->v_type == VREG) {
    147 		const int advice = IO_ADV_DECODE(ap->a_ioflag);
    148 
    149 		while (uio->uio_resid > 0) {
    150 			bytelen = MIN(ext2fs_size(ip) - uio->uio_offset,
    151 			    uio->uio_resid);
    152 			if (bytelen == 0)
    153 				break;
    154 
    155 			win = ubc_alloc(&vp->v_uobj, uio->uio_offset,
    156 			    &bytelen, UBC_READ);
    157 			uvm_ra_request(vp->v_ractx, advice, &vp->v_uobj,
    158 			    uio->uio_offset, bytelen);
    159 			error = uiomove(win, bytelen, uio);
    160 			flags = UBC_WANT_UNMAP(vp) ? UBC_UNMAP : 0;
    161 			ubc_release(win, flags);
    162 			if (error)
    163 				break;
    164 		}
    165 		goto out;
    166 	}
    167 
    168 	for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
    169 		bytesinfile = ext2fs_size(ip) - uio->uio_offset;
    170 		if (bytesinfile <= 0)
    171 			break;
    172 		lbn = lblkno(fs, uio->uio_offset);
    173 		nextlbn = lbn + 1;
    174 		size = fs->e2fs_bsize;
    175 		blkoffset = blkoff(fs, uio->uio_offset);
    176 		xfersize = fs->e2fs_bsize - blkoffset;
    177 		if (uio->uio_resid < xfersize)
    178 			xfersize = uio->uio_resid;
    179 		if (bytesinfile < xfersize)
    180 			xfersize = bytesinfile;
    181 
    182 		if (lblktosize(fs, nextlbn) >= ext2fs_size(ip))
    183 			error = bread(vp, lbn, size, NOCRED, &bp);
    184 		else {
    185 			int nextsize = fs->e2fs_bsize;
    186 			error = breadn(vp, lbn,
    187 				size, &nextlbn, &nextsize, 1, NOCRED, &bp);
    188 		}
    189 		if (error)
    190 			break;
    191 
    192 		/*
    193 		 * We should only get non-zero b_resid when an I/O error
    194 		 * has occurred, which should cause us to break above.
    195 		 * However, if the short read did not cause an error,
    196 		 * then we want to ensure that we do not uiomove bad
    197 		 * or uninitialized data.
    198 		 */
    199 		size -= bp->b_resid;
    200 		if (size < xfersize) {
    201 			if (size == 0)
    202 				break;
    203 			xfersize = size;
    204 		}
    205 		error = uiomove((char *)bp->b_data + blkoffset, xfersize, uio);
    206 		if (error)
    207 			break;
    208 		brelse(bp);
    209 	}
    210 	if (bp != NULL)
    211 		brelse(bp);
    212 
    213 out:
    214 	if (!(vp->v_mount->mnt_flag & MNT_NOATIME)) {
    215 		ip->i_flag |= IN_ACCESS;
    216 		if ((ap->a_ioflag & IO_SYNC) == IO_SYNC)
    217 			error = ext2fs_update(vp, NULL, NULL, UPDATE_WAIT);
    218 	}
    219 	return (error);
    220 }
    221 
    222 /*
    223  * Vnode op for writing.
    224  */
    225 int
    226 ext2fs_write(void *v)
    227 {
    228 	struct vop_write_args /* {
    229 		struct vnode *a_vp;
    230 		struct uio *a_uio;
    231 		int a_ioflag;
    232 		struct ucred *a_cred;
    233 	} */ *ap = v;
    234 	struct vnode *vp;
    235 	struct uio *uio;
    236 	struct inode *ip;
    237 	struct m_ext2fs *fs;
    238 	struct buf *bp;
    239 	struct proc *p;
    240 	struct ufsmount *ump;
    241 	daddr_t lbn;
    242 	off_t osize;
    243 	int blkoffset, error, flags, ioflag, resid, xfersize;
    244 	vsize_t bytelen;
    245 	void *win;
    246 	off_t oldoff = 0;					/* XXX */
    247 	boolean_t async;
    248 	int extended = 0;
    249 
    250 	ioflag = ap->a_ioflag;
    251 	uio = ap->a_uio;
    252 	vp = ap->a_vp;
    253 	ip = VTOI(vp);
    254 	ump = ip->i_ump;
    255 	error = 0;
    256 
    257 #ifdef DIAGNOSTIC
    258 	if (uio->uio_rw != UIO_WRITE)
    259 		panic("%s: mode", "ext2fs_write");
    260 #endif
    261 
    262 	switch (vp->v_type) {
    263 	case VREG:
    264 		if (ioflag & IO_APPEND)
    265 			uio->uio_offset = ext2fs_size(ip);
    266 		if ((ip->i_e2fs_flags & EXT2_APPEND) &&
    267 		    uio->uio_offset != ext2fs_size(ip))
    268 			return (EPERM);
    269 		/* FALLTHROUGH */
    270 	case VLNK:
    271 		break;
    272 	case VDIR:
    273 		if ((ioflag & IO_SYNC) == 0)
    274 			panic("%s: nonsync dir write", "ext2fs_write");
    275 		break;
    276 	default:
    277 		panic("%s: type", "ext2fs_write");
    278 	}
    279 
    280 	fs = ip->i_e2fs;
    281 	if (uio->uio_offset < 0 ||
    282 	    (u_int64_t)uio->uio_offset + uio->uio_resid > ump->um_maxfilesize)
    283 		return (EFBIG);
    284 	/*
    285 	 * Maybe this should be above the vnode op call, but so long as
    286 	 * file servers have no limits, I don't think it matters.
    287 	 */
    288 	p = uio->uio_procp;
    289 	if (vp->v_type == VREG && p &&
    290 	    uio->uio_offset + uio->uio_resid >
    291 	    p->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
    292 		psignal(p, SIGXFSZ);
    293 		return (EFBIG);
    294 	}
    295 	if (uio->uio_resid == 0)
    296 		return (0);
    297 
    298 	async = vp->v_mount->mnt_flag & MNT_ASYNC;
    299 	resid = uio->uio_resid;
    300 	osize = ext2fs_size(ip);
    301 
    302 	if (vp->v_type == VREG) {
    303 		while (uio->uio_resid > 0) {
    304 			oldoff = uio->uio_offset;
    305 			blkoffset = blkoff(fs, uio->uio_offset);
    306 			bytelen = MIN(fs->e2fs_bsize - blkoffset,
    307 			    uio->uio_resid);
    308 
    309 			error = ufs_balloc_range(vp, uio->uio_offset,
    310 			    bytelen, ap->a_cred, 0);
    311 			if (error)
    312 				break;
    313 			win = ubc_alloc(&vp->v_uobj, uio->uio_offset,
    314 			    &bytelen, UBC_WRITE);
    315 			error = uiomove(win, bytelen, uio);
    316 			flags = UBC_WANT_UNMAP(vp) ? UBC_UNMAP : 0;
    317 			ubc_release(win, flags);
    318 			if (error)
    319 				break;
    320 
    321 			/*
    322 			 * update UVM's notion of the size now that we've
    323 			 * copied the data into the vnode's pages.
    324 			 */
    325 
    326 			if (vp->v_size < uio->uio_offset) {
    327 				uvm_vnp_setsize(vp, uio->uio_offset);
    328 				extended = 1;
    329 			}
    330 
    331 			/*
    332 			 * flush what we just wrote if necessary.
    333 			 * XXXUBC simplistic async flushing.
    334 			 */
    335 
    336 			if (!async && oldoff >> 16 != uio->uio_offset >> 16) {
    337 				simple_lock(&vp->v_interlock);
    338 				error = VOP_PUTPAGES(vp, (oldoff >> 16) << 16,
    339 				    (uio->uio_offset >> 16) << 16, PGO_CLEANIT);
    340 			}
    341 		}
    342 		if (error == 0 && ioflag & IO_SYNC) {
    343 			simple_lock(&vp->v_interlock);
    344 			error = VOP_PUTPAGES(vp, trunc_page(oldoff),
    345 			    round_page(blkroundup(fs, uio->uio_offset)),
    346 			    PGO_CLEANIT | PGO_SYNCIO);
    347 		}
    348 
    349 		goto out;
    350 	}
    351 
    352 	flags = ioflag & IO_SYNC ? B_SYNC : 0;
    353 	for (error = 0; uio->uio_resid > 0;) {
    354 		lbn = lblkno(fs, uio->uio_offset);
    355 		blkoffset = blkoff(fs, uio->uio_offset);
    356 		xfersize = MIN(fs->e2fs_bsize - blkoffset, uio->uio_resid);
    357 		if (xfersize < fs->e2fs_bsize)
    358 			flags |= B_CLRBUF;
    359 		else
    360 			flags &= ~B_CLRBUF;
    361 		error = ext2fs_balloc(ip,
    362 		    lbn, blkoffset + xfersize, ap->a_cred, &bp, flags);
    363 		if (error)
    364 			break;
    365 		if (ext2fs_size(ip) < uio->uio_offset + xfersize) {
    366 			error = ext2fs_setsize(ip, uio->uio_offset + xfersize);
    367 			if (error)
    368 				break;
    369 		}
    370 		error = uiomove((char *)bp->b_data + blkoffset, xfersize, uio);
    371 
    372 		/*
    373 		 * update UVM's notion of the size now that we've
    374 		 * copied the data into the vnode's pages.
    375 		 */
    376 
    377 		if (vp->v_size < uio->uio_offset) {
    378 			uvm_vnp_setsize(vp, uio->uio_offset);
    379 			extended = 1;
    380 		}
    381 
    382 		if (ioflag & IO_SYNC)
    383 			(void)bwrite(bp);
    384 		else if (xfersize + blkoffset == fs->e2fs_bsize)
    385 			bawrite(bp);
    386 		else
    387 			bdwrite(bp);
    388 		if (error || xfersize == 0)
    389 			break;
    390 	}
    391 
    392 	/*
    393 	 * If we successfully wrote any data, and we are not the superuser
    394 	 * we clear the setuid and setgid bits as a precaution against
    395 	 * tampering.
    396 	 */
    397 
    398 out:
    399 	ip->i_flag |= IN_CHANGE | IN_UPDATE;
    400 	if (resid > uio->uio_resid && ap->a_cred && ap->a_cred->cr_uid != 0)
    401 		ip->i_e2fs_mode &= ~(ISUID | ISGID);
    402 	if (resid > uio->uio_resid)
    403 		VN_KNOTE(vp, NOTE_WRITE | (extended ? NOTE_EXTEND : 0));
    404 	if (error) {
    405 		(void) ext2fs_truncate(vp, osize, ioflag & IO_SYNC, ap->a_cred,
    406 		    uio->uio_procp);
    407 		uio->uio_offset -= resid - uio->uio_resid;
    408 		uio->uio_resid = resid;
    409 	} else if (resid > uio->uio_resid && (ioflag & IO_SYNC) == IO_SYNC)
    410 		error = ext2fs_update(vp, NULL, NULL, UPDATE_WAIT);
    411 	KASSERT(vp->v_size == ext2fs_size(ip));
    412 	return (error);
    413 }
    414