xref: /openbsd/sys/ufs/ext2fs/ext2fs_readwrite.c (revision 76d0caae)
1 /*	$OpenBSD: ext2fs_readwrite.c,v 1.45 2019/11/27 16:12:13 beck Exp $	*/
2 /*	$NetBSD: ext2fs_readwrite.c,v 1.16 2001/02/27 04:37:47 chs Exp $	*/
3 
4 /*-
5  * Copyright (c) 1997 Manuel Bouyer.
6  * Copyright (c) 1993
7  *	The Regents of the University of California.  All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)ufs_readwrite.c	8.8 (Berkeley) 8/4/94
34  * Modified for ext2fs by Manuel Bouyer.
35  */
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/resourcevar.h>
40 #include <sys/kernel.h>
41 #include <sys/stat.h>
42 #include <sys/buf.h>
43 #include <sys/mount.h>
44 #include <sys/vnode.h>
45 #include <sys/malloc.h>
46 #include <sys/signalvar.h>
47 #include <sys/event.h>
48 
49 #include <ufs/ufs/quota.h>
50 #include <ufs/ufs/ufsmount.h>
51 #include <ufs/ufs/inode.h>
52 #include <ufs/ext2fs/ext2fs.h>
53 #include <ufs/ext2fs/ext2fs_extern.h>
54 
55 
56 static int	ext2_ind_read(struct vnode *, struct inode *, struct m_ext2fs *, struct uio *);
57 static int	ext4_ext_read(struct vnode *, struct inode *, struct m_ext2fs *, struct uio *);
58 
59 /*
60  * Vnode op for reading.
61  */
62 /* ARGSUSED */
63 int
64 ext2fs_read(void *v)
65 {
66 	struct vop_read_args *ap = v;
67 	struct vnode *vp;
68 	struct inode *ip;
69 	struct uio *uio;
70 	struct m_ext2fs *fs;
71 
72 	vp = ap->a_vp;
73 	ip = VTOI(vp);
74 	uio = ap->a_uio;
75 	fs = ip->i_e2fs;
76 
77 	if (ip->i_e2fs_flags & EXT4_EXTENTS)
78 		return ext4_ext_read(vp, ip, fs, uio);
79 	else
80 		return ext2_ind_read(vp, ip, fs, uio);
81 }
82 
83 static int
84 ext2_ind_read(struct vnode *vp, struct inode *ip, struct m_ext2fs *fs,
85     struct uio *uio)
86 {
87 	struct buf *bp;
88 	daddr_t lbn, nextlbn;
89 	off_t bytesinfile;
90 	int size, xfersize, blkoffset;
91 	int error;
92 
93 #ifdef DIAGNOSTIC
94 	if (uio->uio_rw != UIO_READ)
95 		panic("%s: mode", "ext2fs_read");
96 
97 	if (vp->v_type == VLNK) {
98 		if (ext2fs_size(ip) < EXT2_MAXSYMLINKLEN)
99 			panic("%s: short symlink", "ext2fs_read");
100 	} else if (vp->v_type != VREG && vp->v_type != VDIR)
101 		panic("%s: type %d", "ext2fs_read", vp->v_type);
102 #endif
103 	if (uio->uio_offset < 0)
104 		return (EINVAL);
105 	if (uio->uio_resid == 0)
106 		return (0);
107 
108 	for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
109 		if ((bytesinfile = ext2fs_size(ip) - uio->uio_offset) <= 0)
110 			break;
111 		lbn = lblkno(fs, uio->uio_offset);
112 		nextlbn = lbn + 1;
113 		size = fs->e2fs_bsize;
114 		blkoffset = blkoff(fs, uio->uio_offset);
115 		xfersize = fs->e2fs_bsize - blkoffset;
116 		if (uio->uio_resid < xfersize)
117 			xfersize = uio->uio_resid;
118 		if (bytesinfile < xfersize)
119 			xfersize = bytesinfile;
120 
121 		if (lblktosize(fs, nextlbn) >= ext2fs_size(ip))
122 			error = bread(vp, lbn, size, &bp);
123 		else if (lbn - 1 == ip->i_ci.ci_lastr) {
124 			int nextsize = fs->e2fs_bsize;
125 			error = breadn(vp, lbn, size, &nextlbn, &nextsize,
126 			    1, &bp);
127 		} else
128 			error = bread(vp, lbn, size, &bp);
129 		if (error)
130 			break;
131 		ip->i_ci.ci_lastr = lbn;
132 
133 		/*
134 		 * We should only get non-zero b_resid when an I/O error
135 		 * has occurred, which should cause us to break above.
136 		 * However, if the short read did not cause an error,
137 		 * then we want to ensure that we do not uiomove bad
138 		 * or uninitialized data.
139 		 */
140 		size -= bp->b_resid;
141 		if (size < xfersize) {
142 			if (size == 0)
143 				break;
144 			xfersize = size;
145 		}
146 		error = uiomove((char *)bp->b_data + blkoffset, xfersize, uio);
147 		if (error)
148 			break;
149 		brelse(bp);
150 	}
151 	if (bp != NULL)
152 		brelse(bp);
153 
154 	if (!(vp->v_mount->mnt_flag & MNT_NOATIME)) {
155 		ip->i_flag |= IN_ACCESS;
156 	}
157 	return (error);
158 }
159 
160 int
161 ext4_ext_read(struct vnode *vp, struct inode *ip, struct m_ext2fs *fs, struct uio *uio)
162 {
163 	struct ext4_extent_path path;
164 	struct ext4_extent nex, *ep;
165 	struct buf *bp;
166 	daddr_t lbn, pos;
167 	off_t bytesinfile;
168 	int size, xfersize, blkoffset;
169 	int error, cache_type;
170 
171 	memset(&path, 0, sizeof path);
172 
173 	if (uio->uio_offset < 0)
174 		return (EINVAL);
175 	if (uio->uio_resid == 0)
176 		return (0);
177 
178 	while (uio->uio_resid > 0) {
179 		if ((bytesinfile = ext2fs_size(ip) - uio->uio_offset) <= 0)
180 			break;
181 		lbn = lblkno(fs, uio->uio_offset);
182 		size = fs->e2fs_bsize;
183 		blkoffset = blkoff(fs, uio->uio_offset);
184 
185 		xfersize = fs->e2fs_fsize - blkoffset;
186 		xfersize = MIN(xfersize, uio->uio_resid);
187 		xfersize = MIN(xfersize, bytesinfile);
188 
189 		cache_type = ext4_ext_in_cache(ip, lbn, &nex);
190 		switch (cache_type) {
191 		case EXT4_EXT_CACHE_NO:
192 			ext4_ext_find_extent(fs, ip, lbn, &path);
193 			if ((ep = path.ep_ext) == NULL)
194 				return (EIO);
195 			ext4_ext_put_cache(ip, ep, EXT4_EXT_CACHE_IN);
196 
197 			pos = lbn - ep->e_blk + (((daddr_t) ep->e_start_hi << 32) | ep->e_start_lo);
198 			if (path.ep_bp != NULL) {
199 				brelse(path.ep_bp);
200 				path.ep_bp = NULL;
201 			}
202 			break;
203 		case EXT4_EXT_CACHE_GAP:
204 			/* block has not been allocated yet */
205 			return (0);
206 		case EXT4_EXT_CACHE_IN:
207 			pos = lbn - nex.e_blk + (((daddr_t) nex.e_start_hi << 32) | nex.e_start_lo);
208 			break;
209 		}
210 		error = bread(ip->i_devvp, fsbtodb(fs, pos), size, &bp);
211 		if (error) {
212 			brelse(bp);
213 			return (error);
214 		}
215 		size -= bp->b_resid;
216 		if (size < xfersize) {
217 			if (size == 0) {
218 				brelse(bp);
219 				break;
220 			}
221 			xfersize = size;
222 		}
223 		error = uiomove(bp->b_data + blkoffset, xfersize, uio);
224 		brelse(bp);
225 		if (error)
226 			return (error);
227 	}
228 	return (0);
229 }
230 
231 /*
232  * Vnode op for writing.
233  */
234 int
235 ext2fs_write(void *v)
236 {
237 	struct vop_write_args *ap = v;
238 	struct vnode *vp;
239 	struct uio *uio;
240 	struct inode *ip;
241 	struct m_ext2fs *fs;
242 	struct buf *bp;
243 	int32_t lbn;
244 	off_t osize;
245 	int blkoffset, error, extended, flags, ioflag, size, xfersize;
246 	size_t resid;
247 	ssize_t overrun;
248 
249 	extended = 0;
250 	ioflag = ap->a_ioflag;
251 	uio = ap->a_uio;
252 	vp = ap->a_vp;
253 	ip = VTOI(vp);
254 
255 #ifdef DIAGNOSTIC
256 	if (uio->uio_rw != UIO_WRITE)
257 		panic("%s: mode", "ext2fs_write");
258 #endif
259 
260 	/*
261 	 * If writing 0 bytes, succeed and do not change
262 	 * update time or file offset (standards compliance)
263 	 */
264 	if (uio->uio_resid == 0)
265 		return (0);
266 
267 	switch (vp->v_type) {
268 	case VREG:
269 		if (ioflag & IO_APPEND)
270 			uio->uio_offset = ext2fs_size(ip);
271 		if ((ip->i_e2fs_flags & EXT2_APPEND) &&
272 			uio->uio_offset != ext2fs_size(ip))
273 			return (EPERM);
274 		/* FALLTHROUGH */
275 	case VLNK:
276 		break;
277 	case VDIR:
278 		if ((ioflag & IO_SYNC) == 0)
279 			panic("%s: nonsync dir write", "ext2fs_write");
280 		break;
281 	default:
282 		panic("%s: type", "ext2fs_write");
283 	}
284 
285 	fs = ip->i_e2fs;
286 	if (e2fs_overflow(fs, uio->uio_resid, uio->uio_offset + uio->uio_resid))
287 		return (EFBIG);
288 
289 	/* do the filesize rlimit check */
290 	if ((error = vn_fsizechk(vp, uio, ioflag, &overrun)))
291 		return (error);
292 
293 	resid = uio->uio_resid;
294 	osize = ext2fs_size(ip);
295 	flags = ioflag & IO_SYNC ? B_SYNC : 0;
296 
297 	for (error = 0; uio->uio_resid > 0;) {
298 		lbn = lblkno(fs, uio->uio_offset);
299 		blkoffset = blkoff(fs, uio->uio_offset);
300 		xfersize = fs->e2fs_bsize - blkoffset;
301 		if (uio->uio_resid < xfersize)
302 			xfersize = uio->uio_resid;
303 		if (fs->e2fs_bsize > xfersize)
304 			flags |= B_CLRBUF;
305 		else
306 			flags &= ~B_CLRBUF;
307 
308 		error = ext2fs_buf_alloc(ip,
309 			lbn, blkoffset + xfersize, ap->a_cred, &bp, flags);
310 		if (error)
311 			break;
312 		if (uio->uio_offset + xfersize > ext2fs_size(ip)) {
313 			error = ext2fs_setsize(ip, uio->uio_offset + xfersize);
314 			if (error)
315 				break;
316 			uvm_vnp_setsize(vp, ext2fs_size(ip));
317 			extended = 1;
318 		}
319 		uvm_vnp_uncache(vp);
320 
321 		size = fs->e2fs_bsize - bp->b_resid;
322 		if (size < xfersize)
323 			xfersize = size;
324 
325 		error = uiomove(bp->b_data + blkoffset, xfersize, uio);
326 		/*
327 		 * If the buffer is not already filled and we encounter an
328 		 * error while trying to fill it, we have to clear out any
329 		 * garbage data from the pages instantiated for the buffer.
330 		 * If we do not, a failed uiomove() during a write can leave
331 		 * the prior contents of the pages exposed to a userland mmap.
332 		 *
333 		 * Note that we don't need to clear buffers that were
334 		 * allocated with the B_CLRBUF flag set.
335 		 */
336 		if (error != 0 && !(flags & B_CLRBUF))
337 			memset(bp->b_data + blkoffset, 0, xfersize);
338 
339 		if (ioflag & IO_NOCACHE)
340 			bp->b_flags |= B_NOCACHE;
341 
342 		if (ioflag & IO_SYNC)
343 			(void)bwrite(bp);
344 		else if (xfersize + blkoffset == fs->e2fs_bsize)
345 			bawrite(bp);
346 		else
347 			bdwrite(bp);
348 		if (error || xfersize == 0)
349 			break;
350 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
351 	}
352 	/*
353 	 * If we successfully wrote any data, and we are not the superuser
354 	 * we clear the setuid and setgid bits as a precaution against
355 	 * tampering.
356 	 */
357 	if (resid > uio->uio_resid && ap->a_cred && ap->a_cred->cr_uid != 0)
358 		ip->i_e2fs_mode &= ~(ISUID | ISGID);
359 	if (resid > uio->uio_resid)
360 		VN_KNOTE(vp, NOTE_WRITE | (extended ? NOTE_EXTEND : 0));
361 	if (error) {
362 		if (ioflag & IO_UNIT) {
363 			(void)ext2fs_truncate(ip, osize,
364 				ioflag & IO_SYNC, ap->a_cred);
365 			uio->uio_offset -= resid - uio->uio_resid;
366 			uio->uio_resid = resid;
367 		}
368 	} else if (resid > uio->uio_resid && (ioflag & IO_SYNC)) {
369 		error = ext2fs_update(ip, 1);
370 	}
371 	/* correct the result for writes clamped by vn_fsizechk() */
372 	uio->uio_resid += overrun;
373 	return (error);
374 }
375