1 /*	$OpenBSD: ufs_readwrite.c,v 1.26 2004/07/13 21:04:29 millert Exp $	*/
2 /*	$NetBSD: ufs_readwrite.c,v 1.9 1996/05/11 18:27:57 mycroft Exp $	*/
3 
4 /*-
5  * Copyright (c) 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)ufs_readwrite.c	8.11 (Berkeley) 5/8/95
33  */
34 
35 #ifdef LFS_READWRITE
36 #define	BLKSIZE(a, b, c)	blksize(a, b, c)
37 #define	FS			struct lfs
38 #define	I_FS			i_lfs
39 #define	READ			lfs_read
40 #define	READ_S			"lfs_read"
41 #define	WRITE			lfs_write
42 #define	WRITE_S			"lfs_write"
43 #define	fs_bsize		lfs_bsize
44 #define MAXFILESIZE		fs->lfs_maxfilesize
45 #else
46 #define	BLKSIZE(a, b, c)	blksize(a, b, c)
47 #define	FS			struct fs
48 #define	I_FS			i_fs
49 #define	READ			ffs_read
50 #define	READ_S			"ffs_read"
51 #define	WRITE			ffs_write
52 #define	WRITE_S			"ffs_write"
53 #define MAXFILESIZE		fs->fs_maxfilesize
54 #endif
55 
56 #include <sys/event.h>
57 
58 #define VN_KNOTE(vp, b) \
59 	KNOTE((struct klist *)&vp->v_selectinfo.vsi_selinfo.si_note, (b))
60 
61 /*
62  * Vnode op for reading.
63  */
64 /* ARGSUSED */
65 int
READ(v)66 READ(v)
67 	void *v;
68 {
69 	struct vop_read_args /* {
70 		struct vnode *a_vp;
71 		struct uio *a_uio;
72 		int a_ioflag;
73 		struct ucred *a_cred;
74 	} */ *ap = v;
75 	register struct vnode *vp;
76 	register struct inode *ip;
77 	register struct uio *uio;
78 	register FS *fs;
79 	struct buf *bp;
80 	daddr_t lbn, nextlbn;
81 	off_t bytesinfile;
82 	long size, xfersize, blkoffset;
83 	mode_t mode;
84 	int error;
85 
86 	vp = ap->a_vp;
87 	ip = VTOI(vp);
88 	mode = ip->i_ffs_mode;
89 	uio = ap->a_uio;
90 
91 #ifdef DIAGNOSTIC
92 	if (uio->uio_rw != UIO_READ)
93 		panic("%s: mode", READ_S);
94 
95 	if (vp->v_type == VLNK) {
96 		if ((int)ip->i_ffs_size < vp->v_mount->mnt_maxsymlinklen ||
97 		    (vp->v_mount->mnt_maxsymlinklen == 0 &&
98 		     ip->i_ffs_blocks == 0))
99 			panic("%s: short symlink", READ_S);
100 	} else if (vp->v_type != VREG && vp->v_type != VDIR)
101 		panic("%s: type %d", READ_S, vp->v_type);
102 #endif
103 	fs = ip->I_FS;
104 	if ((u_int64_t)uio->uio_offset > MAXFILESIZE)
105 		return (EFBIG);
106 
107 	if (uio->uio_resid == 0)
108 		return (0);
109 
110 	for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
111 		if ((bytesinfile = ip->i_ffs_size - uio->uio_offset) <= 0)
112 			break;
113 		lbn = lblkno(fs, uio->uio_offset);
114 		nextlbn = lbn + 1;
115 		size = BLKSIZE(fs, ip, lbn);
116 		blkoffset = blkoff(fs, uio->uio_offset);
117 		xfersize = fs->fs_bsize - blkoffset;
118 		if (uio->uio_resid < xfersize)
119 			xfersize = uio->uio_resid;
120 		if (bytesinfile < xfersize)
121 			xfersize = bytesinfile;
122 
123 #ifdef LFS_READWRITE
124 		(void)lfs_check(vp, lbn);
125 		error = cluster_read(vp, &ip->i_ci, ip->i_ffs_size, lbn,
126 		    size, NOCRED, &bp);
127 #else
128 		if (lblktosize(fs, nextlbn) >= ip->i_ffs_size)
129 			error = bread(vp, lbn, size, NOCRED, &bp);
130 		else if (doclusterread)
131 			error = cluster_read(vp, &ip->i_ci,
132 			    ip->i_ffs_size, lbn, size, NOCRED, &bp);
133 		else if (lbn - 1 == ip->i_ci.ci_lastr) {
134 			int nextsize = BLKSIZE(fs, ip, nextlbn);
135 			error = breadn(vp, lbn,
136 			    size, &nextlbn, &nextsize, 1, NOCRED, &bp);
137 		} else
138 			error = bread(vp, lbn, size, NOCRED, &bp);
139 #endif
140 		if (error)
141 			break;
142 		ip->i_ci.ci_lastr = lbn;
143 
144 		/*
145 		 * We should only get non-zero b_resid when an I/O error
146 		 * has occurred, which should cause us to break above.
147 		 * However, if the short read did not cause an error,
148 		 * then we want to ensure that we do not uiomove bad
149 		 * or uninitialized data.
150 		 */
151 		size -= bp->b_resid;
152 		if (size < xfersize) {
153 			if (size == 0)
154 				break;
155 			xfersize = size;
156 		}
157 		error = uiomove((char *)bp->b_data + blkoffset, (int)xfersize,
158 				uio);
159 		if (error)
160 			break;
161 		brelse(bp);
162 	}
163 	if (bp != NULL)
164 		brelse(bp);
165 	ip->i_flag |= IN_ACCESS;
166 	return (error);
167 }
168 
169 /*
170  * Vnode op for writing.
171  */
172 int
WRITE(v)173 WRITE(v)
174 	void *v;
175 {
176 	struct vop_write_args /* {
177 		struct vnode *a_vp;
178 		struct uio *a_uio;
179 		int a_ioflag;
180 		struct ucred *a_cred;
181 	} */ *ap = v;
182 	register struct vnode *vp;
183 	register struct uio *uio;
184 	register struct inode *ip;
185 	register FS *fs;
186 	struct buf *bp;
187 	struct proc *p;
188 	daddr_t lbn;
189 	off_t osize;
190 	int blkoffset, error, extended, flags, ioflag, resid, size, xfersize;
191 
192 	extended = 0;
193 	ioflag = ap->a_ioflag;
194 	uio = ap->a_uio;
195 	vp = ap->a_vp;
196 	ip = VTOI(vp);
197 
198 #ifdef DIAGNOSTIC
199 	if (uio->uio_rw != UIO_WRITE)
200 		panic("%s: mode", WRITE_S);
201 #endif
202 
203 	/*
204 	 * If writing 0 bytes, succeed and do not change
205 	 * update time or file offset (standards compliance)
206 	 */
207 	if (uio->uio_resid == 0)
208 		return (0);
209 
210 	switch (vp->v_type) {
211 	case VREG:
212 		if (ioflag & IO_APPEND)
213 			uio->uio_offset = ip->i_ffs_size;
214 		if ((ip->i_ffs_flags & APPEND) && uio->uio_offset != ip->i_ffs_size)
215 			return (EPERM);
216 		/* FALLTHROUGH */
217 	case VLNK:
218 		break;
219 	case VDIR:
220 		if ((ioflag & IO_SYNC) == 0)
221 			panic("%s: nonsync dir write", WRITE_S);
222 		break;
223 	default:
224 		panic("%s: type", WRITE_S);
225 	}
226 
227 	fs = ip->I_FS;
228 	if (uio->uio_offset < 0 ||
229 	    (u_int64_t)uio->uio_offset + uio->uio_resid > MAXFILESIZE)
230 		return (EFBIG);
231 	/*
232 	 * Maybe this should be above the vnode op call, but so long as
233 	 * file servers have no limits, I don't think it matters.
234 	 */
235 	p = uio->uio_procp;
236 	if (vp->v_type == VREG && p && !(ioflag & IO_NOLIMIT) &&
237 	    uio->uio_offset + uio->uio_resid >
238 	    p->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
239 		psignal(p, SIGXFSZ);
240 		return (EFBIG);
241 	}
242 
243 	resid = uio->uio_resid;
244 	osize = ip->i_ffs_size;
245 	flags = ioflag & IO_SYNC ? B_SYNC : 0;
246 
247 	for (error = 0; uio->uio_resid > 0;) {
248 		lbn = lblkno(fs, uio->uio_offset);
249 		blkoffset = blkoff(fs, uio->uio_offset);
250 		xfersize = fs->fs_bsize - blkoffset;
251 		if (uio->uio_resid < xfersize)
252 			xfersize = uio->uio_resid;
253 		if (fs->fs_bsize > xfersize)
254 			flags |= B_CLRBUF;
255 		else
256 			flags &= ~B_CLRBUF;
257 
258 		if ((error = UFS_BUF_ALLOC(ip, uio->uio_offset, xfersize,
259 			 ap->a_cred, flags, &bp)) != 0)
260 			break;
261 		if (uio->uio_offset + xfersize > ip->i_ffs_size) {
262 			ip->i_ffs_size = uio->uio_offset + xfersize;
263 			uvm_vnp_setsize(vp, ip->i_ffs_size);
264 			extended = 1;
265 		}
266 		(void)uvm_vnp_uncache(vp);
267 
268 		size = BLKSIZE(fs, ip, lbn) - bp->b_resid;
269 		if (size < xfersize)
270 			xfersize = size;
271 
272 		error =
273 		    uiomove((char *)bp->b_data + blkoffset, xfersize, uio);
274 
275 		if (error != 0)
276 			bzero((char *)bp->b_data + blkoffset, xfersize);
277 
278 #ifdef LFS_READWRITE
279 		(void)VOP_BWRITE(bp);
280 #else
281 		if (ioflag & IO_SYNC)
282 			(void)bwrite(bp);
283 		else if (xfersize + blkoffset == fs->fs_bsize) {
284 			if (doclusterwrite)
285 				cluster_write(bp, &ip->i_ci, ip->i_ffs_size);
286 			else
287 				bawrite(bp);
288 		} else
289 			bdwrite(bp);
290 #endif
291 		if (error || xfersize == 0)
292 			break;
293 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
294 	}
295 	/*
296 	 * If we successfully wrote any data, and we are not the superuser
297 	 * we clear the setuid and setgid bits as a precaution against
298 	 * tampering.
299 	 */
300 	if (resid > uio->uio_resid && ap->a_cred && ap->a_cred->cr_uid != 0)
301 		ip->i_ffs_mode &= ~(ISUID | ISGID);
302 	if (resid > uio->uio_resid)
303 		VN_KNOTE(vp, NOTE_WRITE | (extended ? NOTE_EXTEND : 0));
304 	if (error) {
305 		if (ioflag & IO_UNIT) {
306 			(void)UFS_TRUNCATE(ip, osize,
307 			    ioflag & IO_SYNC, ap->a_cred);
308 			uio->uio_offset -= resid - uio->uio_resid;
309 			uio->uio_resid = resid;
310 		}
311 	} else if (resid > uio->uio_resid && (ioflag & IO_SYNC)) {
312 		error = UFS_UPDATE(ip, MNT_WAIT);
313 	}
314 	return (error);
315 }
316