xref: /netbsd-src/sys/ufs/ext2fs/ext2fs_readwrite.c (revision b1c86f5f087524e68db12794ee9c3e3da1ab17a0)
1 /*	$NetBSD: ext2fs_readwrite.c,v 1.56 2010/04/23 15:38:46 pooka 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  *
47  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
48  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
49  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
50  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
51  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
52  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
53  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
54  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
55  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
56  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
57  *
58  *	@(#)ufs_readwrite.c	8.8 (Berkeley) 8/4/94
59  * Modified for ext2fs by Manuel Bouyer.
60  */
61 
62 #include <sys/cdefs.h>
63 __KERNEL_RCSID(0, "$NetBSD: ext2fs_readwrite.c,v 1.56 2010/04/23 15:38:46 pooka Exp $");
64 
65 #include <sys/param.h>
66 #include <sys/systm.h>
67 #include <sys/resourcevar.h>
68 #include <sys/kernel.h>
69 #include <sys/file.h>
70 #include <sys/stat.h>
71 #include <sys/buf.h>
72 #include <sys/proc.h>
73 #include <sys/mount.h>
74 #include <sys/vnode.h>
75 #include <sys/malloc.h>
76 #include <sys/signalvar.h>
77 #include <sys/kauth.h>
78 
79 #include <ufs/ufs/inode.h>
80 #include <ufs/ufs/ufsmount.h>
81 #include <ufs/ufs/ufs_extern.h>
82 #include <ufs/ext2fs/ext2fs.h>
83 #include <ufs/ext2fs/ext2fs_extern.h>
84 
85 
86 #define doclusterread 0 /* XXX underway */
87 #define doclusterwrite 0
88 
89 /*
90  * Vnode op for reading.
91  */
92 /* ARGSUSED */
93 int
94 ext2fs_read(void *v)
95 {
96 	struct vop_read_args /* {
97 		struct vnode *a_vp;
98 		struct uio *a_uio;
99 		int a_ioflag;
100 		kauth_cred_t a_cred;
101 	} */ *ap = v;
102 	struct vnode *vp;
103 	struct inode *ip;
104 	struct uio *uio;
105 	struct m_ext2fs *fs;
106 	struct buf *bp;
107 	struct ufsmount *ump;
108 	vsize_t bytelen;
109 	daddr_t lbn, nextlbn;
110 	off_t bytesinfile;
111 	long size, xfersize, blkoffset;
112 	int error;
113 
114 	vp = ap->a_vp;
115 	ip = VTOI(vp);
116 	ump = ip->i_ump;
117 	uio = ap->a_uio;
118 	error = 0;
119 
120 #ifdef DIAGNOSTIC
121 	if (uio->uio_rw != UIO_READ)
122 		panic("%s: mode", "ext2fs_read");
123 
124 	if (vp->v_type == VLNK) {
125 		if (ext2fs_size(ip) < ump->um_maxsymlinklen ||
126 		    (ump->um_maxsymlinklen == 0 && ip->i_e2fs_nblock == 0))
127 			panic("%s: short symlink", "ext2fs_read");
128 	} else if (vp->v_type != VREG && vp->v_type != VDIR)
129 		panic("%s: type %d", "ext2fs_read", vp->v_type);
130 #endif
131 	fs = ip->i_e2fs;
132 	if ((uint64_t)uio->uio_offset > ump->um_maxfilesize)
133 		return (EFBIG);
134 	if (uio->uio_resid == 0)
135 		return (0);
136 	if (uio->uio_offset >= ext2fs_size(ip))
137 		goto out;
138 
139 	if (vp->v_type == VREG) {
140 		const int advice = IO_ADV_DECODE(ap->a_ioflag);
141 
142 		while (uio->uio_resid > 0) {
143 			bytelen = MIN(ext2fs_size(ip) - uio->uio_offset,
144 			    uio->uio_resid);
145 			if (bytelen == 0)
146 				break;
147 
148 			error = ubc_uiomove(&vp->v_uobj, uio, bytelen, advice,
149 			    UBC_READ | UBC_PARTIALOK | UBC_UNMAP_FLAG(vp));
150 			if (error)
151 				break;
152 		}
153 		goto out;
154 	}
155 
156 	for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
157 		bytesinfile = ext2fs_size(ip) - uio->uio_offset;
158 		if (bytesinfile <= 0)
159 			break;
160 		lbn = lblkno(fs, uio->uio_offset);
161 		nextlbn = lbn + 1;
162 		size = fs->e2fs_bsize;
163 		blkoffset = blkoff(fs, uio->uio_offset);
164 		xfersize = fs->e2fs_bsize - blkoffset;
165 		if (uio->uio_resid < xfersize)
166 			xfersize = uio->uio_resid;
167 		if (bytesinfile < xfersize)
168 			xfersize = bytesinfile;
169 
170 		if (lblktosize(fs, nextlbn) >= ext2fs_size(ip))
171 			error = bread(vp, lbn, size, NOCRED, 0, &bp);
172 		else {
173 			int nextsize = fs->e2fs_bsize;
174 			error = breadn(vp, lbn,
175 				size, &nextlbn, &nextsize, 1, NOCRED, 0, &bp);
176 		}
177 		if (error)
178 			break;
179 
180 		/*
181 		 * We should only get non-zero b_resid when an I/O error
182 		 * has occurred, which should cause us to break above.
183 		 * However, if the short read did not cause an error,
184 		 * then we want to ensure that we do not uiomove bad
185 		 * or uninitialized data.
186 		 */
187 		size -= bp->b_resid;
188 		if (size < xfersize) {
189 			if (size == 0)
190 				break;
191 			xfersize = size;
192 		}
193 		error = uiomove((char *)bp->b_data + blkoffset, xfersize, uio);
194 		if (error)
195 			break;
196 		brelse(bp, 0);
197 	}
198 	if (bp != NULL)
199 		brelse(bp, 0);
200 
201 out:
202 	if (!(vp->v_mount->mnt_flag & MNT_NOATIME)) {
203 		ip->i_flag |= IN_ACCESS;
204 		if ((ap->a_ioflag & IO_SYNC) == IO_SYNC)
205 			error = ext2fs_update(vp, NULL, NULL, UPDATE_WAIT);
206 	}
207 	return (error);
208 }
209 
210 /*
211  * Vnode op for writing.
212  */
213 int
214 ext2fs_write(void *v)
215 {
216 	struct vop_write_args /* {
217 		struct vnode *a_vp;
218 		struct uio *a_uio;
219 		int a_ioflag;
220 		kauth_cred_t a_cred;
221 	} */ *ap = v;
222 	struct vnode *vp;
223 	struct uio *uio;
224 	struct inode *ip;
225 	struct m_ext2fs *fs;
226 	struct buf *bp;
227 	struct ufsmount *ump;
228 	daddr_t lbn;
229 	off_t osize;
230 	int blkoffset, error, flags, ioflag, resid, xfersize;
231 	vsize_t bytelen;
232 	off_t oldoff = 0;					/* XXX */
233 	bool async;
234 	int extended = 0;
235 	int advice;
236 
237 	ioflag = ap->a_ioflag;
238 	advice = IO_ADV_DECODE(ioflag);
239 	uio = ap->a_uio;
240 	vp = ap->a_vp;
241 	ip = VTOI(vp);
242 	ump = ip->i_ump;
243 	error = 0;
244 
245 #ifdef DIAGNOSTIC
246 	if (uio->uio_rw != UIO_WRITE)
247 		panic("%s: mode", "ext2fs_write");
248 #endif
249 
250 	switch (vp->v_type) {
251 	case VREG:
252 		if (ioflag & IO_APPEND)
253 			uio->uio_offset = ext2fs_size(ip);
254 		if ((ip->i_e2fs_flags & EXT2_APPEND) &&
255 		    uio->uio_offset != ext2fs_size(ip))
256 			return (EPERM);
257 		/* FALLTHROUGH */
258 	case VLNK:
259 		break;
260 	case VDIR:
261 		if ((ioflag & IO_SYNC) == 0)
262 			panic("%s: nonsync dir write", "ext2fs_write");
263 		break;
264 	default:
265 		panic("%s: type", "ext2fs_write");
266 	}
267 
268 	fs = ip->i_e2fs;
269 	if (uio->uio_offset < 0 ||
270 	    (uint64_t)uio->uio_offset + uio->uio_resid > ump->um_maxfilesize)
271 		return (EFBIG);
272 	if (uio->uio_resid == 0)
273 		return (0);
274 
275 	async = vp->v_mount->mnt_flag & MNT_ASYNC;
276 	resid = uio->uio_resid;
277 	osize = ext2fs_size(ip);
278 
279 	if (vp->v_type == VREG) {
280 		while (uio->uio_resid > 0) {
281 			oldoff = uio->uio_offset;
282 			blkoffset = blkoff(fs, uio->uio_offset);
283 			bytelen = MIN(fs->e2fs_bsize - blkoffset,
284 			    uio->uio_resid);
285 
286 			if (vp->v_size < oldoff + bytelen) {
287 				uvm_vnp_setwritesize(vp, oldoff + bytelen);
288 			}
289 			error = ufs_balloc_range(vp, uio->uio_offset,
290 			    bytelen, ap->a_cred, 0);
291 			if (error)
292 				break;
293 			error = ubc_uiomove(&vp->v_uobj, uio, bytelen, advice,
294 			    UBC_WRITE | UBC_UNMAP_FLAG(vp));
295 			if (error)
296 				break;
297 
298 			/*
299 			 * update UVM's notion of the size now that we've
300 			 * copied the data into the vnode's pages.
301 			 */
302 
303 			if (vp->v_size < uio->uio_offset) {
304 				uvm_vnp_setsize(vp, uio->uio_offset);
305 				extended = 1;
306 			}
307 
308 			/*
309 			 * flush what we just wrote if necessary.
310 			 * XXXUBC simplistic async flushing.
311 			 */
312 
313 			if (!async && oldoff >> 16 != uio->uio_offset >> 16) {
314 				mutex_enter(&vp->v_interlock);
315 				error = VOP_PUTPAGES(vp, (oldoff >> 16) << 16,
316 				    (uio->uio_offset >> 16) << 16, PGO_CLEANIT);
317 			}
318 		}
319 		if (error == 0 && ioflag & IO_SYNC) {
320 			mutex_enter(&vp->v_interlock);
321 			error = VOP_PUTPAGES(vp, trunc_page(oldoff),
322 			    round_page(blkroundup(fs, uio->uio_offset)),
323 			    PGO_CLEANIT | PGO_SYNCIO);
324 		}
325 
326 		goto out;
327 	}
328 
329 	flags = ioflag & IO_SYNC ? B_SYNC : 0;
330 	for (error = 0; uio->uio_resid > 0;) {
331 		lbn = lblkno(fs, uio->uio_offset);
332 		blkoffset = blkoff(fs, uio->uio_offset);
333 		xfersize = MIN(fs->e2fs_bsize - blkoffset, uio->uio_resid);
334 		if (xfersize < fs->e2fs_bsize)
335 			flags |= B_CLRBUF;
336 		else
337 			flags &= ~B_CLRBUF;
338 		error = ext2fs_balloc(ip,
339 		    lbn, blkoffset + xfersize, ap->a_cred, &bp, flags);
340 		if (error)
341 			break;
342 		if (ext2fs_size(ip) < uio->uio_offset + xfersize) {
343 			error = ext2fs_setsize(ip, uio->uio_offset + xfersize);
344 			if (error)
345 				break;
346 		}
347 		error = uiomove((char *)bp->b_data + blkoffset, xfersize, uio);
348 
349 		/*
350 		 * update UVM's notion of the size now that we've
351 		 * copied the data into the vnode's pages.
352 		 */
353 
354 		if (vp->v_size < uio->uio_offset) {
355 			uvm_vnp_setsize(vp, uio->uio_offset);
356 			extended = 1;
357 		}
358 
359 		if (ioflag & IO_SYNC)
360 			(void)bwrite(bp);
361 		else if (xfersize + blkoffset == fs->e2fs_bsize)
362 			bawrite(bp);
363 		else
364 			bdwrite(bp);
365 		if (error || xfersize == 0)
366 			break;
367 	}
368 
369 	/*
370 	 * If we successfully wrote any data, and we are not the superuser
371 	 * we clear the setuid and setgid bits as a precaution against
372 	 * tampering.
373 	 */
374 
375 out:
376 	ip->i_flag |= IN_CHANGE | IN_UPDATE;
377 	if (resid > uio->uio_resid && ap->a_cred &&
378 	    kauth_authorize_generic(ap->a_cred, KAUTH_GENERIC_ISSUSER, NULL))
379 		ip->i_e2fs_mode &= ~(ISUID | ISGID);
380 	if (resid > uio->uio_resid)
381 		VN_KNOTE(vp, NOTE_WRITE | (extended ? NOTE_EXTEND : 0));
382 	if (error) {
383 		(void) ext2fs_truncate(vp, osize, ioflag & IO_SYNC, ap->a_cred);
384 		uio->uio_offset -= resid - uio->uio_resid;
385 		uio->uio_resid = resid;
386 	} else if (resid > uio->uio_resid && (ioflag & IO_SYNC) == IO_SYNC)
387 		error = ext2fs_update(vp, NULL, NULL, UPDATE_WAIT);
388 	KASSERT(vp->v_size == ext2fs_size(ip));
389 	return (error);
390 }
391