xref: /netbsd-src/sys/ufs/ext2fs/ext2fs_readwrite.c (revision 3816d47b2c42fcd6e549e3407f842a5b1a1d23ad)
1 /*	$NetBSD: ext2fs_readwrite.c,v 1.55 2009/10/19 18:41:17 bouyer 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 1.55 2009/10/19 18:41:17 bouyer 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 proc *p;
228 	struct ufsmount *ump;
229 	daddr_t lbn;
230 	off_t osize;
231 	int blkoffset, error, flags, ioflag, resid, xfersize;
232 	vsize_t bytelen;
233 	off_t oldoff = 0;					/* XXX */
234 	bool async;
235 	int extended = 0;
236 	int advice;
237 
238 	ioflag = ap->a_ioflag;
239 	advice = IO_ADV_DECODE(ioflag);
240 	uio = ap->a_uio;
241 	vp = ap->a_vp;
242 	ip = VTOI(vp);
243 	ump = ip->i_ump;
244 	error = 0;
245 
246 #ifdef DIAGNOSTIC
247 	if (uio->uio_rw != UIO_WRITE)
248 		panic("%s: mode", "ext2fs_write");
249 #endif
250 
251 	switch (vp->v_type) {
252 	case VREG:
253 		if (ioflag & IO_APPEND)
254 			uio->uio_offset = ext2fs_size(ip);
255 		if ((ip->i_e2fs_flags & EXT2_APPEND) &&
256 		    uio->uio_offset != ext2fs_size(ip))
257 			return (EPERM);
258 		/* FALLTHROUGH */
259 	case VLNK:
260 		break;
261 	case VDIR:
262 		if ((ioflag & IO_SYNC) == 0)
263 			panic("%s: nonsync dir write", "ext2fs_write");
264 		break;
265 	default:
266 		panic("%s: type", "ext2fs_write");
267 	}
268 
269 	fs = ip->i_e2fs;
270 	if (uio->uio_offset < 0 ||
271 	    (uint64_t)uio->uio_offset + uio->uio_resid > ump->um_maxfilesize)
272 		return (EFBIG);
273 	/*
274 	 * Maybe this should be above the vnode op call, but so long as
275 	 * file servers have no limits, I don't think it matters.
276 	 */
277 	p = curproc;
278 	if (vp->v_type == VREG && p &&
279 	    uio->uio_offset + uio->uio_resid >
280 	    p->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
281 		mutex_enter(proc_lock);
282 		psignal(p, SIGXFSZ);
283 		mutex_exit(proc_lock);
284 		return (EFBIG);
285 	}
286 	if (uio->uio_resid == 0)
287 		return (0);
288 
289 	async = vp->v_mount->mnt_flag & MNT_ASYNC;
290 	resid = uio->uio_resid;
291 	osize = ext2fs_size(ip);
292 
293 	if (vp->v_type == VREG) {
294 		while (uio->uio_resid > 0) {
295 			oldoff = uio->uio_offset;
296 			blkoffset = blkoff(fs, uio->uio_offset);
297 			bytelen = MIN(fs->e2fs_bsize - blkoffset,
298 			    uio->uio_resid);
299 
300 			if (vp->v_size < oldoff + bytelen) {
301 				uvm_vnp_setwritesize(vp, oldoff + bytelen);
302 			}
303 			error = ufs_balloc_range(vp, uio->uio_offset,
304 			    bytelen, ap->a_cred, 0);
305 			if (error)
306 				break;
307 			error = ubc_uiomove(&vp->v_uobj, uio, bytelen, advice,
308 			    UBC_WRITE | UBC_UNMAP_FLAG(vp));
309 			if (error)
310 				break;
311 
312 			/*
313 			 * update UVM's notion of the size now that we've
314 			 * copied the data into the vnode's pages.
315 			 */
316 
317 			if (vp->v_size < uio->uio_offset) {
318 				uvm_vnp_setsize(vp, uio->uio_offset);
319 				extended = 1;
320 			}
321 
322 			/*
323 			 * flush what we just wrote if necessary.
324 			 * XXXUBC simplistic async flushing.
325 			 */
326 
327 			if (!async && oldoff >> 16 != uio->uio_offset >> 16) {
328 				mutex_enter(&vp->v_interlock);
329 				error = VOP_PUTPAGES(vp, (oldoff >> 16) << 16,
330 				    (uio->uio_offset >> 16) << 16, PGO_CLEANIT);
331 			}
332 		}
333 		if (error == 0 && ioflag & IO_SYNC) {
334 			mutex_enter(&vp->v_interlock);
335 			error = VOP_PUTPAGES(vp, trunc_page(oldoff),
336 			    round_page(blkroundup(fs, uio->uio_offset)),
337 			    PGO_CLEANIT | PGO_SYNCIO);
338 		}
339 
340 		goto out;
341 	}
342 
343 	flags = ioflag & IO_SYNC ? B_SYNC : 0;
344 	for (error = 0; uio->uio_resid > 0;) {
345 		lbn = lblkno(fs, uio->uio_offset);
346 		blkoffset = blkoff(fs, uio->uio_offset);
347 		xfersize = MIN(fs->e2fs_bsize - blkoffset, uio->uio_resid);
348 		if (xfersize < fs->e2fs_bsize)
349 			flags |= B_CLRBUF;
350 		else
351 			flags &= ~B_CLRBUF;
352 		error = ext2fs_balloc(ip,
353 		    lbn, blkoffset + xfersize, ap->a_cred, &bp, flags);
354 		if (error)
355 			break;
356 		if (ext2fs_size(ip) < uio->uio_offset + xfersize) {
357 			error = ext2fs_setsize(ip, uio->uio_offset + xfersize);
358 			if (error)
359 				break;
360 		}
361 		error = uiomove((char *)bp->b_data + blkoffset, xfersize, uio);
362 
363 		/*
364 		 * update UVM's notion of the size now that we've
365 		 * copied the data into the vnode's pages.
366 		 */
367 
368 		if (vp->v_size < uio->uio_offset) {
369 			uvm_vnp_setsize(vp, uio->uio_offset);
370 			extended = 1;
371 		}
372 
373 		if (ioflag & IO_SYNC)
374 			(void)bwrite(bp);
375 		else if (xfersize + blkoffset == fs->e2fs_bsize)
376 			bawrite(bp);
377 		else
378 			bdwrite(bp);
379 		if (error || xfersize == 0)
380 			break;
381 	}
382 
383 	/*
384 	 * If we successfully wrote any data, and we are not the superuser
385 	 * we clear the setuid and setgid bits as a precaution against
386 	 * tampering.
387 	 */
388 
389 out:
390 	ip->i_flag |= IN_CHANGE | IN_UPDATE;
391 	if (resid > uio->uio_resid && ap->a_cred &&
392 	    kauth_authorize_generic(ap->a_cred, KAUTH_GENERIC_ISSUSER, NULL))
393 		ip->i_e2fs_mode &= ~(ISUID | ISGID);
394 	if (resid > uio->uio_resid)
395 		VN_KNOTE(vp, NOTE_WRITE | (extended ? NOTE_EXTEND : 0));
396 	if (error) {
397 		(void) ext2fs_truncate(vp, osize, ioflag & IO_SYNC, ap->a_cred);
398 		uio->uio_offset -= resid - uio->uio_resid;
399 		uio->uio_resid = resid;
400 	} else if (resid > uio->uio_resid && (ioflag & IO_SYNC) == IO_SYNC)
401 		error = ext2fs_update(vp, NULL, NULL, UPDATE_WAIT);
402 	KASSERT(vp->v_size == ext2fs_size(ip));
403 	return (error);
404 }
405