xref: /netbsd-src/sys/ufs/ext2fs/ext2fs_readwrite.c (revision 8b0f9554ff8762542c4defc4f70e1eb76fb508fa)
1 /*	$NetBSD: ext2fs_readwrite.c,v 1.49 2007/12/08 19:29:53 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  * 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 1.49 2007/12/08 19:29:53 pooka 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 #include <sys/kauth.h>
83 
84 #include <ufs/ufs/inode.h>
85 #include <ufs/ufs/ufsmount.h>
86 #include <ufs/ufs/ufs_extern.h>
87 #include <ufs/ext2fs/ext2fs.h>
88 #include <ufs/ext2fs/ext2fs_extern.h>
89 
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 		kauth_cred_t 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 
120 	vp = ap->a_vp;
121 	ip = VTOI(vp);
122 	ump = ip->i_ump;
123 	uio = ap->a_uio;
124 	error = 0;
125 
126 #ifdef DIAGNOSTIC
127 	if (uio->uio_rw != UIO_READ)
128 		panic("%s: mode", "ext2fs_read");
129 
130 	if (vp->v_type == VLNK) {
131 		if (ext2fs_size(ip) < ump->um_maxsymlinklen ||
132 		    (ump->um_maxsymlinklen == 0 && ip->i_e2fs_nblock == 0))
133 			panic("%s: short symlink", "ext2fs_read");
134 	} else if (vp->v_type != VREG && vp->v_type != VDIR)
135 		panic("%s: type %d", "ext2fs_read", vp->v_type);
136 #endif
137 	fs = ip->i_e2fs;
138 	if ((u_int64_t)uio->uio_offset > ump->um_maxfilesize)
139 		return (EFBIG);
140 	if (uio->uio_resid == 0)
141 		return (0);
142 	if (uio->uio_offset >= ext2fs_size(ip))
143 		goto out;
144 
145 	if (vp->v_type == VREG) {
146 		const int advice = IO_ADV_DECODE(ap->a_ioflag);
147 
148 		while (uio->uio_resid > 0) {
149 			bytelen = MIN(ext2fs_size(ip) - uio->uio_offset,
150 			    uio->uio_resid);
151 			if (bytelen == 0)
152 				break;
153 
154 			win = ubc_alloc(&vp->v_uobj, uio->uio_offset,
155 			    &bytelen, advice, UBC_READ);
156 			error = uiomove(win, bytelen, uio);
157 			flags = UBC_WANT_UNMAP(vp) ? UBC_UNMAP : 0;
158 			ubc_release(win, flags);
159 			if (error)
160 				break;
161 		}
162 		goto out;
163 	}
164 
165 	for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
166 		bytesinfile = ext2fs_size(ip) - uio->uio_offset;
167 		if (bytesinfile <= 0)
168 			break;
169 		lbn = lblkno(fs, uio->uio_offset);
170 		nextlbn = lbn + 1;
171 		size = fs->e2fs_bsize;
172 		blkoffset = blkoff(fs, uio->uio_offset);
173 		xfersize = fs->e2fs_bsize - blkoffset;
174 		if (uio->uio_resid < xfersize)
175 			xfersize = uio->uio_resid;
176 		if (bytesinfile < xfersize)
177 			xfersize = bytesinfile;
178 
179 		if (lblktosize(fs, nextlbn) >= ext2fs_size(ip))
180 			error = bread(vp, lbn, size, NOCRED, &bp);
181 		else {
182 			int nextsize = fs->e2fs_bsize;
183 			error = breadn(vp, lbn,
184 				size, &nextlbn, &nextsize, 1, NOCRED, &bp);
185 		}
186 		if (error)
187 			break;
188 
189 		/*
190 		 * We should only get non-zero b_resid when an I/O error
191 		 * has occurred, which should cause us to break above.
192 		 * However, if the short read did not cause an error,
193 		 * then we want to ensure that we do not uiomove bad
194 		 * or uninitialized data.
195 		 */
196 		size -= bp->b_resid;
197 		if (size < xfersize) {
198 			if (size == 0)
199 				break;
200 			xfersize = size;
201 		}
202 		error = uiomove((char *)bp->b_data + blkoffset, xfersize, uio);
203 		if (error)
204 			break;
205 		brelse(bp, 0);
206 	}
207 	if (bp != NULL)
208 		brelse(bp, 0);
209 
210 out:
211 	if (!(vp->v_mount->mnt_flag & MNT_NOATIME)) {
212 		ip->i_flag |= IN_ACCESS;
213 		if ((ap->a_ioflag & IO_SYNC) == IO_SYNC)
214 			error = ext2fs_update(vp, NULL, NULL, UPDATE_WAIT);
215 	}
216 	return (error);
217 }
218 
219 /*
220  * Vnode op for writing.
221  */
222 int
223 ext2fs_write(void *v)
224 {
225 	struct vop_write_args /* {
226 		struct vnode *a_vp;
227 		struct uio *a_uio;
228 		int a_ioflag;
229 		kauth_cred_t a_cred;
230 	} */ *ap = v;
231 	struct vnode *vp;
232 	struct uio *uio;
233 	struct inode *ip;
234 	struct m_ext2fs *fs;
235 	struct buf *bp;
236 	struct proc *p;
237 	struct ufsmount *ump;
238 	daddr_t lbn;
239 	off_t osize;
240 	int blkoffset, error, flags, ioflag, resid, xfersize;
241 	vsize_t bytelen;
242 	void *win;
243 	off_t oldoff = 0;					/* XXX */
244 	bool async;
245 	int extended = 0;
246 
247 	ioflag = ap->a_ioflag;
248 	uio = ap->a_uio;
249 	vp = ap->a_vp;
250 	ip = VTOI(vp);
251 	ump = ip->i_ump;
252 	error = 0;
253 
254 #ifdef DIAGNOSTIC
255 	if (uio->uio_rw != UIO_WRITE)
256 		panic("%s: mode", "ext2fs_write");
257 #endif
258 
259 	switch (vp->v_type) {
260 	case VREG:
261 		if (ioflag & IO_APPEND)
262 			uio->uio_offset = ext2fs_size(ip);
263 		if ((ip->i_e2fs_flags & EXT2_APPEND) &&
264 		    uio->uio_offset != ext2fs_size(ip))
265 			return (EPERM);
266 		/* FALLTHROUGH */
267 	case VLNK:
268 		break;
269 	case VDIR:
270 		if ((ioflag & IO_SYNC) == 0)
271 			panic("%s: nonsync dir write", "ext2fs_write");
272 		break;
273 	default:
274 		panic("%s: type", "ext2fs_write");
275 	}
276 
277 	fs = ip->i_e2fs;
278 	if (uio->uio_offset < 0 ||
279 	    (u_int64_t)uio->uio_offset + uio->uio_resid > ump->um_maxfilesize)
280 		return (EFBIG);
281 	/*
282 	 * Maybe this should be above the vnode op call, but so long as
283 	 * file servers have no limits, I don't think it matters.
284 	 */
285 	p = curproc;
286 	if (vp->v_type == VREG && p &&
287 	    uio->uio_offset + uio->uio_resid >
288 	    p->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
289 		mutex_enter(&proclist_mutex);
290 		psignal(p, SIGXFSZ);
291 		mutex_exit(&proclist_mutex);
292 		return (EFBIG);
293 	}
294 	if (uio->uio_resid == 0)
295 		return (0);
296 
297 	async = vp->v_mount->mnt_flag & MNT_ASYNC;
298 	resid = uio->uio_resid;
299 	osize = ext2fs_size(ip);
300 
301 	if (vp->v_type == VREG) {
302 		while (uio->uio_resid > 0) {
303 			oldoff = uio->uio_offset;
304 			blkoffset = blkoff(fs, uio->uio_offset);
305 			bytelen = MIN(fs->e2fs_bsize - blkoffset,
306 			    uio->uio_resid);
307 
308 			if (vp->v_size < oldoff + bytelen) {
309 				uvm_vnp_setwritesize(vp, oldoff + bytelen);
310 			}
311 			error = ufs_balloc_range(vp, uio->uio_offset,
312 			    bytelen, ap->a_cred, 0);
313 			if (error)
314 				break;
315 			win = ubc_alloc(&vp->v_uobj, uio->uio_offset,
316 			    &bytelen, UVM_ADV_NORMAL, UBC_WRITE);
317 			error = uiomove(win, bytelen, uio);
318 			flags = UBC_WANT_UNMAP(vp) ? UBC_UNMAP : 0;
319 			ubc_release(win, flags);
320 			if (error)
321 				break;
322 
323 			/*
324 			 * update UVM's notion of the size now that we've
325 			 * copied the data into the vnode's pages.
326 			 */
327 
328 			if (vp->v_size < uio->uio_offset) {
329 				uvm_vnp_setsize(vp, uio->uio_offset);
330 				extended = 1;
331 			}
332 
333 			/*
334 			 * flush what we just wrote if necessary.
335 			 * XXXUBC simplistic async flushing.
336 			 */
337 
338 			if (!async && oldoff >> 16 != uio->uio_offset >> 16) {
339 				simple_lock(&vp->v_interlock);
340 				error = VOP_PUTPAGES(vp, (oldoff >> 16) << 16,
341 				    (uio->uio_offset >> 16) << 16, PGO_CLEANIT);
342 			}
343 		}
344 		if (error == 0 && ioflag & IO_SYNC) {
345 			simple_lock(&vp->v_interlock);
346 			error = VOP_PUTPAGES(vp, trunc_page(oldoff),
347 			    round_page(blkroundup(fs, uio->uio_offset)),
348 			    PGO_CLEANIT | PGO_SYNCIO);
349 		}
350 
351 		goto out;
352 	}
353 
354 	flags = ioflag & IO_SYNC ? B_SYNC : 0;
355 	for (error = 0; uio->uio_resid > 0;) {
356 		lbn = lblkno(fs, uio->uio_offset);
357 		blkoffset = blkoff(fs, uio->uio_offset);
358 		xfersize = MIN(fs->e2fs_bsize - blkoffset, uio->uio_resid);
359 		if (xfersize < fs->e2fs_bsize)
360 			flags |= B_CLRBUF;
361 		else
362 			flags &= ~B_CLRBUF;
363 		error = ext2fs_balloc(ip,
364 		    lbn, blkoffset + xfersize, ap->a_cred, &bp, flags);
365 		if (error)
366 			break;
367 		if (ext2fs_size(ip) < uio->uio_offset + xfersize) {
368 			error = ext2fs_setsize(ip, uio->uio_offset + xfersize);
369 			if (error)
370 				break;
371 		}
372 		error = uiomove((char *)bp->b_data + blkoffset, xfersize, uio);
373 
374 		/*
375 		 * update UVM's notion of the size now that we've
376 		 * copied the data into the vnode's pages.
377 		 */
378 
379 		if (vp->v_size < uio->uio_offset) {
380 			uvm_vnp_setsize(vp, uio->uio_offset);
381 			extended = 1;
382 		}
383 
384 		if (ioflag & IO_SYNC)
385 			(void)bwrite(bp);
386 		else if (xfersize + blkoffset == fs->e2fs_bsize)
387 			bawrite(bp);
388 		else
389 			bdwrite(bp);
390 		if (error || xfersize == 0)
391 			break;
392 	}
393 
394 	/*
395 	 * If we successfully wrote any data, and we are not the superuser
396 	 * we clear the setuid and setgid bits as a precaution against
397 	 * tampering.
398 	 */
399 
400 out:
401 	ip->i_flag |= IN_CHANGE | IN_UPDATE;
402 	if (resid > uio->uio_resid && ap->a_cred &&
403 	    kauth_authorize_generic(ap->a_cred, KAUTH_GENERIC_ISSUSER, NULL))
404 		ip->i_e2fs_mode &= ~(ISUID | ISGID);
405 	if (resid > uio->uio_resid)
406 		VN_KNOTE(vp, NOTE_WRITE | (extended ? NOTE_EXTEND : 0));
407 	if (error) {
408 		(void) ext2fs_truncate(vp, osize, ioflag & IO_SYNC, ap->a_cred);
409 		uio->uio_offset -= resid - uio->uio_resid;
410 		uio->uio_resid = resid;
411 	} else if (resid > uio->uio_resid && (ioflag & IO_SYNC) == IO_SYNC)
412 		error = ext2fs_update(vp, NULL, NULL, UPDATE_WAIT);
413 	KASSERT(vp->v_size == ext2fs_size(ip));
414 	return (error);
415 }
416