linux_getcwd.c revision 204523 
1/* $OpenBSD: linux_getcwd.c,v 1.2 2001/05/16 12:50:21 ho Exp $ */
2/* $NetBSD: vfs_getcwd.c,v 1.3.2.3 1999/07/11 10:24:09 sommerfeld Exp $ */
3/*-
4 * Copyright (c) 1999 The NetBSD Foundation, Inc.
5 * All rights reserved.
6 *
7 * This code is derived from software contributed to The NetBSD Foundation
8 * by Bill Sommerfeld.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 *    notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 *    notice, this list of conditions and the following disclaimer in the
17 *    documentation and/or other materials provided with the distribution.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22 * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29 * POSSIBILITY OF SUCH DAMAGE.
30 */
31
32#include <sys/cdefs.h>
33__FBSDID("$FreeBSD: head/sys/compat/linux/linux_getcwd.c 204523 2010-03-01 17:20:04Z joel $");
34
35#include "opt_compat.h"
36
37#include <sys/param.h>
38#include <sys/systm.h>
39#include <sys/namei.h>
40#include <sys/filedesc.h>
41#include <sys/kernel.h>
42#include <sys/file.h>
43#include <sys/stat.h>
44#include <sys/syscallsubr.h>
45#include <sys/vnode.h>
46#include <sys/mount.h>
47#include <sys/proc.h>
48#include <sys/uio.h>
49#include <sys/malloc.h>
50#include <sys/dirent.h>
51#include <ufs/ufs/dir.h>	/* XXX only for DIRBLKSIZ */
52
53#ifdef COMPAT_LINUX32
54#include <machine/../linux32/linux.h>
55#include <machine/../linux32/linux32_proto.h>
56#else
57#include <machine/../linux/linux.h>
58#include <machine/../linux/linux_proto.h>
59#endif
60#include <compat/linux/linux_util.h>
61
62#include <security/mac/mac_framework.h>
63
64static int
65linux_getcwd_scandir(struct vnode **, struct vnode **,
66    char **, char *, struct thread *);
67static int
68linux_getcwd_common(struct vnode *, struct vnode *,
69		   char **, char *, int, int, struct thread *);
70
71#define DIRENT_MINSIZE (sizeof(struct dirent) - (MAXNAMLEN+1) + 4)
72
73/*
74 * Vnode variable naming conventions in this file:
75 *
76 * rvp: the current root we're aiming towards.
77 * lvp, *lvpp: the "lower" vnode
78 * uvp, *uvpp: the "upper" vnode.
79 *
80 * Since all the vnodes we're dealing with are directories, and the
81 * lookups are going *up* in the filesystem rather than *down*, the
82 * usual "pvp" (parent) or "dvp" (directory) naming conventions are
83 * too confusing.
84 */
85
86/*
87 * XXX Will infinite loop in certain cases if a directory read reliably
88 *	returns EINVAL on last block.
89 * XXX is EINVAL the right thing to return if a directory is malformed?
90 */
91
92/*
93 * XXX Untested vs. mount -o union; probably does the wrong thing.
94 */
95
96/*
97 * Find parent vnode of *lvpp, return in *uvpp
98 *
99 * If we care about the name, scan it looking for name of directory
100 * entry pointing at lvp.
101 *
102 * Place the name in the buffer which starts at bufp, immediately
103 * before *bpp, and move bpp backwards to point at the start of it.
104 *
105 * On entry, *lvpp is a locked vnode reference; on exit, it is vput and NULL'ed
106 * On exit, *uvpp is either NULL or is a locked vnode reference.
107 */
108static int
109linux_getcwd_scandir(lvpp, uvpp, bpp, bufp, td)
110	struct vnode **lvpp;
111	struct vnode **uvpp;
112	char **bpp;
113	char *bufp;
114	struct thread *td;
115{
116	int     error = 0;
117	int     eofflag;
118	off_t   off;
119	int     tries;
120	struct uio uio;
121	struct iovec iov;
122	char   *dirbuf = NULL;
123	int	dirbuflen;
124	ino_t   fileno;
125	struct vattr va;
126	struct vnode *uvp = NULL;
127	struct vnode *lvp = *lvpp;
128	struct componentname cn;
129	int len, reclen;
130	tries = 0;
131
132	/*
133	 * If we want the filename, get some info we need while the
134	 * current directory is still locked.
135	 */
136	if (bufp != NULL) {
137		error = VOP_GETATTR(lvp, &va, td->td_ucred);
138		if (error) {
139			vput(lvp);
140			*lvpp = NULL;
141			*uvpp = NULL;
142			return error;
143		}
144	}
145
146	/*
147	 * Ok, we have to do it the hard way..
148	 * Next, get parent vnode using lookup of ..
149	 */
150	cn.cn_nameiop = LOOKUP;
151	cn.cn_flags = ISLASTCN | ISDOTDOT | RDONLY;
152	cn.cn_thread = td;
153	cn.cn_cred = td->td_ucred;
154	cn.cn_pnbuf = NULL;
155	cn.cn_nameptr = "..";
156	cn.cn_namelen = 2;
157	cn.cn_consume = 0;
158	cn.cn_lkflags = LK_SHARED;
159
160	/*
161	 * At this point, lvp is locked and will be unlocked by the lookup.
162	 * On successful return, *uvpp will be locked
163	 */
164#ifdef MAC
165	error = mac_vnode_check_lookup(td->td_ucred, lvp, &cn);
166	if (error == 0)
167#endif
168		error = VOP_LOOKUP(lvp, uvpp, &cn);
169	if (error) {
170		vput(lvp);
171		*lvpp = NULL;
172		*uvpp = NULL;
173		return error;
174	}
175	uvp = *uvpp;
176
177	/* If we don't care about the pathname, we're done */
178	if (bufp == NULL) {
179		vput(lvp);
180		*lvpp = NULL;
181		return 0;
182	}
183
184	fileno = va.va_fileid;
185
186	dirbuflen = DIRBLKSIZ;
187	if (dirbuflen < va.va_blocksize)
188		dirbuflen = va.va_blocksize;
189	dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
190
191#if 0
192unionread:
193#endif
194	off = 0;
195	do {
196		/* call VOP_READDIR of parent */
197		iov.iov_base = dirbuf;
198		iov.iov_len = dirbuflen;
199
200		uio.uio_iov = &iov;
201		uio.uio_iovcnt = 1;
202		uio.uio_offset = off;
203		uio.uio_resid = dirbuflen;
204		uio.uio_segflg = UIO_SYSSPACE;
205		uio.uio_rw = UIO_READ;
206		uio.uio_td = td;
207
208		eofflag = 0;
209
210#ifdef MAC
211		error = mac_vnode_check_readdir(td->td_ucred, uvp);
212		if (error == 0)
213#endif /* MAC */
214			error = VOP_READDIR(uvp, &uio, td->td_ucred, &eofflag,
215			    0, 0);
216
217		off = uio.uio_offset;
218
219		/*
220		 * Try again if NFS tosses its cookies.
221		 * XXX this can still loop forever if the directory is busted
222		 * such that the second or subsequent page of it always
223		 * returns EINVAL
224		 */
225		if ((error == EINVAL) && (tries < 3)) {
226			off = 0;
227			tries++;
228			continue;	/* once more, with feeling */
229		}
230
231		if (!error) {
232			char   *cpos;
233			struct dirent *dp;
234
235			cpos = dirbuf;
236			tries = 0;
237
238			/* scan directory page looking for matching vnode */
239			for (len = (dirbuflen - uio.uio_resid); len > 0; len -= reclen) {
240				dp = (struct dirent *) cpos;
241				reclen = dp->d_reclen;
242
243				/* check for malformed directory.. */
244				if (reclen < DIRENT_MINSIZE) {
245					error = EINVAL;
246					goto out;
247				}
248				/*
249				 * XXX should perhaps do VOP_LOOKUP to
250				 * check that we got back to the right place,
251				 * but getting the locking games for that
252				 * right would be heinous.
253				 */
254				if ((dp->d_type != DT_WHT) &&
255				    (dp->d_fileno == fileno)) {
256					char *bp = *bpp;
257					bp -= dp->d_namlen;
258
259					if (bp <= bufp) {
260						error = ERANGE;
261						goto out;
262					}
263					bcopy(dp->d_name, bp, dp->d_namlen);
264					error = 0;
265					*bpp = bp;
266					goto out;
267				}
268				cpos += reclen;
269			}
270		}
271	} while (!eofflag);
272	error = ENOENT;
273
274out:
275	vput(lvp);
276	*lvpp = NULL;
277	free(dirbuf, M_TEMP);
278	return error;
279}
280
281
282/*
283 * common routine shared by sys___getcwd() and linux_vn_isunder()
284 */
285
286#define GETCWD_CHECK_ACCESS 0x0001
287
288static int
289linux_getcwd_common (lvp, rvp, bpp, bufp, limit, flags, td)
290	struct vnode *lvp;
291	struct vnode *rvp;
292	char **bpp;
293	char *bufp;
294	int limit;
295	int flags;
296	struct thread *td;
297{
298	struct filedesc *fdp = td->td_proc->p_fd;
299	struct vnode *uvp = NULL;
300	char *bp = NULL;
301	int error;
302	accmode_t accmode = VEXEC;
303
304	if (rvp == NULL) {
305		rvp = fdp->fd_rdir;
306		if (rvp == NULL)
307			rvp = rootvnode;
308	}
309
310	VREF(rvp);
311	VREF(lvp);
312
313	/*
314	 * Error handling invariant:
315	 * Before a `goto out':
316	 *	lvp is either NULL, or locked and held.
317	 *	uvp is either NULL, or locked and held.
318	 */
319
320	error = vn_lock(lvp, LK_EXCLUSIVE | LK_RETRY);
321	if (error != 0)
322		panic("vn_lock LK_RETRY returned error %d", error);
323	if (bufp)
324		bp = *bpp;
325	/*
326	 * this loop will terminate when one of the following happens:
327	 *	- we hit the root
328	 *	- getdirentries or lookup fails
329	 *	- we run out of space in the buffer.
330	 */
331	if (lvp == rvp) {
332		if (bp)
333			*(--bp) = '/';
334		goto out;
335	}
336	do {
337		if (lvp->v_type != VDIR) {
338			error = ENOTDIR;
339			goto out;
340		}
341
342		/*
343		 * access check here is optional, depending on
344		 * whether or not caller cares.
345		 */
346		if (flags & GETCWD_CHECK_ACCESS) {
347			error = VOP_ACCESS(lvp, accmode, td->td_ucred, td);
348			if (error)
349				goto out;
350			accmode = VEXEC|VREAD;
351		}
352
353		/*
354		 * step up if we're a covered vnode..
355		 */
356		while (lvp->v_vflag & VV_ROOT) {
357			struct vnode *tvp;
358
359			if (lvp == rvp)
360				goto out;
361
362			tvp = lvp;
363			lvp = lvp->v_mount->mnt_vnodecovered;
364			vput(tvp);
365			/*
366			 * hodie natus est radici frater
367			 */
368			if (lvp == NULL) {
369				error = ENOENT;
370				goto out;
371			}
372			VREF(lvp);
373			error = vn_lock(lvp, LK_EXCLUSIVE | LK_RETRY);
374			if (error != 0)
375				panic("vn_lock LK_RETRY returned %d", error);
376		}
377		error = linux_getcwd_scandir(&lvp, &uvp, &bp, bufp, td);
378		if (error)
379			goto out;
380#ifdef DIAGNOSTIC
381		if (lvp != NULL)
382			panic("getcwd: oops, forgot to null lvp");
383		if (bufp && (bp <= bufp)) {
384			panic("getcwd: oops, went back too far");
385		}
386#endif
387		if (bp)
388			*(--bp) = '/';
389		lvp = uvp;
390		uvp = NULL;
391		limit--;
392	} while ((lvp != rvp) && (limit > 0));
393
394out:
395	if (bpp)
396		*bpp = bp;
397	if (uvp)
398		vput(uvp);
399	if (lvp)
400		vput(lvp);
401	vrele(rvp);
402	return error;
403}
404
405
406/*
407 * Find pathname of process's current directory.
408 *
409 * Use vfs vnode-to-name reverse cache; if that fails, fall back
410 * to reading directory contents.
411 */
412
413int
414linux_getcwd(struct thread *td, struct linux_getcwd_args *args)
415{
416	caddr_t bp, bend, path;
417	int error, len, lenused;
418
419#ifdef DEBUG
420	if (ldebug(getcwd))
421		printf(ARGS(getcwd, "%p, %ld"), args->buf, (long)args->bufsize);
422#endif
423
424	len = args->bufsize;
425
426	if (len > MAXPATHLEN*4)
427		len = MAXPATHLEN*4;
428	else if (len < 2)
429		return ERANGE;
430
431	path = (char *)malloc(len, M_TEMP, M_WAITOK);
432
433	error = kern___getcwd(td, path, UIO_SYSSPACE, len);
434	if (!error) {
435		lenused = strlen(path) + 1;
436		if (lenused <= args->bufsize) {
437			td->td_retval[0] = lenused;
438			error = copyout(path, args->buf, lenused);
439		}
440		else
441			error = ERANGE;
442	} else {
443		bp = &path[len];
444		bend = bp;
445		*(--bp) = '\0';
446
447		/*
448		 * 5th argument here is "max number of vnodes to traverse".
449		 * Since each entry takes up at least 2 bytes in the output buffer,
450		 * limit it to N/2 vnodes for an N byte buffer.
451		 */
452
453		mtx_lock(&Giant);
454		error = linux_getcwd_common (td->td_proc->p_fd->fd_cdir, NULL,
455		    &bp, path, len/2, GETCWD_CHECK_ACCESS, td);
456		mtx_unlock(&Giant);
457
458		if (error)
459			goto out;
460		lenused = bend - bp;
461		td->td_retval[0] = lenused;
462		/* put the result into user buffer */
463		error = copyout(bp, args->buf, lenused);
464	}
465out:
466	free(path, M_TEMP);
467	return (error);
468}
469
470