1/*	$NetBSD: linux_file64.c,v 1.52 2011/09/01 12:44:10 njoly Exp $	*/
2
3/*-
4 * Copyright (c) 1995, 1998, 2000, 2008 The NetBSD Foundation, Inc.
5 * All rights reserved.
6 *
7 * This code is derived from software contributed to The NetBSD Foundation
8 * by Frank van der Linden and Eric Haszlakiewicz.
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/*
33 * Linux 64bit filesystem calls. Used on 32bit archs, not used on 64bit ones.
34 */
35
36#include <sys/cdefs.h>
37__KERNEL_RCSID(0, "$NetBSD: linux_file64.c,v 1.52 2011/09/01 12:44:10 njoly Exp $");
38
39#include <sys/param.h>
40#include <sys/systm.h>
41#include <sys/namei.h>
42#include <sys/proc.h>
43#include <sys/dirent.h>
44#include <sys/file.h>
45#include <sys/stat.h>
46#include <sys/filedesc.h>
47#include <sys/ioctl.h>
48#include <sys/kernel.h>
49#include <sys/mount.h>
50#include <sys/malloc.h>
51#include <sys/namei.h>
52#include <sys/vfs_syscalls.h>
53#include <sys/vnode.h>
54#include <sys/tty.h>
55#include <sys/conf.h>
56
57#include <sys/syscallargs.h>
58
59#include <compat/linux/common/linux_types.h>
60#include <compat/linux/common/linux_signal.h>
61#include <compat/linux/common/linux_fcntl.h>
62#include <compat/linux/common/linux_util.h>
63#include <compat/linux/common/linux_machdep.h>
64#include <compat/linux/common/linux_dirent.h>
65#include <compat/linux/common/linux_ipc.h>
66#include <compat/linux/common/linux_sem.h>
67
68#include <compat/linux/linux_syscallargs.h>
69
70#ifndef alpha
71
72static void bsd_to_linux_stat(struct stat *, struct linux_stat64 *);
73
74/*
75 * Convert a NetBSD stat structure to a Linux stat structure.
76 * Only the order of the fields and the padding in the structure
77 * is different. linux_fakedev is a machine-dependent function
78 * which optionally converts device driver major/minor numbers
79 * (XXX horrible, but what can you do against code that compares
80 * things against constant major device numbers? sigh)
81 */
82static void
83bsd_to_linux_stat(struct stat *bsp, struct linux_stat64 *lsp)
84{
85	lsp->lst_dev     = linux_fakedev(bsp->st_dev, 0);
86	lsp->lst_ino     = bsp->st_ino;
87	lsp->lst_mode    = (linux_mode_t)bsp->st_mode;
88	if (bsp->st_nlink >= (1 << 15))
89		lsp->lst_nlink = (1 << 15) - 1;
90	else
91		lsp->lst_nlink = (linux_nlink_t)bsp->st_nlink;
92	lsp->lst_uid     = bsp->st_uid;
93	lsp->lst_gid     = bsp->st_gid;
94	lsp->lst_rdev    = linux_fakedev(bsp->st_rdev, 1);
95	lsp->lst_size    = bsp->st_size;
96	lsp->lst_blksize = bsp->st_blksize;
97	lsp->lst_blocks  = bsp->st_blocks;
98	lsp->lst_atime   = bsp->st_atime;
99	lsp->lst_mtime   = bsp->st_mtime;
100	lsp->lst_ctime   = bsp->st_ctime;
101#  ifdef LINUX_STAT64_HAS_NSEC
102	lsp->lst_atime_nsec   = bsp->st_atimensec;
103	lsp->lst_mtime_nsec   = bsp->st_mtimensec;
104	lsp->lst_ctime_nsec   = bsp->st_ctimensec;
105#  endif
106#  if LINUX_STAT64_HAS_BROKEN_ST_INO
107	lsp->__lst_ino   = (linux_ino_t) bsp->st_ino;
108#  endif
109}
110
111/*
112 * The stat functions below are plain sailing. stat and lstat are handled
113 * by one function to avoid code duplication.
114 */
115int
116linux_sys_fstat64(struct lwp *l, const struct linux_sys_fstat64_args *uap, register_t *retval)
117{
118	/* {
119		syscallarg(int) fd;
120		syscallarg(struct linux_stat64 *) sp;
121	} */
122	struct linux_stat64 tmplst;
123	struct stat tmpst;
124	int error;
125
126	error = do_sys_fstat(SCARG(uap, fd), &tmpst);
127	if (error != 0)
128		return error;
129
130	bsd_to_linux_stat(&tmpst, &tmplst);
131
132	return copyout(&tmplst, SCARG(uap, sp), sizeof tmplst);
133}
134
135static int
136linux_do_stat64(struct lwp *l, const struct linux_sys_stat64_args *uap, register_t *retval, int flags)
137{
138	struct linux_stat64 tmplst;
139	struct stat tmpst;
140	int error;
141
142	error = do_sys_stat(SCARG(uap, path), flags, &tmpst);
143	if (error != 0)
144		return error;
145
146	bsd_to_linux_stat(&tmpst, &tmplst);
147
148	return copyout(&tmplst, SCARG(uap, sp), sizeof tmplst);
149}
150
151int
152linux_sys_stat64(struct lwp *l, const struct linux_sys_stat64_args *uap, register_t *retval)
153{
154	/* {
155		syscallarg(const char *) path;
156		syscallarg(struct linux_stat64 *) sp;
157	} */
158
159	return linux_do_stat64(l, uap, retval, FOLLOW);
160}
161
162int
163linux_sys_lstat64(struct lwp *l, const struct linux_sys_lstat64_args *uap, register_t *retval)
164{
165	/* {
166		syscallarg(const char *) path;
167		syscallarg(struct linux_stat64 *) sp;
168	} */
169
170	return linux_do_stat64(l, (const void *)uap, retval, NOFOLLOW);
171}
172
173int
174linux_sys_truncate64(struct lwp *l, const struct linux_sys_truncate64_args *uap, register_t *retval)
175{
176	/* {
177		syscallarg(const char *) path;
178		syscallarg(off_t) length;
179	} */
180	struct sys_truncate_args ta;
181
182	/* Linux doesn't have the 'pad' pseudo-parameter */
183	SCARG(&ta, path) = SCARG(uap, path);
184	SCARG(&ta, PAD) = 0;
185	SCARG(&ta, length) = SCARG(uap, length);
186
187	return sys_truncate(l, &ta, retval);
188}
189
190int
191linux_sys_ftruncate64(struct lwp *l, const struct linux_sys_ftruncate64_args *uap, register_t *retval)
192{
193	/* {
194		syscallarg(unsigned int) fd;
195		syscallarg(off_t) length;
196	} */
197	struct sys_ftruncate_args ta;
198
199	/* Linux doesn't have the 'pad' pseudo-parameter */
200	SCARG(&ta, fd) = SCARG(uap, fd);
201	SCARG(&ta, PAD) = 0;
202	SCARG(&ta, length) = SCARG(uap, length);
203
204	return sys_ftruncate(l, &ta, retval);
205}
206#endif /* !alpha */
207
208/*
209 * Linux 'readdir' call. This code is mostly taken from the
210 * SunOS getdents call (see compat/sunos/sunos_misc.c), though
211 * an attempt has been made to keep it a little cleaner.
212 *
213 * The d_off field contains the offset of the next valid entry,
214 * unless the older Linux getdents(2), which used to have it set
215 * to the offset of the entry itself. This function also doesn't
216 * need to deal with the old count == 1 glibc problem.
217 *
218 * Read in BSD-style entries, convert them, and copy them out.
219 *
220 * Note that this doesn't handle union-mounted filesystems.
221 */
222int
223linux_sys_getdents64(struct lwp *l, const struct linux_sys_getdents64_args *uap, register_t *retval)
224{
225	/* {
226		syscallarg(int) fd;
227		syscallarg(struct linux_dirent64 *) dent;
228		syscallarg(unsigned int) count;
229	} */
230	struct dirent *bdp;
231	struct vnode *vp;
232	char *inp, *tbuf;		/* BSD-format */
233	int len, reclen;		/* BSD-format */
234	char *outp;			/* Linux-format */
235	int resid, linux_reclen = 0;	/* Linux-format */
236	file_t *fp;
237	struct uio auio;
238	struct iovec aiov;
239	struct linux_dirent64 idb;
240	off_t off;		/* true file offset */
241	int buflen, error, eofflag, nbytes;
242	struct vattr va;
243	off_t *cookiebuf = NULL, *cookie;
244	int ncookies;
245
246	/* fd_getvnode() will use the descriptor for us */
247	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
248		return (error);
249
250	if ((fp->f_flag & FREAD) == 0) {
251		error = EBADF;
252		goto out1;
253	}
254
255	vp = (struct vnode *)fp->f_data;
256	if (vp->v_type != VDIR) {
257		error = ENOTDIR;
258		goto out1;
259	}
260
261	vn_lock(vp, LK_SHARED | LK_RETRY);
262	error = VOP_GETATTR(vp, &va, l->l_cred);
263	VOP_UNLOCK(vp);
264	if (error)
265		goto out1;
266
267	nbytes = SCARG(uap, count);
268	buflen = min(MAXBSIZE, nbytes);
269	if (buflen < va.va_blocksize)
270		buflen = va.va_blocksize;
271	tbuf = malloc(buflen, M_TEMP, M_WAITOK);
272
273	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
274	off = fp->f_offset;
275again:
276	aiov.iov_base = tbuf;
277	aiov.iov_len = buflen;
278	auio.uio_iov = &aiov;
279	auio.uio_iovcnt = 1;
280	auio.uio_rw = UIO_READ;
281	auio.uio_resid = buflen;
282	auio.uio_offset = off;
283	UIO_SETUP_SYSSPACE(&auio);
284	/*
285         * First we read into the malloc'ed buffer, then
286         * we massage it into user space, one record at a time.
287         */
288	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
289	    &ncookies);
290	if (error)
291		goto out;
292
293	inp = tbuf;
294	outp = (void *)SCARG(uap, dent);
295	resid = nbytes;
296	if ((len = buflen - auio.uio_resid) == 0)
297		goto eof;
298
299	for (cookie = cookiebuf; len > 0; len -= reclen) {
300		bdp = (struct dirent *)inp;
301		reclen = bdp->d_reclen;
302		if (reclen & 3)
303			panic("linux_readdir");
304		if (bdp->d_fileno == 0) {
305			inp += reclen;	/* it is a hole; squish it out */
306			if (cookie)
307				off = *cookie++;
308			else
309				off += reclen;
310			continue;
311		}
312		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
313		if (reclen > len || resid < linux_reclen) {
314			/* entry too big for buffer, so just stop */
315			outp++;
316			break;
317		}
318		if (cookie)
319			off = *cookie++;	/* each entry points to next */
320		else
321			off += reclen;
322		/*
323		 * Massage in place to make a Linux-shaped dirent (otherwise
324		 * we have to worry about touching user memory outside of
325		 * the copyout() call).
326		 */
327		idb.d_ino = bdp->d_fileno;
328		idb.d_type = bdp->d_type;
329		idb.d_off = off;
330		idb.d_reclen = (u_short)linux_reclen;
331		strcpy(idb.d_name, bdp->d_name);
332		if ((error = copyout((void *)&idb, outp, linux_reclen)))
333			goto out;
334		/* advance past this real entry */
335		inp += reclen;
336		/* advance output past Linux-shaped entry */
337		outp += linux_reclen;
338		resid -= linux_reclen;
339	}
340
341	/* if we squished out the whole block, try again */
342	if (outp == (void *)SCARG(uap, dent)) {
343		if (cookiebuf)
344			free(cookiebuf, M_TEMP);
345		cookiebuf = NULL;
346		goto again;
347	}
348	fp->f_offset = off;	/* update the vnode offset */
349
350eof:
351	*retval = nbytes - resid;
352out:
353	VOP_UNLOCK(vp);
354	if (cookiebuf)
355		free(cookiebuf, M_TEMP);
356	free(tbuf, M_TEMP);
357out1:
358	fd_putfile(SCARG(uap, fd));
359	return error;
360}
361