Home | History | Annotate | Line # | Download | only in common
linux_file64.c revision 1.36.2.1
      1 /*	$NetBSD: linux_file64.c,v 1.36.2.1 2007/03/13 16:50:18 ad Exp $	*/
      2 
      3 /*-
      4  * Copyright (c) 1995, 1998, 2000 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  * 3. All advertising materials mentioning features or use of this software
     19  *    must display the following acknowledgement:
     20  *	This product includes software developed by the NetBSD
     21  *	Foundation, Inc. and its contributors.
     22  * 4. Neither the name of The NetBSD Foundation nor the names of its
     23  *    contributors may be used to endorse or promote products derived
     24  *    from this software without specific prior written permission.
     25  *
     26  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     27  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     28  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     29  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     30  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     31  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     32  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     33  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     34  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     35  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     36  * POSSIBILITY OF SUCH DAMAGE.
     37  */
     38 
     39 /*
     40  * Linux 64bit filesystem calls. Used on 32bit archs, not used on 64bit ones.
     41  */
     42 
     43 #include <sys/cdefs.h>
     44 __KERNEL_RCSID(0, "$NetBSD: linux_file64.c,v 1.36.2.1 2007/03/13 16:50:18 ad Exp $");
     45 
     46 #include <sys/param.h>
     47 #include <sys/systm.h>
     48 #include <sys/namei.h>
     49 #include <sys/proc.h>
     50 #include <sys/dirent.h>
     51 #include <sys/file.h>
     52 #include <sys/stat.h>
     53 #include <sys/filedesc.h>
     54 #include <sys/ioctl.h>
     55 #include <sys/kernel.h>
     56 #include <sys/mount.h>
     57 #include <sys/malloc.h>
     58 #include <sys/namei.h>
     59 #include <sys/vfs_syscalls.h>
     60 #include <sys/vnode.h>
     61 #include <sys/tty.h>
     62 #include <sys/conf.h>
     63 
     64 #include <sys/syscallargs.h>
     65 
     66 #include <compat/linux/common/linux_types.h>
     67 #include <compat/linux/common/linux_signal.h>
     68 #include <compat/linux/common/linux_fcntl.h>
     69 #include <compat/linux/common/linux_util.h>
     70 #include <compat/linux/common/linux_machdep.h>
     71 #include <compat/linux/common/linux_dirent.h>
     72 
     73 #include <compat/linux/linux_syscallargs.h>
     74 
     75 #ifndef alpha
     76 
     77 # ifndef COMPAT_LINUX32
     78 
     79 static void bsd_to_linux_stat __P((struct stat *, struct linux_stat64 *));
     80 static int linux_do_stat64 __P((struct lwp *, void *, register_t *, int));
     81 
     82 /*
     83  * Convert a NetBSD stat structure to a Linux stat structure.
     84  * Only the order of the fields and the padding in the structure
     85  * is different. linux_fakedev is a machine-dependent function
     86  * which optionally converts device driver major/minor numbers
     87  * (XXX horrible, but what can you do against code that compares
     88  * things against constant major device numbers? sigh)
     89  */
     90 static void
     91 bsd_to_linux_stat(bsp, lsp)
     92 	struct stat *bsp;
     93 	struct linux_stat64 *lsp;
     94 {
     95 	lsp->lst_dev     = linux_fakedev(bsp->st_dev, 0);
     96 	lsp->lst_ino     = bsp->st_ino;
     97 	lsp->lst_mode    = (linux_mode_t)bsp->st_mode;
     98 	if (bsp->st_nlink >= (1 << 15))
     99 		lsp->lst_nlink = (1 << 15) - 1;
    100 	else
    101 		lsp->lst_nlink = (linux_nlink_t)bsp->st_nlink;
    102 	lsp->lst_uid     = bsp->st_uid;
    103 	lsp->lst_gid     = bsp->st_gid;
    104 	lsp->lst_rdev    = linux_fakedev(bsp->st_rdev, 1);
    105 	lsp->lst_size    = bsp->st_size;
    106 	lsp->lst_blksize = bsp->st_blksize;
    107 	lsp->lst_blocks  = bsp->st_blocks;
    108 	lsp->lst_atime   = bsp->st_atime;
    109 	lsp->lst_mtime   = bsp->st_mtime;
    110 	lsp->lst_ctime   = bsp->st_ctime;
    111 #  ifdef LINUX_STAT64_HAS_NSEC
    112 	lsp->lst_atime_nsec   = bsp->st_atimensec;
    113 	lsp->lst_mtime_nsec   = bsp->st_mtimensec;
    114 	lsp->lst_ctime_nsec   = bsp->st_ctimensec;
    115 #  endif
    116 #  if LINUX_STAT64_HAS_BROKEN_ST_INO
    117 	lsp->__lst_ino   = (linux_ino_t) bsp->st_ino;
    118 #  endif
    119 }
    120 
    121 /*
    122  * The stat functions below are plain sailing. stat and lstat are handled
    123  * by one function to avoid code duplication.
    124  */
    125 int
    126 linux_sys_fstat64(l, v, retval)
    127 	struct lwp *l;
    128 	void *v;
    129 	register_t *retval;
    130 {
    131 	struct linux_sys_fstat64_args /* {
    132 		syscallarg(int) fd;
    133 		syscallarg(struct linux_stat64 *) sp;
    134 	} */ *uap = v;
    135 	struct linux_stat64 tmplst;
    136 	struct stat tmpst;
    137 	int error;
    138 
    139 	error = do_sys_fstat(l,  SCARG(uap, fd), &tmpst);
    140 	if (error != 0)
    141 		return error;
    142 
    143 	bsd_to_linux_stat(&tmpst, &tmplst);
    144 
    145 	return copyout(&tmplst, SCARG(uap, sp), sizeof tmplst);
    146 }
    147 
    148 static int
    149 linux_do_stat64(l, v, retval, flags)
    150 	struct lwp *l;
    151 	void *v;
    152 	register_t *retval;
    153 	int flags;
    154 {
    155 	struct linux_stat64 tmplst;
    156 	struct stat tmpst;
    157 	void *sg;
    158 	int error;
    159 	struct linux_sys_stat64_args *uap = v;
    160 
    161 	sg = stackgap_init(l->l_proc, 0);
    162 	CHECK_ALT_EXIST(l, &sg, SCARG(uap, path));
    163 
    164 	error = do_sys_stat(l, SCARG(uap, path), flags, &tmpst);
    165 	if (error != 0)
    166 		return error;
    167 
    168 	bsd_to_linux_stat(&tmpst, &tmplst);
    169 
    170 	return copyout(&tmplst, SCARG(uap, sp), sizeof tmplst);
    171 }
    172 
    173 int
    174 linux_sys_stat64(l, v, retval)
    175 	struct lwp *l;
    176 	void *v;
    177 	register_t *retval;
    178 {
    179 	struct linux_sys_stat64_args /* {
    180 		syscallarg(const char *) path;
    181 		syscallarg(struct linux_stat64 *) sp;
    182 	} */ *uap = v;
    183 
    184 	return linux_do_stat64(l, uap, retval, FOLLOW);
    185 }
    186 
    187 int
    188 linux_sys_lstat64(l, v, retval)
    189 	struct lwp *l;
    190 	void *v;
    191 	register_t *retval;
    192 {
    193 	struct linux_sys_lstat64_args /* {
    194 		syscallarg(const char *) path;
    195 		syscallarg(struct linux_stat64 *) sp;
    196 	} */ *uap = v;
    197 
    198 	return linux_do_stat64(l, uap, retval, NOFOLLOW);
    199 }
    200 
    201 int
    202 linux_sys_truncate64(l, v, retval)
    203 	struct lwp *l;
    204 	void *v;
    205 	register_t *retval;
    206 {
    207 	struct linux_sys_truncate64_args /* {
    208 		syscallarg(const char *) path;
    209 		syscallarg(off_t) length;
    210 	} */ *uap = v;
    211 	struct sys_truncate_args ta;
    212 	struct proc *p = l->l_proc;
    213 	void *sg = stackgap_init(p, 0);
    214 
    215 	CHECK_ALT_EXIST(l, &sg, SCARG(uap, path));
    216 
    217 	/* Linux doesn't have the 'pad' pseudo-parameter */
    218 	SCARG(&ta, path) = SCARG(uap, path);
    219 	SCARG(&ta, pad) = 0;
    220 	SCARG(&ta, length) = SCARG(uap, length);
    221 
    222 	return sys_truncate(l, &ta, retval);
    223 }
    224 
    225 int
    226 linux_sys_ftruncate64(l, v, retval)
    227 	struct lwp *l;
    228 	void *v;
    229 	register_t *retval;
    230 {
    231 	struct linux_sys_ftruncate64_args /* {
    232 		syscallarg(unsigned int) fd;
    233 		syscallarg(off_t) length;
    234 	} */ *uap = v;
    235 	struct sys_ftruncate_args ta;
    236 
    237 	/* Linux doesn't have the 'pad' pseudo-parameter */
    238 	SCARG(&ta, fd) = SCARG(uap, fd);
    239 	SCARG(&ta, pad) = 0;
    240 	SCARG(&ta, length) = SCARG(uap, length);
    241 
    242 	return sys_ftruncate(l, &ta, retval);
    243 }
    244 # endif /* !COMPAT_LINUX32 */
    245 
    246 # if !defined(__m68k__) && (!defined(__amd64__) || defined(COMPAT_LINUX32))
    247 static void bsd_to_linux_flock64 __P((struct linux_flock64 *,
    248     const struct flock *));
    249 static void linux_to_bsd_flock64 __P((struct flock *,
    250     const struct linux_flock64 *));
    251 
    252 static void
    253 bsd_to_linux_flock64(lfp, bfp)
    254 	struct linux_flock64 *lfp;
    255 	const struct flock *bfp;
    256 {
    257 
    258 	lfp->l_start = bfp->l_start;
    259 	lfp->l_len = bfp->l_len;
    260 	lfp->l_pid = bfp->l_pid;
    261 	lfp->l_whence = bfp->l_whence;
    262 	switch (bfp->l_type) {
    263 	case F_RDLCK:
    264 		lfp->l_type = LINUX_F_RDLCK;
    265 		break;
    266 	case F_UNLCK:
    267 		lfp->l_type = LINUX_F_UNLCK;
    268 		break;
    269 	case F_WRLCK:
    270 		lfp->l_type = LINUX_F_WRLCK;
    271 		break;
    272 	}
    273 }
    274 
    275 static void
    276 linux_to_bsd_flock64(bfp, lfp)
    277 	struct flock *bfp;
    278 	const struct linux_flock64 *lfp;
    279 {
    280 
    281 	bfp->l_start = lfp->l_start;
    282 	bfp->l_len = lfp->l_len;
    283 	bfp->l_pid = lfp->l_pid;
    284 	bfp->l_whence = lfp->l_whence;
    285 	switch (lfp->l_type) {
    286 	case LINUX_F_RDLCK:
    287 		bfp->l_type = F_RDLCK;
    288 		break;
    289 	case LINUX_F_UNLCK:
    290 		bfp->l_type = F_UNLCK;
    291 		break;
    292 	case LINUX_F_WRLCK:
    293 		bfp->l_type = F_WRLCK;
    294 		break;
    295 	}
    296 }
    297 
    298 int
    299 linux_sys_fcntl64(l, v, retval)
    300 	struct lwp *l;
    301 	void *v;
    302 	register_t *retval;
    303 {
    304 	struct linux_sys_fcntl64_args /* {
    305 		syscallarg(int) fd;
    306 		syscallarg(int) cmd;
    307 		syscallarg(void *) arg;
    308 	} */ *uap = v;
    309 	struct proc *p = l->l_proc;
    310 	struct sys_fcntl_args fca;
    311 	struct linux_flock64 lfl;
    312 	struct flock bfl, *bfp;
    313 	int error;
    314 	void *sg;
    315 	void *arg = SCARG(uap, arg);
    316 	int cmd = SCARG(uap, cmd);
    317 	int fd = SCARG(uap, fd);
    318 
    319 	switch (cmd) {
    320 	case LINUX_F_GETLK64:
    321 		sg = stackgap_init(p, 0);
    322 		bfp = (struct flock *) stackgap_alloc(p, &sg, sizeof *bfp);
    323 		if ((error = copyin(arg, &lfl, sizeof lfl)) != 0)
    324 			return error;
    325 		linux_to_bsd_flock64(&bfl, &lfl);
    326 		if ((error = copyout(&bfl, bfp, sizeof bfl)) != 0)
    327 			return error;
    328 		SCARG(&fca, fd) = fd;
    329 		SCARG(&fca, cmd) = F_GETLK;
    330 		SCARG(&fca, arg) = bfp;
    331 		if ((error = sys_fcntl(l, &fca, retval)) != 0)
    332 			return error;
    333 		if ((error = copyin(bfp, &bfl, sizeof bfl)) != 0)
    334 			return error;
    335 		bsd_to_linux_flock64(&lfl, &bfl);
    336 		return copyout(&lfl, arg, sizeof lfl);
    337 	case LINUX_F_SETLK64:
    338 	case LINUX_F_SETLKW64:
    339 		cmd = (cmd == LINUX_F_SETLK64 ? F_SETLK : F_SETLKW);
    340 		if ((error = copyin(arg, &lfl, sizeof lfl)) != 0)
    341 			return error;
    342 		linux_to_bsd_flock64(&bfl, &lfl);
    343 		sg = stackgap_init(p, 0);
    344 		bfp = (struct flock *) stackgap_alloc(p, &sg, sizeof *bfp);
    345 		if ((error = copyout(&bfl, bfp, sizeof bfl)) != 0)
    346 			return error;
    347 		SCARG(&fca, fd) = fd;
    348 		SCARG(&fca, cmd) = cmd;
    349 		SCARG(&fca, arg) = bfp;
    350 		return sys_fcntl(l, &fca, retval);
    351 	default:
    352 		return linux_sys_fcntl(l, v, retval);
    353 	}
    354 }
    355 # endif /* !m69k && !amd64  && !COMPAT_LINUX32 */
    356 
    357 #endif /* !alpha */
    358 
    359 /*
    360  * Linux 'readdir' call. This code is mostly taken from the
    361  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
    362  * an attempt has been made to keep it a little cleaner.
    363  *
    364  * The d_off field contains the offset of the next valid entry,
    365  * unless the older Linux getdents(2), which used to have it set
    366  * to the offset of the entry itself. This function also doesn't
    367  * need to deal with the old count == 1 glibc problem.
    368  *
    369  * Read in BSD-style entries, convert them, and copy them out.
    370  *
    371  * Note that this doesn't handle union-mounted filesystems.
    372  */
    373 #ifndef COMPAT_LINUX32
    374 int
    375 linux_sys_getdents64(l, v, retval)
    376 	struct lwp *l;
    377 	void *v;
    378 	register_t *retval;
    379 {
    380 	struct linux_sys_getdents_args /* {
    381 		syscallarg(int) fd;
    382 		syscallarg(struct linux_dirent64 *) dent;
    383 		syscallarg(unsigned int) count;
    384 	} */ *uap = v;
    385 	struct dirent *bdp;
    386 	struct vnode *vp;
    387 	char *inp, *tbuf;		/* BSD-format */
    388 	int len, reclen;		/* BSD-format */
    389 	char *outp;			/* Linux-format */
    390 	int resid, linux_reclen = 0;	/* Linux-format */
    391 	struct file *fp;
    392 	struct uio auio;
    393 	struct iovec aiov;
    394 	struct linux_dirent64 idb;
    395 	off_t off;		/* true file offset */
    396 	int buflen, error, eofflag, nbytes;
    397 	struct vattr va;
    398 	off_t *cookiebuf = NULL, *cookie;
    399 	int ncookies;
    400 
    401 	/* getvnode() will use the descriptor for us */
    402 	if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0)
    403 		return (error);
    404 
    405 	if ((fp->f_flag & FREAD) == 0) {
    406 		error = EBADF;
    407 		goto out1;
    408 	}
    409 
    410 	vp = (struct vnode *)fp->f_data;
    411 	if (vp->v_type != VDIR) {
    412 		error = EINVAL;
    413 		goto out1;
    414 	}
    415 
    416 	if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)))
    417 		goto out1;
    418 
    419 	nbytes = SCARG(uap, count);
    420 	buflen = min(MAXBSIZE, nbytes);
    421 	if (buflen < va.va_blocksize)
    422 		buflen = va.va_blocksize;
    423 	tbuf = malloc(buflen, M_TEMP, M_WAITOK);
    424 
    425 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    426 	off = fp->f_offset;
    427 again:
    428 	aiov.iov_base = tbuf;
    429 	aiov.iov_len = buflen;
    430 	auio.uio_iov = &aiov;
    431 	auio.uio_iovcnt = 1;
    432 	auio.uio_rw = UIO_READ;
    433 	auio.uio_resid = buflen;
    434 	auio.uio_offset = off;
    435 	UIO_SETUP_SYSSPACE(&auio);
    436 	/*
    437          * First we read into the malloc'ed buffer, then
    438          * we massage it into user space, one record at a time.
    439          */
    440 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
    441 	    &ncookies);
    442 	if (error)
    443 		goto out;
    444 
    445 	inp = tbuf;
    446 	outp = (void *)SCARG(uap, dent);
    447 	resid = nbytes;
    448 	if ((len = buflen - auio.uio_resid) == 0)
    449 		goto eof;
    450 
    451 	for (cookie = cookiebuf; len > 0; len -= reclen) {
    452 		bdp = (struct dirent *)inp;
    453 		reclen = bdp->d_reclen;
    454 		if (reclen & 3)
    455 			panic("linux_readdir");
    456 		if (bdp->d_fileno == 0) {
    457 			inp += reclen;	/* it is a hole; squish it out */
    458 			if (cookie)
    459 				off = *cookie++;
    460 			else
    461 				off += reclen;
    462 			continue;
    463 		}
    464 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
    465 		if (reclen > len || resid < linux_reclen) {
    466 			/* entry too big for buffer, so just stop */
    467 			outp++;
    468 			break;
    469 		}
    470 		if (cookie)
    471 			off = *cookie++;	/* each entry points to next */
    472 		else
    473 			off += reclen;
    474 		/*
    475 		 * Massage in place to make a Linux-shaped dirent (otherwise
    476 		 * we have to worry about touching user memory outside of
    477 		 * the copyout() call).
    478 		 */
    479 		idb.d_ino = bdp->d_fileno;
    480 		idb.d_type = bdp->d_type;
    481 		idb.d_off = off;
    482 		idb.d_reclen = (u_short)linux_reclen;
    483 		strcpy(idb.d_name, bdp->d_name);
    484 		if ((error = copyout((void *)&idb, outp, linux_reclen)))
    485 			goto out;
    486 		/* advance past this real entry */
    487 		inp += reclen;
    488 		/* advance output past Linux-shaped entry */
    489 		outp += linux_reclen;
    490 		resid -= linux_reclen;
    491 	}
    492 
    493 	/* if we squished out the whole block, try again */
    494 	if (outp == (void *)SCARG(uap, dent))
    495 		goto again;
    496 	fp->f_offset = off;	/* update the vnode offset */
    497 
    498 eof:
    499 	*retval = nbytes - resid;
    500 out:
    501 	VOP_UNLOCK(vp, 0);
    502 	if (cookiebuf)
    503 		free(cookiebuf, M_TEMP);
    504 	free(tbuf, M_TEMP);
    505 out1:
    506 	FILE_UNUSE(fp, l);
    507 	return error;
    508 }
    509 #endif /* !COMPAT_LINUX32 */
    510