Home | History | Annotate | Line # | Download | only in common
linux_file.c revision 1.88
      1 /*	$NetBSD: linux_file.c,v 1.88 2007/12/08 18:36:07 dsl Exp $	*/
      2 
      3 /*-
      4  * Copyright (c) 1995, 1998 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  * Functions in multiarch:
     41  *	linux_sys_llseek	: linux_llseek.c
     42  */
     43 
     44 #include <sys/cdefs.h>
     45 __KERNEL_RCSID(0, "$NetBSD: linux_file.c,v 1.88 2007/12/08 18:36:07 dsl Exp $");
     46 
     47 #include <sys/param.h>
     48 #include <sys/systm.h>
     49 #include <sys/namei.h>
     50 #include <sys/proc.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/vnode.h>
     60 #include <sys/tty.h>
     61 #include <sys/socketvar.h>
     62 #include <sys/conf.h>
     63 #include <sys/pipe.h>
     64 
     65 #include <sys/syscallargs.h>
     66 #include <sys/vfs_syscalls.h>
     67 
     68 #include <compat/linux/common/linux_types.h>
     69 #include <compat/linux/common/linux_signal.h>
     70 #include <compat/linux/common/linux_fcntl.h>
     71 #include <compat/linux/common/linux_util.h>
     72 #include <compat/linux/common/linux_machdep.h>
     73 #include <compat/linux/common/linux_ipc.h>
     74 #include <compat/linux/common/linux_sem.h>
     75 
     76 #include <compat/linux/linux_syscallargs.h>
     77 
     78 static int linux_to_bsd_ioflags(int);
     79 static int bsd_to_linux_ioflags(int);
     80 static void bsd_to_linux_flock(struct flock *, struct linux_flock *);
     81 static void linux_to_bsd_flock(struct linux_flock *, struct flock *);
     82 #ifndef __amd64__
     83 static void bsd_to_linux_stat(struct stat *, struct linux_stat *);
     84 static int linux_stat1(struct lwp *, void *, register_t *, int);
     85 #endif
     86 
     87 /*
     88  * Some file-related calls are handled here. The usual flag conversion
     89  * an structure conversion is done, and alternate emul path searching.
     90  */
     91 
     92 /*
     93  * The next two functions convert between the Linux and NetBSD values
     94  * of the flags used in open(2) and fcntl(2).
     95  */
     96 static int
     97 linux_to_bsd_ioflags(int lflags)
     98 {
     99 	int res = 0;
    100 
    101 	res |= cvtto_bsd_mask(lflags, LINUX_O_WRONLY, O_WRONLY);
    102 	res |= cvtto_bsd_mask(lflags, LINUX_O_RDONLY, O_RDONLY);
    103 	res |= cvtto_bsd_mask(lflags, LINUX_O_RDWR, O_RDWR);
    104 	res |= cvtto_bsd_mask(lflags, LINUX_O_CREAT, O_CREAT);
    105 	res |= cvtto_bsd_mask(lflags, LINUX_O_EXCL, O_EXCL);
    106 	res |= cvtto_bsd_mask(lflags, LINUX_O_NOCTTY, O_NOCTTY);
    107 	res |= cvtto_bsd_mask(lflags, LINUX_O_TRUNC, O_TRUNC);
    108 	res |= cvtto_bsd_mask(lflags, LINUX_O_NDELAY, O_NDELAY);
    109 	res |= cvtto_bsd_mask(lflags, LINUX_O_SYNC, O_FSYNC);
    110 	res |= cvtto_bsd_mask(lflags, LINUX_FASYNC, O_ASYNC);
    111 	res |= cvtto_bsd_mask(lflags, LINUX_O_APPEND, O_APPEND);
    112 
    113 	return res;
    114 }
    115 
    116 static int
    117 bsd_to_linux_ioflags(int bflags)
    118 {
    119 	int res = 0;
    120 
    121 	res |= cvtto_linux_mask(bflags, O_WRONLY, LINUX_O_WRONLY);
    122 	res |= cvtto_linux_mask(bflags, O_RDONLY, LINUX_O_RDONLY);
    123 	res |= cvtto_linux_mask(bflags, O_RDWR, LINUX_O_RDWR);
    124 	res |= cvtto_linux_mask(bflags, O_CREAT, LINUX_O_CREAT);
    125 	res |= cvtto_linux_mask(bflags, O_EXCL, LINUX_O_EXCL);
    126 	res |= cvtto_linux_mask(bflags, O_NOCTTY, LINUX_O_NOCTTY);
    127 	res |= cvtto_linux_mask(bflags, O_TRUNC, LINUX_O_TRUNC);
    128 	res |= cvtto_linux_mask(bflags, O_NDELAY, LINUX_O_NDELAY);
    129 	res |= cvtto_linux_mask(bflags, O_FSYNC, LINUX_O_SYNC);
    130 	res |= cvtto_linux_mask(bflags, O_ASYNC, LINUX_FASYNC);
    131 	res |= cvtto_linux_mask(bflags, O_APPEND, LINUX_O_APPEND);
    132 
    133 	return res;
    134 }
    135 
    136 /*
    137  * creat(2) is an obsolete function, but it's present as a Linux
    138  * system call, so let's deal with it.
    139  *
    140  * Note: On the Alpha this doesn't really exist in Linux, but it's defined
    141  * in syscalls.master anyway so this doesn't have to be special cased.
    142  *
    143  * Just call open(2) with the TRUNC, CREAT and WRONLY flags.
    144  */
    145 int
    146 linux_sys_creat(struct lwp *l, void *v, register_t *retval)
    147 {
    148 	struct linux_sys_creat_args /* {
    149 		syscallarg(const char *) path;
    150 		syscallarg(int) mode;
    151 	} */ *uap = v;
    152 	struct sys_open_args oa;
    153 
    154 	SCARG(&oa, path) = SCARG(uap, path);
    155 	SCARG(&oa, flags) = O_CREAT | O_TRUNC | O_WRONLY;
    156 	SCARG(&oa, mode) = SCARG(uap, mode);
    157 
    158 	return sys_open(l, &oa, retval);
    159 }
    160 
    161 /*
    162  * open(2). Take care of the different flag values, and let the
    163  * NetBSD syscall do the real work. See if this operation
    164  * gives the current process a controlling terminal.
    165  * (XXX is this necessary?)
    166  */
    167 int
    168 linux_sys_open(struct lwp *l, void *v, register_t *retval)
    169 {
    170 	struct linux_sys_open_args /* {
    171 		syscallarg(const char *) path;
    172 		syscallarg(int) flags;
    173 		syscallarg(int) mode;
    174 	} */ *uap = v;
    175 	struct proc *p = l->l_proc;
    176 	int error, fl;
    177 	struct sys_open_args boa;
    178 
    179 	fl = linux_to_bsd_ioflags(SCARG(uap, flags));
    180 
    181 	SCARG(&boa, path) = SCARG(uap, path);
    182 	SCARG(&boa, flags) = fl;
    183 	SCARG(&boa, mode) = SCARG(uap, mode);
    184 
    185 	if ((error = sys_open(l, &boa, retval)))
    186 		return error;
    187 
    188 	/*
    189 	 * this bit from sunos_misc.c (and svr4_fcntl.c).
    190 	 * If we are a session leader, and we don't have a controlling
    191 	 * terminal yet, and the O_NOCTTY flag is not set, try to make
    192 	 * this the controlling terminal.
    193 	 */
    194         if (!(fl & O_NOCTTY) && SESS_LEADER(p) && !(p->p_lflag & PL_CONTROLT)) {
    195                 struct filedesc *fdp = p->p_fd;
    196                 struct file     *fp;
    197 
    198 		fp = fd_getfile(fdp, *retval);
    199 
    200                 /* ignore any error, just give it a try */
    201                 if (fp != NULL) {
    202 			FILE_USE(fp);
    203 			if (fp->f_type == DTYPE_VNODE) {
    204 				(fp->f_ops->fo_ioctl) (fp, TIOCSCTTY,
    205 				    (void *) 0, l);
    206 			}
    207 			FILE_UNUSE(fp, l);
    208 		}
    209         }
    210 	return 0;
    211 }
    212 
    213 /*
    214  * The next two functions take care of converting the flock
    215  * structure back and forth between Linux and NetBSD format.
    216  * The only difference in the structures is the order of
    217  * the fields, and the 'whence' value.
    218  */
    219 static void
    220 bsd_to_linux_flock(struct flock *bfp, struct linux_flock *lfp)
    221 {
    222 
    223 	lfp->l_start = bfp->l_start;
    224 	lfp->l_len = bfp->l_len;
    225 	lfp->l_pid = bfp->l_pid;
    226 	lfp->l_whence = bfp->l_whence;
    227 	switch (bfp->l_type) {
    228 	case F_RDLCK:
    229 		lfp->l_type = LINUX_F_RDLCK;
    230 		break;
    231 	case F_UNLCK:
    232 		lfp->l_type = LINUX_F_UNLCK;
    233 		break;
    234 	case F_WRLCK:
    235 		lfp->l_type = LINUX_F_WRLCK;
    236 		break;
    237 	}
    238 }
    239 
    240 static void
    241 linux_to_bsd_flock(struct linux_flock *lfp, struct flock *bfp)
    242 {
    243 
    244 	bfp->l_start = lfp->l_start;
    245 	bfp->l_len = lfp->l_len;
    246 	bfp->l_pid = lfp->l_pid;
    247 	bfp->l_whence = lfp->l_whence;
    248 	switch (lfp->l_type) {
    249 	case LINUX_F_RDLCK:
    250 		bfp->l_type = F_RDLCK;
    251 		break;
    252 	case LINUX_F_UNLCK:
    253 		bfp->l_type = F_UNLCK;
    254 		break;
    255 	case LINUX_F_WRLCK:
    256 		bfp->l_type = F_WRLCK;
    257 		break;
    258 	}
    259 }
    260 
    261 /*
    262  * Most actions in the fcntl() call are straightforward; simply
    263  * pass control to the NetBSD system call. A few commands need
    264  * conversions after the actual system call has done its work,
    265  * because the flag values and lock structure are different.
    266  */
    267 int
    268 linux_sys_fcntl(struct lwp *l, void *v, register_t *retval)
    269 {
    270 	struct linux_sys_fcntl_args /* {
    271 		syscallarg(int) fd;
    272 		syscallarg(int) cmd;
    273 		syscallarg(void *) arg;
    274 	} */ *uap = v;
    275 	struct proc *p = l->l_proc;
    276 	int fd, cmd, error;
    277 	u_long val;
    278 	void *arg;
    279 	struct linux_flock lfl;
    280 	struct flock bfl;
    281 	struct sys_fcntl_args fca;
    282 	struct filedesc *fdp;
    283 	struct file *fp;
    284 	struct vnode *vp;
    285 	struct vattr va;
    286 	const struct cdevsw *cdev;
    287 	long pgid;
    288 	struct pgrp *pgrp;
    289 	struct tty *tp, *(*d_tty)(dev_t);
    290 
    291 	fd = SCARG(uap, fd);
    292 	cmd = SCARG(uap, cmd);
    293 	arg = (void *) SCARG(uap, arg);
    294 
    295 	switch (cmd) {
    296 	case LINUX_F_DUPFD:
    297 		cmd = F_DUPFD;
    298 		break;
    299 	case LINUX_F_GETFD:
    300 		cmd = F_GETFD;
    301 		break;
    302 	case LINUX_F_SETFD:
    303 		cmd = F_SETFD;
    304 		break;
    305 	case LINUX_F_GETFL:
    306 		SCARG(&fca, fd) = fd;
    307 		SCARG(&fca, cmd) = F_GETFL;
    308 		SCARG(&fca, arg) = arg;
    309 		if ((error = sys_fcntl(l, &fca, retval)))
    310 			return error;
    311 		retval[0] = bsd_to_linux_ioflags(retval[0]);
    312 		return 0;
    313 	case LINUX_F_SETFL: {
    314 		struct file	*fp1 = NULL;
    315 
    316 		val = linux_to_bsd_ioflags((unsigned long)SCARG(uap, arg));
    317 		/*
    318 		 * Linux seems to have same semantics for sending SIGIO to the
    319 		 * read side of socket, but slightly different semantics
    320 		 * for SIGIO to the write side.  Rather than sending the SIGIO
    321 		 * every time it's possible to write (directly) more data, it
    322 		 * only sends SIGIO if last write(2) failed due to insufficient
    323 		 * memory to hold the data. This is compatible enough
    324 		 * with NetBSD semantics to not do anything about the
    325 		 * difference.
    326 		 *
    327 		 * Linux does NOT send SIGIO for pipes. Deal with socketpair
    328 		 * ones and DTYPE_PIPE ones. For these, we don't set
    329 		 * the underlying flags (we don't pass O_ASYNC flag down
    330 		 * to sys_fcntl()), but set the FASYNC flag for file descriptor,
    331 		 * so that F_GETFL would report the ASYNC i/o is on.
    332 		 */
    333 		if (val & O_ASYNC) {
    334 			if (((fp1 = fd_getfile(p->p_fd, fd)) == NULL))
    335 			    return (EBADF);
    336 
    337 			FILE_USE(fp1);
    338 
    339 			if (((fp1->f_type == DTYPE_SOCKET) && fp1->f_data
    340 			      && ((struct socket *)fp1->f_data)->so_state & SS_ISAPIPE)
    341 			    || (fp1->f_type == DTYPE_PIPE))
    342 				val &= ~O_ASYNC;
    343 			else {
    344 				/* not a pipe, do not modify anything */
    345 				FILE_UNUSE(fp1, l);
    346 				fp1 = NULL;
    347 			}
    348 		}
    349 
    350 		SCARG(&fca, fd) = fd;
    351 		SCARG(&fca, cmd) = F_SETFL;
    352 		SCARG(&fca, arg) = (void *) val;
    353 
    354 		error = sys_fcntl(l, &fca, retval);
    355 
    356 		/* Now set the FASYNC flag for pipes */
    357 		if (fp1) {
    358 			if (!error)
    359 				fp1->f_flag |= FASYNC;
    360 			FILE_UNUSE(fp1, l);
    361 		}
    362 
    363 		return (error);
    364 	    }
    365 	case LINUX_F_GETLK:
    366 		if ((error = copyin(arg, &lfl, sizeof lfl)))
    367 			return error;
    368 		linux_to_bsd_flock(&lfl, &bfl);
    369 		error = do_fcntl_lock(l, fd, F_GETLK, &bfl);
    370 		if (error)
    371 			return error;
    372 		bsd_to_linux_flock(&bfl, &lfl);
    373 		return copyout(&lfl, arg, sizeof lfl);
    374 
    375 	case LINUX_F_SETLK:
    376 	case LINUX_F_SETLKW:
    377 		cmd = (cmd == LINUX_F_SETLK ? F_SETLK : F_SETLKW);
    378 		if ((error = copyin(arg, &lfl, sizeof lfl)))
    379 			return error;
    380 		linux_to_bsd_flock(&lfl, &bfl);
    381 		return do_fcntl_lock(l, fd, cmd, &bfl);
    382 
    383 	case LINUX_F_SETOWN:
    384 	case LINUX_F_GETOWN:
    385 		/*
    386 		 * We need to route fcntl() for tty descriptors around normal
    387 		 * fcntl(), since NetBSD tty TIOC{G,S}PGRP semantics is too
    388 		 * restrictive for Linux F_{G,S}ETOWN. For non-tty descriptors,
    389 		 * this is not a problem.
    390 		 */
    391 		fdp = p->p_fd;
    392 		if ((fp = fd_getfile(fdp, fd)) == NULL)
    393 			return EBADF;
    394 		FILE_USE(fp);
    395 
    396 		/* Check it's a character device vnode */
    397 		if (fp->f_type != DTYPE_VNODE
    398 		    || (vp = (struct vnode *)fp->f_data) == NULL
    399 		    || vp->v_type != VCHR) {
    400 			FILE_UNUSE(fp, l);
    401 
    402 	    not_tty:
    403 			/* Not a tty, proceed with common fcntl() */
    404 			cmd = cmd == LINUX_F_SETOWN ? F_SETOWN : F_GETOWN;
    405 			break;
    406 		}
    407 
    408 		error = VOP_GETATTR(vp, &va, l->l_cred);
    409 
    410 		FILE_UNUSE(fp, l);
    411 
    412 		if (error)
    413 			return error;
    414 
    415 		cdev = cdevsw_lookup(va.va_rdev);
    416 		if (cdev == NULL)
    417 			return (ENXIO);
    418 		d_tty = cdev->d_tty;
    419 		if (!d_tty || (!(tp = (*d_tty)(va.va_rdev))))
    420 			goto not_tty;
    421 
    422 		/* set tty pg_id appropriately */
    423 		if (cmd == LINUX_F_GETOWN) {
    424 			retval[0] = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID;
    425 			return 0;
    426 		}
    427 		mutex_enter(&proclist_lock);
    428 		if ((long)arg <= 0) {
    429 			pgid = -(long)arg;
    430 		} else {
    431 			struct proc *p1 = p_find((long)arg, PFIND_LOCKED | PFIND_UNLOCK_FAIL);
    432 			if (p1 == NULL)
    433 				return (ESRCH);
    434 			pgid = (long)p1->p_pgrp->pg_id;
    435 		}
    436 		pgrp = pg_find(pgid, PFIND_LOCKED);
    437 		if (pgrp == NULL || pgrp->pg_session != p->p_session) {
    438 			mutex_exit(&proclist_lock);
    439 			return EPERM;
    440 		}
    441 		tp->t_pgrp = pgrp;
    442 		mutex_exit(&proclist_lock);
    443 		return 0;
    444 
    445 	default:
    446 		return EOPNOTSUPP;
    447 	}
    448 
    449 	SCARG(&fca, fd) = fd;
    450 	SCARG(&fca, cmd) = cmd;
    451 	SCARG(&fca, arg) = arg;
    452 
    453 	return sys_fcntl(l, &fca, retval);
    454 }
    455 
    456 #if !defined(__amd64__)
    457 /*
    458  * Convert a NetBSD stat structure to a Linux stat structure.
    459  * Only the order of the fields and the padding in the structure
    460  * is different. linux_fakedev is a machine-dependent function
    461  * which optionally converts device driver major/minor numbers
    462  * (XXX horrible, but what can you do against code that compares
    463  * things against constant major device numbers? sigh)
    464  */
    465 static void
    466 bsd_to_linux_stat(struct stat *bsp, struct linux_stat *lsp)
    467 {
    468 
    469 	lsp->lst_dev     = linux_fakedev(bsp->st_dev, 0);
    470 	lsp->lst_ino     = bsp->st_ino;
    471 	lsp->lst_mode    = (linux_mode_t)bsp->st_mode;
    472 	if (bsp->st_nlink >= (1 << 15))
    473 		lsp->lst_nlink = (1 << 15) - 1;
    474 	else
    475 		lsp->lst_nlink = (linux_nlink_t)bsp->st_nlink;
    476 	lsp->lst_uid     = bsp->st_uid;
    477 	lsp->lst_gid     = bsp->st_gid;
    478 	lsp->lst_rdev    = linux_fakedev(bsp->st_rdev, 1);
    479 	lsp->lst_size    = bsp->st_size;
    480 	lsp->lst_blksize = bsp->st_blksize;
    481 	lsp->lst_blocks  = bsp->st_blocks;
    482 	lsp->lst_atime   = bsp->st_atime;
    483 	lsp->lst_mtime   = bsp->st_mtime;
    484 	lsp->lst_ctime   = bsp->st_ctime;
    485 #ifdef LINUX_STAT_HAS_NSEC
    486 	lsp->lst_atime_nsec   = bsp->st_atimensec;
    487 	lsp->lst_mtime_nsec   = bsp->st_mtimensec;
    488 	lsp->lst_ctime_nsec   = bsp->st_ctimensec;
    489 #endif
    490 }
    491 
    492 /*
    493  * The stat functions below are plain sailing. stat and lstat are handled
    494  * by one function to avoid code duplication.
    495  */
    496 int
    497 linux_sys_fstat(struct lwp *l, void *v, register_t *retval)
    498 {
    499 	struct linux_sys_fstat_args /* {
    500 		syscallarg(int) fd;
    501 		syscallarg(linux_stat *) sp;
    502 	} */ *uap = v;
    503 	struct linux_stat tmplst;
    504 	struct stat tmpst;
    505 	int error;
    506 
    507 	error = do_sys_fstat(l, SCARG(uap, fd), &tmpst);
    508 	if (error != 0)
    509 		return error;
    510 	bsd_to_linux_stat(&tmpst, &tmplst);
    511 
    512 	return copyout(&tmplst, SCARG(uap, sp), sizeof tmplst);
    513 }
    514 
    515 static int
    516 linux_stat1(struct lwp *l, void *v, register_t *retval, int flags)
    517 {
    518 	struct linux_stat tmplst;
    519 	struct stat tmpst;
    520 	int error;
    521 	struct linux_sys_stat_args *uap = v;
    522 
    523 	error = do_sys_stat(l, SCARG(uap, path), flags, &tmpst);
    524 	if (error != 0)
    525 		return error;
    526 
    527 	bsd_to_linux_stat(&tmpst, &tmplst);
    528 
    529 	return copyout(&tmplst, SCARG(uap, sp), sizeof tmplst);
    530 }
    531 
    532 int
    533 linux_sys_stat(struct lwp *l, void *v, register_t *retval)
    534 {
    535 	struct linux_sys_stat_args /* {
    536 		syscallarg(const char *) path;
    537 		syscallarg(struct linux_stat *) sp;
    538 	} */ *uap = v;
    539 
    540 	return linux_stat1(l, uap, retval, FOLLOW);
    541 }
    542 
    543 /* Note: this is "newlstat" in the Linux sources */
    544 /*	(we don't bother with the old lstat currently) */
    545 int
    546 linux_sys_lstat(struct lwp *l, void *v, register_t *retval)
    547 {
    548 	struct linux_sys_lstat_args /* {
    549 		syscallarg(const char *) path;
    550 		syscallarg(struct linux_stat *) sp;
    551 	} */ *uap = v;
    552 
    553 	return linux_stat1(l, uap, retval, NOFOLLOW);
    554 }
    555 #endif /* !__amd64__ */
    556 
    557 /*
    558  * The following syscalls are mostly here because of the alternate path check.
    559  */
    560 int
    561 linux_sys_unlink(l, v, retval)
    562 	struct lwp *l;
    563 	void *v;
    564 	register_t *retval;
    565 
    566 {
    567 	struct linux_sys_unlink_args /* {
    568 		syscallarg(const char *) path;
    569 	} */ *uap = v;
    570 	int error;
    571 	struct nameidata nd;
    572 
    573 	error = sys_unlink(l, uap, retval);
    574 	if (error != EPERM)
    575 		return (error);
    576 
    577 	/*
    578 	 * Linux returns EISDIR if unlink(2) is called on a directory.
    579 	 * We return EPERM in such cases. To emulate correct behaviour,
    580 	 * check if the path points to directory and return EISDIR if this
    581 	 * is the case.
    582 	 */
    583 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
    584 	    SCARG(uap, path), l);
    585 	if (namei(&nd) == 0) {
    586 		struct stat sb;
    587 
    588 		if (vn_stat(nd.ni_vp, &sb, l) == 0
    589 		    && S_ISDIR(sb.st_mode))
    590 			error = EISDIR;
    591 
    592 		vput(nd.ni_vp);
    593 	}
    594 
    595 	return (error);
    596 }
    597 
    598 int
    599 linux_sys_mknod(struct lwp *l, void *v, register_t *retval)
    600 {
    601 	struct linux_sys_mknod_args /* {
    602 		syscallarg(const char *) path;
    603 		syscallarg(int) mode;
    604 		syscallarg(int) dev;
    605 	} */ *uap = v;
    606 
    607 	/*
    608 	 * BSD handles FIFOs separately
    609 	 */
    610 	if (S_ISFIFO(SCARG(uap, mode))) {
    611 		struct sys_mkfifo_args bma;
    612 
    613 		SCARG(&bma, path) = SCARG(uap, path);
    614 		SCARG(&bma, mode) = SCARG(uap, mode);
    615 		return sys_mkfifo(l, &bma, retval);
    616 	} else {
    617 		struct sys_mknod_args bma;
    618 
    619 		SCARG(&bma, path) = SCARG(uap, path);
    620 		SCARG(&bma, mode) = SCARG(uap, mode);
    621 		/*
    622 		 * Linux device numbers uses 8 bits for minor and 8 bits
    623 		 * for major. Due to how we map our major and minor,
    624 		 * this just fits into our dev_t. Just mask off the
    625 		 * upper 16bit to remove any random junk.
    626 		 */
    627 		SCARG(&bma, dev) = SCARG(uap, dev) & 0xffff;
    628 		return sys_mknod(l, &bma, retval);
    629 	}
    630 }
    631 
    632 #if defined(__i386__) || defined(__m68k__) || \
    633     defined(__arm__)
    634 int
    635 linux_sys_chown16(struct lwp *l, void *v, register_t *retval)
    636 {
    637 	struct linux_sys_chown16_args /* {
    638 		syscallarg(const char *) path;
    639 		syscallarg(int) uid;
    640 		syscallarg(int) gid;
    641 	} */ *uap = v;
    642 	struct sys___posix_chown_args bca;
    643 
    644 	SCARG(&bca, path) = SCARG(uap, path);
    645 	SCARG(&bca, uid) = ((linux_uid_t)SCARG(uap, uid) == (linux_uid_t)-1) ?
    646 		(uid_t)-1 : SCARG(uap, uid);
    647 	SCARG(&bca, gid) = ((linux_gid_t)SCARG(uap, gid) == (linux_gid_t)-1) ?
    648 		(gid_t)-1 : SCARG(uap, gid);
    649 
    650 	return sys___posix_chown(l, &bca, retval);
    651 }
    652 
    653 int
    654 linux_sys_fchown16(struct lwp *l, void *v, register_t *retval)
    655 {
    656 	struct linux_sys_fchown16_args /* {
    657 		syscallarg(int) fd;
    658 		syscallarg(int) uid;
    659 		syscallarg(int) gid;
    660 	} */ *uap = v;
    661 	struct sys___posix_fchown_args bfa;
    662 
    663 	SCARG(&bfa, fd) = SCARG(uap, fd);
    664 	SCARG(&bfa, uid) = ((linux_uid_t)SCARG(uap, uid) == (linux_uid_t)-1) ?
    665 		(uid_t)-1 : SCARG(uap, uid);
    666 	SCARG(&bfa, gid) = ((linux_gid_t)SCARG(uap, gid) == (linux_gid_t)-1) ?
    667 		(gid_t)-1 : SCARG(uap, gid);
    668 
    669 	return sys___posix_fchown(l, &bfa, retval);
    670 }
    671 
    672 int
    673 linux_sys_lchown16(struct lwp *l, void *v, register_t *retval)
    674 {
    675 	struct linux_sys_lchown16_args /* {
    676 		syscallarg(char *) path;
    677 		syscallarg(int) uid;
    678 		syscallarg(int) gid;
    679 	} */ *uap = v;
    680 	struct sys___posix_lchown_args bla;
    681 
    682 	SCARG(&bla, path) = SCARG(uap, path);
    683 	SCARG(&bla, uid) = ((linux_uid_t)SCARG(uap, uid) == (linux_uid_t)-1) ?
    684 		(uid_t)-1 : SCARG(uap, uid);
    685 	SCARG(&bla, gid) = ((linux_gid_t)SCARG(uap, gid) == (linux_gid_t)-1) ?
    686 		(gid_t)-1 : SCARG(uap, gid);
    687 
    688 	return sys___posix_lchown(l, &bla, retval);
    689 }
    690 #endif /* __i386__ || __m68k__ || __arm__ || __amd64__ */
    691 
    692 /*
    693  * This is just fsync() for now (just as it is in the Linux kernel)
    694  * Note: this is not implemented under Linux on Alpha and Arm
    695  *	but should still be defined in our syscalls.master.
    696  *	(syscall #148 on the arm)
    697  */
    698 int
    699 linux_sys_fdatasync(struct lwp *l, void *v, register_t *retval)
    700 {
    701 #ifdef notdef
    702 	struct linux_sys_fdatasync_args /* {
    703 		syscallarg(int) fd;
    704 	} */ *uap = v;
    705 #endif
    706 	return sys_fsync(l, v, retval);
    707 }
    708 
    709 /*
    710  * pread(2).
    711  */
    712 int
    713 linux_sys_pread(struct lwp *l, void *v, register_t *retval)
    714 {
    715 	struct linux_sys_pread_args /* {
    716 		syscallarg(int) fd;
    717 		syscallarg(void *) buf;
    718 		syscallarg(size_t) nbyte;
    719 		syscallarg(linux_off_t) offset;
    720 	} */ *uap = v;
    721 	struct sys_pread_args pra;
    722 
    723 	SCARG(&pra, fd) = SCARG(uap, fd);
    724 	SCARG(&pra, buf) = SCARG(uap, buf);
    725 	SCARG(&pra, nbyte) = SCARG(uap, nbyte);
    726 	SCARG(&pra, offset) = SCARG(uap, offset);
    727 
    728 	return sys_pread(l, &pra, retval);
    729 }
    730 
    731 /*
    732  * pwrite(2).
    733  */
    734 int
    735 linux_sys_pwrite(struct lwp *l, void *v, register_t *retval)
    736 {
    737 	struct linux_sys_pwrite_args /* {
    738 		syscallarg(int) fd;
    739 		syscallarg(void *) buf;
    740 		syscallarg(size_t) nbyte;
    741 		syscallarg(linux_off_t) offset;
    742 	} */ *uap = v;
    743 	struct sys_pwrite_args pra;
    744 
    745 	SCARG(&pra, fd) = SCARG(uap, fd);
    746 	SCARG(&pra, buf) = SCARG(uap, buf);
    747 	SCARG(&pra, nbyte) = SCARG(uap, nbyte);
    748 	SCARG(&pra, offset) = SCARG(uap, offset);
    749 
    750 	return sys_pwrite(l, &pra, retval);
    751 }
    752 
    753 #define LINUX_NOT_SUPPORTED(fun) \
    754 int \
    755 fun(struct lwp *l, void *v, register_t *retval) \
    756 { \
    757 	return EOPNOTSUPP; \
    758 }
    759 
    760 LINUX_NOT_SUPPORTED(linux_sys_setxattr)
    761 LINUX_NOT_SUPPORTED(linux_sys_lsetxattr)
    762 LINUX_NOT_SUPPORTED(linux_sys_fsetxattr)
    763 
    764 LINUX_NOT_SUPPORTED(linux_sys_getxattr)
    765 LINUX_NOT_SUPPORTED(linux_sys_lgetxattr)
    766 LINUX_NOT_SUPPORTED(linux_sys_fgetxattr)
    767 
    768 LINUX_NOT_SUPPORTED(linux_sys_listxattr)
    769 LINUX_NOT_SUPPORTED(linux_sys_llistxattr)
    770 LINUX_NOT_SUPPORTED(linux_sys_flistxattr)
    771 
    772 LINUX_NOT_SUPPORTED(linux_sys_removexattr)
    773 LINUX_NOT_SUPPORTED(linux_sys_lremovexattr)
    774 LINUX_NOT_SUPPORTED(linux_sys_fremovexattr)
    775