Home | History | Annotate | Line # | Download | only in common
linux_misc.c revision 1.87
      1 /*	$NetBSD: linux_misc.c,v 1.87 2001/03/30 17:16:34 jdolecek Exp $	*/
      2 
      3 /*-
      4  * Copyright (c) 1995, 1998, 1999 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; by Jason R. Thorpe
      9  * of the Numerical Aerospace Simulation Facility, NASA Ames Research Center.
     10  *
     11  * Redistribution and use in source and binary forms, with or without
     12  * modification, are permitted provided that the following conditions
     13  * are met:
     14  * 1. Redistributions of source code must retain the above copyright
     15  *    notice, this list of conditions and the following disclaimer.
     16  * 2. Redistributions in binary form must reproduce the above copyright
     17  *    notice, this list of conditions and the following disclaimer in the
     18  *    documentation and/or other materials provided with the distribution.
     19  * 3. All advertising materials mentioning features or use of this software
     20  *    must display the following acknowledgement:
     21  *	This product includes software developed by the NetBSD
     22  *	Foundation, Inc. and its contributors.
     23  * 4. Neither the name of The NetBSD Foundation nor the names of its
     24  *    contributors may be used to endorse or promote products derived
     25  *    from this software without specific prior written permission.
     26  *
     27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     37  * POSSIBILITY OF SUCH DAMAGE.
     38  */
     39 
     40 /*
     41  * Linux compatibility module. Try to deal with various Linux system calls.
     42  */
     43 
     44 /*
     45  * These functions have been moved to multiarch to allow
     46  * selection of which machines include them to be
     47  * determined by the individual files.linux_<arch> files.
     48  *
     49  * Function in multiarch:
     50  *	linux_sys_break			: linux_break.c
     51  *	linux_sys_alarm			: linux_misc_notalpha.c
     52  *	linux_sys_getresgid		: linux_misc_notalpha.c
     53  *	linux_sys_nice			: linux_misc_notalpha.c
     54  *	linux_sys_readdir		: linux_misc_notalpha.c
     55  *	linux_sys_setresgid		: linux_misc_notalpha.c
     56  *	linux_sys_time			: linux_misc_notalpha.c
     57  *	linux_sys_utime			: linux_misc_notalpha.c
     58  *	linux_sys_waitpid		: linux_misc_notalpha.c
     59  *	linux_sys_old_mmap		: linux_oldmmap.c
     60  *	linux_sys_oldolduname		: linux_oldolduname.c
     61  *	linux_sys_oldselect		: linux_oldselect.c
     62  *	linux_sys_olduname		: linux_olduname.c
     63  *	linux_sys_pipe			: linux_pipe.c
     64  */
     65 
     66 #include <sys/param.h>
     67 #include <sys/systm.h>
     68 #include <sys/namei.h>
     69 #include <sys/proc.h>
     70 #include <sys/dirent.h>
     71 #include <sys/file.h>
     72 #include <sys/stat.h>
     73 #include <sys/filedesc.h>
     74 #include <sys/ioctl.h>
     75 #include <sys/kernel.h>
     76 #include <sys/malloc.h>
     77 #include <sys/mbuf.h>
     78 #include <sys/mman.h>
     79 #include <sys/mount.h>
     80 #include <sys/reboot.h>
     81 #include <sys/resource.h>
     82 #include <sys/resourcevar.h>
     83 #include <sys/signal.h>
     84 #include <sys/signalvar.h>
     85 #include <sys/socket.h>
     86 #include <sys/time.h>
     87 #include <sys/times.h>
     88 #include <sys/vnode.h>
     89 #include <sys/uio.h>
     90 #include <sys/wait.h>
     91 #include <sys/utsname.h>
     92 #include <sys/unistd.h>
     93 #include <sys/swap.h>		/* for SWAP_ON */
     94 #include <sys/sysctl.h>		/* for KERN_DOMAINNAME */
     95 
     96 #include <sys/ptrace.h>
     97 #include <machine/ptrace.h>
     98 
     99 #include <sys/syscallargs.h>
    100 
    101 #include <compat/linux/common/linux_types.h>
    102 #include <compat/linux/common/linux_signal.h>
    103 
    104 #include <compat/linux/linux_syscallargs.h>
    105 
    106 #include <compat/linux/common/linux_fcntl.h>
    107 #include <compat/linux/common/linux_mmap.h>
    108 #include <compat/linux/common/linux_dirent.h>
    109 #include <compat/linux/common/linux_util.h>
    110 #include <compat/linux/common/linux_misc.h>
    111 #include <compat/linux/common/linux_ptrace.h>
    112 #include <compat/linux/common/linux_reboot.h>
    113 #include <compat/linux/common/linux_emuldata.h>
    114 
    115 const int linux_ptrace_request_map[] = {
    116 	LINUX_PTRACE_TRACEME,	PT_TRACE_ME,
    117 	LINUX_PTRACE_PEEKTEXT,	PT_READ_I,
    118 	LINUX_PTRACE_PEEKDATA,	PT_READ_D,
    119 	LINUX_PTRACE_POKETEXT,	PT_WRITE_I,
    120 	LINUX_PTRACE_POKEDATA,	PT_WRITE_D,
    121 	LINUX_PTRACE_CONT,	PT_CONTINUE,
    122 	LINUX_PTRACE_KILL,	PT_KILL,
    123 	LINUX_PTRACE_ATTACH,	PT_ATTACH,
    124 	LINUX_PTRACE_DETACH,	PT_DETACH,
    125 #ifdef PT_STEP
    126 	LINUX_PTRACE_SINGLESTEP,	PT_STEP,
    127 #endif
    128 	-1
    129 };
    130 
    131 /* Local linux_misc.c functions: */
    132 static void bsd_to_linux_statfs __P((struct statfs *, struct linux_statfs *));
    133 
    134 /*
    135  * The information on a terminated (or stopped) process needs
    136  * to be converted in order for Linux binaries to get a valid signal
    137  * number out of it.
    138  */
    139 void
    140 bsd_to_linux_wstat(st)
    141 	int *st;
    142 {
    143 
    144 	int sig;
    145 
    146 	if (WIFSIGNALED(*st)) {
    147 		sig = WTERMSIG(*st);
    148 		if (sig >= 0 && sig < NSIG)
    149 			*st= (*st& ~0177) | native_to_linux_sig[sig];
    150 	} else if (WIFSTOPPED(*st)) {
    151 		sig = WSTOPSIG(*st);
    152 		if (sig >= 0 && sig < NSIG)
    153 			*st = (*st & ~0xff00) | (native_to_linux_sig[sig] << 8);
    154 	}
    155 }
    156 
    157 /*
    158  * This is very much the same as waitpid()
    159  */
    160 int
    161 linux_sys_wait4(p, v, retval)
    162 	struct proc *p;
    163 	void *v;
    164 	register_t *retval;
    165 {
    166 	struct linux_sys_wait4_args /* {
    167 		syscallarg(int) pid;
    168 		syscallarg(int *) status;
    169 		syscallarg(int) options;
    170 		syscallarg(struct rusage *) rusage;
    171 	} */ *uap = v;
    172 	struct sys_wait4_args w4a;
    173 	int error, *status, tstat, options, linux_options;
    174 	caddr_t sg;
    175 
    176 	if (SCARG(uap, status) != NULL) {
    177 		sg = stackgap_init(p->p_emul);
    178 		status = (int *) stackgap_alloc(&sg, sizeof *status);
    179 	} else
    180 		status = NULL;
    181 
    182 	linux_options = SCARG(uap, options);
    183 	options = 0;
    184 	if (linux_options &
    185 	    ~(LINUX_WAIT4_WNOHANG|LINUX_WAIT4_WUNTRACED|LINUX_WAIT4_WCLONE))
    186 		return (EINVAL);
    187 
    188 	if (linux_options & LINUX_WAIT4_WNOHANG)
    189 		options |= WNOHANG;
    190 	if (linux_options & LINUX_WAIT4_WUNTRACED)
    191 		options |= WUNTRACED;
    192 	if (linux_options & LINUX_WAIT4_WCLONE)
    193 		options |= WALTSIG;
    194 
    195 	SCARG(&w4a, pid) = SCARG(uap, pid);
    196 	SCARG(&w4a, status) = status;
    197 	SCARG(&w4a, options) = options;
    198 	SCARG(&w4a, rusage) = SCARG(uap, rusage);
    199 
    200 	if ((error = sys_wait4(p, &w4a, retval)))
    201 		return error;
    202 
    203 	sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD);
    204 
    205 	if (status != NULL) {
    206 		if ((error = copyin(status, &tstat, sizeof tstat)))
    207 			return error;
    208 
    209 		bsd_to_linux_wstat(&tstat);
    210 		return copyout(&tstat, SCARG(uap, status), sizeof tstat);
    211 	}
    212 
    213 	return 0;
    214 }
    215 
    216 /*
    217  * Linux brk(2). The check if the new address is >= the old one is
    218  * done in the kernel in Linux. NetBSD does it in the library.
    219  */
    220 int
    221 linux_sys_brk(p, v, retval)
    222 	struct proc *p;
    223 	void *v;
    224 	register_t *retval;
    225 {
    226 	struct linux_sys_brk_args /* {
    227 		syscallarg(char *) nsize;
    228 	} */ *uap = v;
    229 	char *nbrk = SCARG(uap, nsize);
    230 	struct sys_obreak_args oba;
    231 	struct vmspace *vm = p->p_vmspace;
    232 	struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata;
    233 
    234 	SCARG(&oba, nsize) = nbrk;
    235 
    236 	if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(p, &oba, retval) == 0)
    237 		ed->p_break = (char*)nbrk;
    238 	else
    239 		nbrk = ed->p_break;
    240 
    241 	retval[0] = (register_t)nbrk;
    242 
    243 	return 0;
    244 }
    245 
    246 /*
    247  * Convert BSD statfs structure to Linux statfs structure.
    248  * The Linux structure has less fields, and it also wants
    249  * the length of a name in a dir entry in a field, which
    250  * we fake (probably the wrong way).
    251  */
    252 static void
    253 bsd_to_linux_statfs(bsp, lsp)
    254 	struct statfs *bsp;
    255 	struct linux_statfs *lsp;
    256 {
    257 
    258 	lsp->l_ftype = bsp->f_type;
    259 	lsp->l_fbsize = bsp->f_bsize;
    260 	lsp->l_fblocks = bsp->f_blocks;
    261 	lsp->l_fbfree = bsp->f_bfree;
    262 	lsp->l_fbavail = bsp->f_bavail;
    263 	lsp->l_ffiles = bsp->f_files;
    264 	lsp->l_fffree = bsp->f_ffree;
    265 	lsp->l_ffsid.val[0] = bsp->f_fsid.val[0];
    266 	lsp->l_ffsid.val[1] = bsp->f_fsid.val[1];
    267 	lsp->l_fnamelen = MAXNAMLEN;	/* XXX */
    268 }
    269 
    270 /*
    271  * Implement the fs stat functions. Straightforward.
    272  */
    273 int
    274 linux_sys_statfs(p, v, retval)
    275 	struct proc *p;
    276 	void *v;
    277 	register_t *retval;
    278 {
    279 	struct linux_sys_statfs_args /* {
    280 		syscallarg(const char *) path;
    281 		syscallarg(struct linux_statfs *) sp;
    282 	} */ *uap = v;
    283 	struct statfs btmp, *bsp;
    284 	struct linux_statfs ltmp;
    285 	struct sys_statfs_args bsa;
    286 	caddr_t sg;
    287 	int error;
    288 
    289 	sg = stackgap_init(p->p_emul);
    290 	bsp = (struct statfs *) stackgap_alloc(&sg, sizeof (struct statfs));
    291 
    292 	CHECK_ALT_EXIST(p, &sg, SCARG(uap, path));
    293 
    294 	SCARG(&bsa, path) = SCARG(uap, path);
    295 	SCARG(&bsa, buf) = bsp;
    296 
    297 	if ((error = sys_statfs(p, &bsa, retval)))
    298 		return error;
    299 
    300 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
    301 		return error;
    302 
    303 	bsd_to_linux_statfs(&btmp, &ltmp);
    304 
    305 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
    306 }
    307 
    308 int
    309 linux_sys_fstatfs(p, v, retval)
    310 	struct proc *p;
    311 	void *v;
    312 	register_t *retval;
    313 {
    314 	struct linux_sys_fstatfs_args /* {
    315 		syscallarg(int) fd;
    316 		syscallarg(struct linux_statfs *) sp;
    317 	} */ *uap = v;
    318 	struct statfs btmp, *bsp;
    319 	struct linux_statfs ltmp;
    320 	struct sys_fstatfs_args bsa;
    321 	caddr_t sg;
    322 	int error;
    323 
    324 	sg = stackgap_init(p->p_emul);
    325 	bsp = (struct statfs *) stackgap_alloc(&sg, sizeof (struct statfs));
    326 
    327 	SCARG(&bsa, fd) = SCARG(uap, fd);
    328 	SCARG(&bsa, buf) = bsp;
    329 
    330 	if ((error = sys_fstatfs(p, &bsa, retval)))
    331 		return error;
    332 
    333 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
    334 		return error;
    335 
    336 	bsd_to_linux_statfs(&btmp, &ltmp);
    337 
    338 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
    339 }
    340 
    341 char linux_sysname[] = "Linux";
    342 char linux_release[] = "2.0.38";
    343 char linux_version[] = "#0 Sun Apr 1 11:11:11 MET 2000";
    344 
    345 /*
    346  * uname(). Just copy the info from the various strings stored in the
    347  * kernel, and put it in the Linux utsname structure. That structure
    348  * is almost the same as the NetBSD one, only it has fields 65 characters
    349  * long, and an extra domainname field.
    350  */
    351 int
    352 linux_sys_uname(p, v, retval)
    353 	struct proc *p;
    354 	void *v;
    355 	register_t *retval;
    356 {
    357 	struct linux_sys_uname_args /* {
    358 		syscallarg(struct linux_utsname *) up;
    359 	} */ *uap = v;
    360 	struct linux_utsname luts;
    361 
    362 	strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
    363 	strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
    364 	strncpy(luts.l_release, linux_release, sizeof(luts.l_release));
    365 	strncpy(luts.l_version, linux_version, sizeof(luts.l_version));
    366 	strncpy(luts.l_machine, machine, sizeof(luts.l_machine));
    367 	strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
    368 
    369 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
    370 }
    371 
    372 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
    373 /* Used indirectly on: arm, i386, m68k */
    374 
    375 /*
    376  * New type Linux mmap call.
    377  * Only called directly on machines with >= 6 free regs.
    378  */
    379 int
    380 linux_sys_mmap(p, v, retval)
    381 	struct proc *p;
    382 	void *v;
    383 	register_t *retval;
    384 {
    385 	struct linux_sys_mmap_args /* {
    386 		syscallarg(unsigned long) addr;
    387 		syscallarg(size_t) len;
    388 		syscallarg(int) prot;
    389 		syscallarg(int) flags;
    390 		syscallarg(int) fd;
    391 		syscallarg(off_t) offset;
    392 	} */ *uap = v;
    393 	struct sys_mmap_args cma;
    394 	int flags;
    395 
    396 	flags = 0;
    397 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_SHARED, MAP_SHARED);
    398 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_PRIVATE, MAP_PRIVATE);
    399 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_FIXED, MAP_FIXED);
    400 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_ANON, MAP_ANON);
    401 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
    402 
    403 	SCARG(&cma,addr) = (void *)SCARG(uap, addr);
    404 	SCARG(&cma,len) = SCARG(uap, len);
    405 	SCARG(&cma,prot) = SCARG(uap, prot);
    406 	if (SCARG(&cma,prot) & VM_PROT_WRITE) /* XXX */
    407 		SCARG(&cma,prot) |= VM_PROT_READ;
    408 	SCARG(&cma,flags) = flags;
    409 	SCARG(&cma,fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
    410 	SCARG(&cma,pad) = 0;
    411 	SCARG(&cma,pos) = SCARG(uap, offset);
    412 
    413 	return sys_mmap(p, &cma, retval);
    414 }
    415 
    416 int
    417 linux_sys_mremap(p, v, retval)
    418 	struct proc *p;
    419 	void *v;
    420 	register_t *retval;
    421 {
    422 	struct linux_sys_mremap_args /* {
    423 		syscallarg(void *) old_address;
    424 		syscallarg(size_t) old_size;
    425 		syscallarg(size_t) new_size;
    426 		syscallarg(u_long) flags;
    427 	} */ *uap = v;
    428 	struct sys_munmap_args mua;
    429 	size_t old_size, new_size;
    430 	int error;
    431 
    432 	old_size = round_page(SCARG(uap, old_size));
    433 	new_size = round_page(SCARG(uap, new_size));
    434 
    435 	/*
    436 	 * Growing mapped region.
    437 	 */
    438 	if (new_size > old_size) {
    439 		/*
    440 		 * XXX Implement me.  What we probably want to do is
    441 		 * XXX dig out the guts of the old mapping, mmap that
    442 		 * XXX object again with the new size, then munmap
    443 		 * XXX the old mapping.
    444 		 */
    445 		*retval = 0;
    446 		return (ENOMEM);
    447 	}
    448 
    449 	/*
    450 	 * Shrinking mapped region.
    451 	 */
    452 	if (new_size < old_size) {
    453 		SCARG(&mua, addr) = (caddr_t)SCARG(uap, old_address) +
    454 		    new_size;
    455 		SCARG(&mua, len) = old_size - new_size;
    456 		error = sys_munmap(p, &mua, retval);
    457 		*retval = error ? 0 : (register_t)SCARG(uap, old_address);
    458 		return (error);
    459 	}
    460 
    461 	/*
    462 	 * No change.
    463 	 */
    464 	*retval = (register_t)SCARG(uap, old_address);
    465 	return (0);
    466 }
    467 
    468 int
    469 linux_sys_msync(p, v, retval)
    470 	struct proc *p;
    471 	void *v;
    472 	register_t *retval;
    473 {
    474 	struct linux_sys_msync_args /* {
    475 		syscallarg(caddr_t) addr;
    476 		syscallarg(int) len;
    477 		syscallarg(int) fl;
    478 	} */ *uap = v;
    479 
    480 	struct sys___msync13_args bma;
    481 
    482 	/* flags are ignored */
    483 	SCARG(&bma, addr) = SCARG(uap, addr);
    484 	SCARG(&bma, len) = SCARG(uap, len);
    485 	SCARG(&bma, flags) = SCARG(uap, fl);
    486 
    487 	return sys___msync13(p, &bma, retval);
    488 }
    489 
    490 /*
    491  * This code is partly stolen from src/lib/libc/compat-43/times.c
    492  * XXX - CLK_TCK isn't declared in /sys, just in <time.h>, done here
    493  */
    494 
    495 #define CLK_TCK 100
    496 #define	CONVTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
    497 
    498 int
    499 linux_sys_times(p, v, retval)
    500 	struct proc *p;
    501 	void *v;
    502 	register_t *retval;
    503 {
    504 	struct linux_sys_times_args /* {
    505 		syscallarg(struct times *) tms;
    506 	} */ *uap = v;
    507 	struct timeval t;
    508 	struct linux_tms ltms;
    509 	struct rusage ru;
    510 	int error, s;
    511 
    512 	calcru(p, &ru.ru_utime, &ru.ru_stime, NULL);
    513 	ltms.ltms_utime = CONVTCK(ru.ru_utime);
    514 	ltms.ltms_stime = CONVTCK(ru.ru_stime);
    515 
    516 	ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
    517 	ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
    518 
    519 	if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
    520 		return error;
    521 
    522 	s = splclock();
    523 	timersub(&time, &boottime, &t);
    524 	splx(s);
    525 
    526 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
    527 	return 0;
    528 }
    529 
    530 /*
    531  * Linux 'readdir' call. This code is mostly taken from the
    532  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
    533  * an attempt has been made to keep it a little cleaner (failing
    534  * miserably, because of the cruft needed if count 1 is passed).
    535  *
    536  * The d_off field should contain the offset of the next valid entry,
    537  * but in Linux it has the offset of the entry itself. We emulate
    538  * that bug here.
    539  *
    540  * Read in BSD-style entries, convert them, and copy them out.
    541  *
    542  * Note that this doesn't handle union-mounted filesystems.
    543  */
    544 int
    545 linux_sys_getdents(p, v, retval)
    546 	struct proc *p;
    547 	void *v;
    548 	register_t *retval;
    549 {
    550 	struct linux_sys_getdents_args /* {
    551 		syscallarg(int) fd;
    552 		syscallarg(struct linux_dirent *) dent;
    553 		syscallarg(unsigned int) count;
    554 	} */ *uap = v;
    555 	struct dirent *bdp;
    556 	struct vnode *vp;
    557 	caddr_t	inp, buf;		/* BSD-format */
    558 	int len, reclen;		/* BSD-format */
    559 	caddr_t outp;			/* Linux-format */
    560 	int resid, linux_reclen = 0;	/* Linux-format */
    561 	struct file *fp;
    562 	struct uio auio;
    563 	struct iovec aiov;
    564 	struct linux_dirent idb;
    565 	off_t off;		/* true file offset */
    566 	int buflen, error, eofflag, nbytes, oldcall;
    567 	struct vattr va;
    568 	off_t *cookiebuf = NULL, *cookie;
    569 	int ncookies;
    570 
    571 	/* getvnode() will use the descriptor for us */
    572 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
    573 		return (error);
    574 
    575 	if ((fp->f_flag & FREAD) == 0) {
    576 		error = EBADF;
    577 		goto out1;
    578 	}
    579 
    580 	vp = (struct vnode *)fp->f_data;
    581 	if (vp->v_type != VDIR) {
    582 		error = EINVAL;
    583 		goto out1;
    584 	}
    585 
    586 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)))
    587 		goto out1;
    588 
    589 	nbytes = SCARG(uap, count);
    590 	if (nbytes == 1) {	/* emulating old, broken behaviour */
    591 		nbytes = sizeof (struct linux_dirent);
    592 		buflen = max(va.va_blocksize, nbytes);
    593 		oldcall = 1;
    594 	} else {
    595 		buflen = min(MAXBSIZE, nbytes);
    596 		if (buflen < va.va_blocksize)
    597 			buflen = va.va_blocksize;
    598 		oldcall = 0;
    599 	}
    600 	buf = malloc(buflen, M_TEMP, M_WAITOK);
    601 
    602 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    603 	off = fp->f_offset;
    604 again:
    605 	aiov.iov_base = buf;
    606 	aiov.iov_len = buflen;
    607 	auio.uio_iov = &aiov;
    608 	auio.uio_iovcnt = 1;
    609 	auio.uio_rw = UIO_READ;
    610 	auio.uio_segflg = UIO_SYSSPACE;
    611 	auio.uio_procp = p;
    612 	auio.uio_resid = buflen;
    613 	auio.uio_offset = off;
    614 	/*
    615          * First we read into the malloc'ed buffer, then
    616          * we massage it into user space, one record at a time.
    617          */
    618 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
    619 	    &ncookies);
    620 	if (error)
    621 		goto out;
    622 
    623 	inp = buf;
    624 	outp = (caddr_t)SCARG(uap, dent);
    625 	resid = nbytes;
    626 	if ((len = buflen - auio.uio_resid) == 0)
    627 		goto eof;
    628 
    629 	for (cookie = cookiebuf; len > 0; len -= reclen) {
    630 		bdp = (struct dirent *)inp;
    631 		reclen = bdp->d_reclen;
    632 		if (reclen & 3)
    633 			panic("linux_readdir");
    634 		if (bdp->d_fileno == 0) {
    635 			inp += reclen;	/* it is a hole; squish it out */
    636 			off = *cookie++;
    637 			continue;
    638 		}
    639 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
    640 		if (reclen > len || resid < linux_reclen) {
    641 			/* entry too big for buffer, so just stop */
    642 			outp++;
    643 			break;
    644 		}
    645 		/*
    646 		 * Massage in place to make a Linux-shaped dirent (otherwise
    647 		 * we have to worry about touching user memory outside of
    648 		 * the copyout() call).
    649 		 */
    650 		idb.d_ino = (linux_ino_t)bdp->d_fileno;
    651 		/*
    652 		 * The old readdir() call misuses the offset and reclen fields.
    653 		 */
    654 		if (oldcall) {
    655 			idb.d_off = (linux_off_t)linux_reclen;
    656 			idb.d_reclen = (u_short)bdp->d_namlen;
    657 		} else {
    658 			if (sizeof (linux_off_t) < 4 && (off >> 32) != 0) {
    659 				compat_offseterr(vp, "linux_getdents");
    660 				error = EINVAL;
    661 				goto out;
    662 			}
    663 			idb.d_off = (linux_off_t)off;
    664 			idb.d_reclen = (u_short)linux_reclen;
    665 		}
    666 		strcpy(idb.d_name, bdp->d_name);
    667 		if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
    668 			goto out;
    669 		/* advance past this real entry */
    670 		inp += reclen;
    671 		off = *cookie++;	/* each entry points to itself */
    672 		/* advance output past Linux-shaped entry */
    673 		outp += linux_reclen;
    674 		resid -= linux_reclen;
    675 		if (oldcall)
    676 			break;
    677 	}
    678 
    679 	/* if we squished out the whole block, try again */
    680 	if (outp == (caddr_t)SCARG(uap, dent))
    681 		goto again;
    682 	fp->f_offset = off;	/* update the vnode offset */
    683 
    684 	if (oldcall)
    685 		nbytes = resid + linux_reclen;
    686 
    687 eof:
    688 	*retval = nbytes - resid;
    689 out:
    690 	VOP_UNLOCK(vp, 0);
    691 	if (cookiebuf)
    692 		free(cookiebuf, M_TEMP);
    693 	free(buf, M_TEMP);
    694  out1:
    695 	FILE_UNUSE(fp, p);
    696 	return error;
    697 }
    698 
    699 /*
    700  * Even when just using registers to pass arguments to syscalls you can
    701  * have 5 of them on the i386. So this newer version of select() does
    702  * this.
    703  */
    704 int
    705 linux_sys_select(p, v, retval)
    706 	struct proc *p;
    707 	void *v;
    708 	register_t *retval;
    709 {
    710 	struct linux_sys_select_args /* {
    711 		syscallarg(int) nfds;
    712 		syscallarg(fd_set *) readfds;
    713 		syscallarg(fd_set *) writefds;
    714 		syscallarg(fd_set *) exceptfds;
    715 		syscallarg(struct timeval *) timeout;
    716 	} */ *uap = v;
    717 
    718 	return linux_select1(p, retval, SCARG(uap, nfds), SCARG(uap, readfds),
    719 	    SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout));
    720 }
    721 
    722 /*
    723  * Common code for the old and new versions of select(). A couple of
    724  * things are important:
    725  * 1) return the amount of time left in the 'timeout' parameter
    726  * 2) select never returns ERESTART on Linux, always return EINTR
    727  */
    728 int
    729 linux_select1(p, retval, nfds, readfds, writefds, exceptfds, timeout)
    730 	struct proc *p;
    731 	register_t *retval;
    732 	int nfds;
    733 	fd_set *readfds, *writefds, *exceptfds;
    734 	struct timeval *timeout;
    735 {
    736 	struct sys_select_args bsa;
    737 	struct timeval tv0, tv1, utv, *tvp;
    738 	caddr_t sg;
    739 	int error;
    740 
    741 	SCARG(&bsa, nd) = nfds;
    742 	SCARG(&bsa, in) = readfds;
    743 	SCARG(&bsa, ou) = writefds;
    744 	SCARG(&bsa, ex) = exceptfds;
    745 	SCARG(&bsa, tv) = timeout;
    746 
    747 	/*
    748 	 * Store current time for computation of the amount of
    749 	 * time left.
    750 	 */
    751 	if (timeout) {
    752 		if ((error = copyin(timeout, &utv, sizeof(utv))))
    753 			return error;
    754 		if (itimerfix(&utv)) {
    755 			/*
    756 			 * The timeval was invalid.  Convert it to something
    757 			 * valid that will act as it does under Linux.
    758 			 */
    759 			sg = stackgap_init(p->p_emul);
    760 			tvp = stackgap_alloc(&sg, sizeof(utv));
    761 			utv.tv_sec += utv.tv_usec / 1000000;
    762 			utv.tv_usec %= 1000000;
    763 			if (utv.tv_usec < 0) {
    764 				utv.tv_sec -= 1;
    765 				utv.tv_usec += 1000000;
    766 			}
    767 			if (utv.tv_sec < 0)
    768 				timerclear(&utv);
    769 			if ((error = copyout(&utv, tvp, sizeof(utv))))
    770 				return error;
    771 			SCARG(&bsa, tv) = tvp;
    772 		}
    773 		microtime(&tv0);
    774 	}
    775 
    776 	error = sys_select(p, &bsa, retval);
    777 	if (error) {
    778 		/*
    779 		 * See fs/select.c in the Linux kernel.  Without this,
    780 		 * Maelstrom doesn't work.
    781 		 */
    782 		if (error == ERESTART)
    783 			error = EINTR;
    784 		return error;
    785 	}
    786 
    787 	if (timeout) {
    788 		if (*retval) {
    789 			/*
    790 			 * Compute how much time was left of the timeout,
    791 			 * by subtracting the current time and the time
    792 			 * before we started the call, and subtracting
    793 			 * that result from the user-supplied value.
    794 			 */
    795 			microtime(&tv1);
    796 			timersub(&tv1, &tv0, &tv1);
    797 			timersub(&utv, &tv1, &utv);
    798 			if (utv.tv_sec < 0)
    799 				timerclear(&utv);
    800 		} else
    801 			timerclear(&utv);
    802 		if ((error = copyout(&utv, timeout, sizeof(utv))))
    803 			return error;
    804 	}
    805 
    806 	return 0;
    807 }
    808 
    809 /*
    810  * Get the process group of a certain process. Look it up
    811  * and return the value.
    812  */
    813 int
    814 linux_sys_getpgid(p, v, retval)
    815 	struct proc *p;
    816 	void *v;
    817 	register_t *retval;
    818 {
    819 	struct linux_sys_getpgid_args /* {
    820 		syscallarg(int) pid;
    821 	} */ *uap = v;
    822 	struct proc *targp;
    823 
    824 	if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) {
    825 		if ((targp = pfind(SCARG(uap, pid))) == 0)
    826 			return ESRCH;
    827 	}
    828 	else
    829 		targp = p;
    830 
    831 	retval[0] = targp->p_pgid;
    832 	return 0;
    833 }
    834 
    835 /*
    836  * Set the 'personality' (emulation mode) for the current process. Only
    837  * accept the Linux personality here (0). This call is needed because
    838  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
    839  * ELF binaries run in Linux mode, not SVR4 mode.
    840  */
    841 int
    842 linux_sys_personality(p, v, retval)
    843 	struct proc *p;
    844 	void *v;
    845 	register_t *retval;
    846 {
    847 	struct linux_sys_personality_args /* {
    848 		syscallarg(int) per;
    849 	} */ *uap = v;
    850 
    851 	if (SCARG(uap, per) != 0)
    852 		return EINVAL;
    853 	retval[0] = 0;
    854 	return 0;
    855 }
    856 
    857 #if defined(__i386__) || defined(__m68k__)
    858 /*
    859  * The calls are here because of type conversions.
    860  */
    861 int
    862 linux_sys_setreuid16(p, v, retval)
    863 	struct proc *p;
    864 	void *v;
    865 	register_t *retval;
    866 {
    867 	struct linux_sys_setreuid16_args /* {
    868 		syscallarg(int) ruid;
    869 		syscallarg(int) euid;
    870 	} */ *uap = v;
    871 	struct sys_setreuid_args bsa;
    872 
    873 	SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
    874 		(uid_t)-1 : SCARG(uap, ruid);
    875 	SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
    876 		(uid_t)-1 : SCARG(uap, euid);
    877 
    878 	return sys_setreuid(p, &bsa, retval);
    879 }
    880 
    881 int
    882 linux_sys_setregid16(p, v, retval)
    883 	struct proc *p;
    884 	void *v;
    885 	register_t *retval;
    886 {
    887 	struct linux_sys_setregid16_args /* {
    888 		syscallarg(int) rgid;
    889 		syscallarg(int) egid;
    890 	} */ *uap = v;
    891 	struct sys_setregid_args bsa;
    892 
    893 	SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
    894 		(uid_t)-1 : SCARG(uap, rgid);
    895 	SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
    896 		(uid_t)-1 : SCARG(uap, egid);
    897 
    898 	return sys_setregid(p, &bsa, retval);
    899 }
    900 
    901 int
    902 linux_sys_setresuid16(p, v, retval)
    903 	struct proc *p;
    904 	void *v;
    905 	register_t *retval;
    906 {
    907 	struct linux_sys_setresuid16_args /* {
    908 		syscallarg(uid_t) ruid;
    909 		syscallarg(uid_t) euid;
    910 		syscallarg(uid_t) suid;
    911 	} */ *uap = v;
    912 	struct linux_sys_setresuid16_args lsa;
    913 
    914 	SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
    915 		(uid_t)-1 : SCARG(uap, ruid);
    916 	SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
    917 		(uid_t)-1 : SCARG(uap, euid);
    918 	SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ?
    919 		(uid_t)-1 : SCARG(uap, suid);
    920 
    921 	return linux_sys_setresuid(p, &lsa, retval);
    922 }
    923 
    924 int
    925 linux_sys_setresgid16(p, v, retval)
    926 	struct proc *p;
    927 	void *v;
    928 	register_t *retval;
    929 {
    930 	struct linux_sys_setresgid16_args /* {
    931 		syscallarg(gid_t) rgid;
    932 		syscallarg(gid_t) egid;
    933 		syscallarg(gid_t) sgid;
    934 	} */ *uap = v;
    935 	struct linux_sys_setresgid16_args lsa;
    936 
    937 	SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
    938 		(gid_t)-1 : SCARG(uap, rgid);
    939 	SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
    940 		(gid_t)-1 : SCARG(uap, egid);
    941 	SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ?
    942 		(gid_t)-1 : SCARG(uap, sgid);
    943 
    944 	return linux_sys_setresgid(p, &lsa, retval);
    945 }
    946 
    947 int
    948 linux_sys_getgroups16(p, v, retval)
    949 	struct proc *p;
    950 	void *v;
    951 	register_t *retval;
    952 {
    953 	struct linux_sys_getgroups16_args /* {
    954 		syscallarg(int) gidsetsize;
    955 		syscallarg(linux_gid_t *) gidset;
    956 	} */ *uap = v;
    957 	caddr_t sg;
    958 	int n, error, i;
    959 	struct sys_getgroups_args bsa;
    960 	gid_t *bset, *kbset;
    961 	linux_gid_t *lset;
    962 	struct pcred *pc = p->p_cred;
    963 
    964 	n = SCARG(uap, gidsetsize);
    965 	if (n < 0)
    966 		return EINVAL;
    967 	error = 0;
    968 	bset = kbset = NULL;
    969 	lset = NULL;
    970 	if (n > 0) {
    971 		n = min(pc->pc_ucred->cr_ngroups, n);
    972 		sg = stackgap_init(p->p_emul);
    973 		bset = stackgap_alloc(&sg, n * sizeof (gid_t));
    974 		kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
    975 		lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
    976 		if (bset == NULL || kbset == NULL || lset == NULL)
    977 			return ENOMEM;
    978 		SCARG(&bsa, gidsetsize) = n;
    979 		SCARG(&bsa, gidset) = bset;
    980 		error = sys_getgroups(p, &bsa, retval);
    981 		if (error != 0)
    982 			goto out;
    983 		error = copyin(bset, kbset, n * sizeof (gid_t));
    984 		if (error != 0)
    985 			goto out;
    986 		for (i = 0; i < n; i++)
    987 			lset[i] = (linux_gid_t)kbset[i];
    988 		error = copyout(lset, SCARG(uap, gidset),
    989 		    n * sizeof (linux_gid_t));
    990 	} else
    991 		*retval = pc->pc_ucred->cr_ngroups;
    992 out:
    993 	if (kbset != NULL)
    994 		free(kbset, M_TEMP);
    995 	if (lset != NULL)
    996 		free(lset, M_TEMP);
    997 	return error;
    998 }
    999 
   1000 int
   1001 linux_sys_setgroups16(p, v, retval)
   1002 	struct proc *p;
   1003 	void *v;
   1004 	register_t *retval;
   1005 {
   1006 	struct linux_sys_setgroups16_args /* {
   1007 		syscallarg(int) gidsetsize;
   1008 		syscallarg(linux_gid_t *) gidset;
   1009 	} */ *uap = v;
   1010 	caddr_t sg;
   1011 	int n;
   1012 	int error, i;
   1013 	struct sys_setgroups_args bsa;
   1014 	gid_t *bset, *kbset;
   1015 	linux_gid_t *lset;
   1016 
   1017 	n = SCARG(uap, gidsetsize);
   1018 	if (n < 0 || n > NGROUPS)
   1019 		return EINVAL;
   1020 	sg = stackgap_init(p->p_emul);
   1021 	bset = stackgap_alloc(&sg, n * sizeof (gid_t));
   1022 	lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
   1023 	kbset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
   1024 	if (lset == NULL || bset == NULL)
   1025 		return ENOMEM;
   1026 	error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t));
   1027 	if (error != 0)
   1028 		goto out;
   1029 	for (i = 0; i < n; i++)
   1030 		kbset[i] = (gid_t)lset[i];
   1031 	error = copyout(kbset, bset, n * sizeof (gid_t));
   1032 	if (error != 0)
   1033 		goto out;
   1034 	SCARG(&bsa, gidsetsize) = n;
   1035 	SCARG(&bsa, gidset) = bset;
   1036 	error = sys_setgroups(p, &bsa, retval);
   1037 
   1038 out:
   1039 	if (lset != NULL)
   1040 		free(lset, M_TEMP);
   1041 	if (kbset != NULL)
   1042 		free(kbset, M_TEMP);
   1043 
   1044 	return error;
   1045 }
   1046 
   1047 #endif /* __i386__ || __m68k__ */
   1048 
   1049 /*
   1050  * We have nonexistent fsuid equal to uid.
   1051  * If modification is requested, refuse.
   1052  */
   1053 int
   1054 linux_sys_setfsuid(p, v, retval)
   1055 	 struct proc *p;
   1056 	 void *v;
   1057 	 register_t *retval;
   1058 {
   1059 	 struct linux_sys_setfsuid_args /* {
   1060 		 syscallarg(uid_t) uid;
   1061 	 } */ *uap = v;
   1062 	 uid_t uid;
   1063 
   1064 	 uid = SCARG(uap, uid);
   1065 	 if (p->p_cred->p_ruid != uid)
   1066 		 return sys_nosys(p, v, retval);
   1067 	 else
   1068 		 return (0);
   1069 }
   1070 
   1071 /* XXX XXX XXX */
   1072 #ifndef alpha
   1073 int
   1074 linux_sys_getfsuid(p, v, retval)
   1075 	struct proc *p;
   1076 	void *v;
   1077 	register_t *retval;
   1078 {
   1079 	return sys_getuid(p, v, retval);
   1080 }
   1081 #endif
   1082 
   1083 int
   1084 linux_sys___sysctl(p, v, retval)
   1085 	struct proc *p;
   1086 	void *v;
   1087 	register_t *retval;
   1088 {
   1089 	struct linux_sys___sysctl_args /* {
   1090 		syscallarg(struct linux___sysctl *) lsp;
   1091 	} */ *uap = v;
   1092 	struct linux___sysctl ls;
   1093 	struct sys___sysctl_args bsa;
   1094 	int error;
   1095 
   1096 	if ((error = copyin(SCARG(uap, lsp), &ls, sizeof ls)))
   1097 		return error;
   1098 	SCARG(&bsa, name) = ls.name;
   1099 	SCARG(&bsa, namelen) = ls.namelen;
   1100 	SCARG(&bsa, old) = ls.old;
   1101 	SCARG(&bsa, oldlenp) = ls.oldlenp;
   1102 	SCARG(&bsa, new) = ls.new;
   1103 	SCARG(&bsa, newlen) = ls.newlen;
   1104 
   1105 	return sys___sysctl(p, &bsa, retval);
   1106 }
   1107 
   1108 int
   1109 linux_sys_setresuid(p, v, retval)
   1110 	struct proc *p;
   1111 	void *v;
   1112 	register_t *retval;
   1113 {
   1114 	struct linux_sys_setresuid_args /* {
   1115 		syscallarg(uid_t) ruid;
   1116 		syscallarg(uid_t) euid;
   1117 		syscallarg(uid_t) suid;
   1118 	} */ *uap = v;
   1119 	struct pcred *pc = p->p_cred;
   1120 	uid_t ruid, euid, suid;
   1121 	int error;
   1122 
   1123 	ruid = SCARG(uap, ruid);
   1124 	euid = SCARG(uap, euid);
   1125 	suid = SCARG(uap, suid);
   1126 
   1127 	/*
   1128 	 * Note: These checks are a little different than the NetBSD
   1129 	 * setreuid(2) call performs.  This precisely follows the
   1130 	 * behavior of the Linux kernel.
   1131 	 */
   1132 	if (ruid != (uid_t)-1 &&
   1133 	    ruid != pc->p_ruid &&
   1134 	    ruid != pc->pc_ucred->cr_uid &&
   1135 	    ruid != pc->p_svuid &&
   1136 	    (error = suser(pc->pc_ucred, &p->p_acflag)))
   1137 		return (error);
   1138 
   1139 	if (euid != (uid_t)-1 &&
   1140 	    euid != pc->p_ruid &&
   1141 	    euid != pc->pc_ucred->cr_uid &&
   1142 	    euid != pc->p_svuid &&
   1143 	    (error = suser(pc->pc_ucred, &p->p_acflag)))
   1144 		return (error);
   1145 
   1146 	if (suid != (uid_t)-1 &&
   1147 	    suid != pc->p_ruid &&
   1148 	    suid != pc->pc_ucred->cr_uid &&
   1149 	    suid != pc->p_svuid &&
   1150 	    (error = suser(pc->pc_ucred, &p->p_acflag)))
   1151 		return (error);
   1152 
   1153 	/*
   1154 	 * Now assign the new real, effective, and saved UIDs.
   1155 	 * Note that Linux, unlike NetBSD in setreuid(2), does not
   1156 	 * set the saved UID in this call unless the user specifies
   1157 	 * it.
   1158 	 */
   1159 	if (ruid != (uid_t)-1) {
   1160 		(void)chgproccnt(pc->p_ruid, -1);
   1161 		(void)chgproccnt(ruid, 1);
   1162 		pc->p_ruid = ruid;
   1163 	}
   1164 
   1165 	if (euid != (uid_t)-1) {
   1166 		pc->pc_ucred = crcopy(pc->pc_ucred);
   1167 		pc->pc_ucred->cr_uid = euid;
   1168 	}
   1169 
   1170 	if (suid != (uid_t)-1)
   1171 		pc->p_svuid = suid;
   1172 
   1173 	if (ruid != (uid_t)-1 && euid != (uid_t)-1 && suid != (uid_t)-1)
   1174 		p->p_flag |= P_SUGID;
   1175 	return (0);
   1176 }
   1177 
   1178 int
   1179 linux_sys_getresuid(p, v, retval)
   1180 	struct proc *p;
   1181 	void *v;
   1182 	register_t *retval;
   1183 {
   1184 	struct linux_sys_getresuid_args /* {
   1185 		syscallarg(uid_t *) ruid;
   1186 		syscallarg(uid_t *) euid;
   1187 		syscallarg(uid_t *) suid;
   1188 	} */ *uap = v;
   1189 	struct pcred *pc = p->p_cred;
   1190 	int error;
   1191 
   1192 	/*
   1193 	 * Linux copies these values out to userspace like so:
   1194 	 *
   1195 	 *	1. Copy out ruid.
   1196 	 *	2. If that succeeds, copy out euid.
   1197 	 *	3. If both of those succeed, copy out suid.
   1198 	 */
   1199 	if ((error = copyout(&pc->p_ruid, SCARG(uap, ruid),
   1200 			     sizeof(uid_t))) != 0)
   1201 		return (error);
   1202 
   1203 	if ((error = copyout(&pc->pc_ucred->cr_uid, SCARG(uap, euid),
   1204 			     sizeof(uid_t))) != 0)
   1205 		return (error);
   1206 
   1207 	return (copyout(&pc->p_svuid, SCARG(uap, suid), sizeof(uid_t)));
   1208 }
   1209 
   1210 int
   1211 linux_sys_ptrace(p, v, retval)
   1212 	struct proc *p;
   1213 	void *v;
   1214 	register_t *retval;
   1215 {
   1216 	struct linux_sys_ptrace_args /* {
   1217 		i386, m68k: T=int
   1218 		alpha: T=long
   1219 		syscallarg(T) request;
   1220 		syscallarg(T) pid;
   1221 		syscallarg(T) addr;
   1222 		syscallarg(T) data;
   1223 	} */ *uap = v;
   1224 	const int *ptr;
   1225 	int request;
   1226 
   1227 	ptr = linux_ptrace_request_map;
   1228 	request = SCARG(uap, request);
   1229 	while (*ptr != -1)
   1230 		if (*ptr++ == request) {
   1231 			struct sys_ptrace_args pta;
   1232 			caddr_t sg;
   1233 
   1234 			sg = stackgap_init(p->p_emul);
   1235 
   1236 			SCARG(&pta, req) = *ptr;
   1237 			SCARG(&pta, pid) = SCARG(uap, pid);
   1238 			SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr);
   1239 			SCARG(&pta, data) = SCARG(uap, data);
   1240 
   1241 			/*
   1242 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
   1243 			 * to continue as the process left off previously,
   1244 			 * i.e. same as if NetBSD ptrace called with
   1245 			 * addr == (caddr_t) 1.
   1246 			 */
   1247 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
   1248 				SCARG(&pta, addr) = (caddr_t) 1;
   1249 
   1250 			return sys_ptrace(p, &pta, retval);
   1251 		}
   1252 		else
   1253 			ptr++;
   1254 
   1255 	return LINUX_SYS_PTRACE_ARCH(p, uap, retval);
   1256 }
   1257 
   1258 int
   1259 linux_sys_reboot(struct proc *p, void *v, register_t *retval)
   1260 {
   1261 	struct linux_sys_reboot_args /* {
   1262 		syscallarg(int) magic1;
   1263 		syscallarg(int) magic2;
   1264 		syscallarg(int) cmd;
   1265 		syscallarg(void *) arg;
   1266 	} */ *uap = v;
   1267 	struct sys_reboot_args /* {
   1268 		syscallarg(int) opt;
   1269 		syscallarg(char *) bootstr;
   1270 	} */ sra;
   1271 	int error;
   1272 
   1273 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
   1274 		return(error);
   1275 
   1276 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
   1277 		return(EINVAL);
   1278 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
   1279 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
   1280 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
   1281 		return(EINVAL);
   1282 
   1283 	switch (SCARG(uap, cmd)) {
   1284 	case LINUX_REBOOT_CMD_RESTART:
   1285 		SCARG(&sra, opt) = RB_AUTOBOOT;
   1286 		break;
   1287 	case LINUX_REBOOT_CMD_HALT:
   1288 		SCARG(&sra, opt) = RB_HALT;
   1289 		break;
   1290 	case LINUX_REBOOT_CMD_POWER_OFF:
   1291 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
   1292 		break;
   1293 	case LINUX_REBOOT_CMD_RESTART2:
   1294 		/* Reboot with an argument. */
   1295 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
   1296 		SCARG(&sra, bootstr) = SCARG(uap, arg);
   1297 		break;
   1298 	case LINUX_REBOOT_CMD_CAD_ON:
   1299 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
   1300 	case LINUX_REBOOT_CMD_CAD_OFF:
   1301 		return(0);
   1302 	default:
   1303 		return(EINVAL);
   1304 	}
   1305 
   1306 	return(sys_reboot(p, &sra, retval));
   1307 }
   1308 
   1309 /*
   1310  * Copy of compat_12_sys_swapon().
   1311  */
   1312 int
   1313 linux_sys_swapon(p, v, retval)
   1314 	struct proc *p;
   1315 	void *v;
   1316 	register_t *retval;
   1317 {
   1318 	struct sys_swapctl_args ua;
   1319 	struct linux_sys_swapon_args /* {
   1320 		syscallarg(const char *) name;
   1321 	} */ *uap = v;
   1322 
   1323 	SCARG(&ua, cmd) = SWAP_ON;
   1324 	SCARG(&ua, arg) = (void *)SCARG(uap, name);
   1325 	SCARG(&ua, misc) = 0;	/* priority */
   1326 	return (sys_swapctl(p, &ua, retval));
   1327 }
   1328 
   1329 /*
   1330  * Stop swapping to the file or block device specified by path.
   1331  */
   1332 int
   1333 linux_sys_swapoff(p, v, retval)
   1334 	struct proc *p;
   1335 	void *v;
   1336 	register_t *retval;
   1337 {
   1338 	struct sys_swapctl_args ua;
   1339 	struct linux_sys_swapoff_args /* {
   1340 		syscallarg(const char *) path;
   1341 	} */ *uap = v;
   1342 
   1343 	SCARG(&ua, cmd) = SWAP_OFF;
   1344 	SCARG(&ua, arg) = (void *)SCARG(uap, path);
   1345 	return (sys_swapctl(p, &ua, retval));
   1346 }
   1347 
   1348 /*
   1349  * Copy of compat_09_sys_setdomainname()
   1350  */
   1351 /* ARGSUSED */
   1352 int
   1353 linux_sys_setdomainname(p, v, retval)
   1354 	struct proc *p;
   1355 	void *v;
   1356 	register_t *retval;
   1357 {
   1358 	struct linux_sys_setdomainname_args /* {
   1359 		syscallarg(char *) domainname;
   1360 		syscallarg(int) len;
   1361 	} */ *uap = v;
   1362 	int name;
   1363 	int error;
   1364 
   1365 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
   1366 		return (error);
   1367 	name = KERN_DOMAINNAME;
   1368 	return (kern_sysctl(&name, 1, 0, 0, SCARG(uap, domainname),
   1369 			    SCARG(uap, len), p));
   1370 }
   1371 
   1372 /*
   1373  * sysinfo()
   1374  */
   1375 /* ARGSUSED */
   1376 int
   1377 linux_sys_sysinfo(p, v, retval)
   1378 	struct proc *p;
   1379 	void *v;
   1380 	register_t *retval;
   1381 {
   1382 	struct linux_sys_sysinfo_args /* {
   1383 		syscallarg(struct linux_sysinfo *) arg;
   1384 	} */ *uap = v;
   1385 	struct linux_sysinfo si;
   1386 	struct loadavg *la;
   1387 
   1388 	si.uptime = time.tv_sec - boottime.tv_sec;
   1389 	la = &averunnable;
   1390 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
   1391 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
   1392 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
   1393 	si.totalram = ctob(physmem);
   1394 	si.freeram = uvmexp.free * uvmexp.pagesize;
   1395 	si.sharedram = 0;	/* XXX */
   1396 	si.bufferram = uvmexp.vnodepages * uvmexp.pagesize;
   1397 	si.totalswap = uvmexp.swpages * uvmexp.pagesize;
   1398 	si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
   1399 	si.procs = nprocs;
   1400 
   1401 	/* The following are only present in newer Linux kernels. */
   1402 	si.totalbig = 0;
   1403 	si.freebig = 0;
   1404 	si.mem_unit = 1;
   1405 
   1406 	return (copyout(&si, SCARG(uap, arg), sizeof si));
   1407 }
   1408 
   1409 /*
   1410  * This gets called for unsupported syscalls. The difference to sys_nosys()
   1411  * is that process does not get SIGSYS, the call just returns with ENOSYS.
   1412  * This is the way Linux does it and glibc depends on this behaviour.
   1413  */
   1414 int
   1415 linux_sys_nosys(p, v, retval)
   1416 	struct proc *p;
   1417 	void *v;
   1418 	register_t *retval;
   1419 {
   1420 	return (ENOSYS);
   1421 }
   1422