Home | History | Annotate | Line # | Download | only in common
linux_misc.c revision 1.83.2.8
      1 /*	$NetBSD: linux_misc.c,v 1.83.2.8 2002/02/28 04:12:56 nathanw 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/cdefs.h>
     67 __KERNEL_RCSID(0, "$NetBSD: linux_misc.c,v 1.83.2.8 2002/02/28 04:12:56 nathanw Exp $");
     68 
     69 #include <sys/param.h>
     70 #include <sys/systm.h>
     71 #include <sys/namei.h>
     72 #include <sys/lwp.h>
     73 #include <sys/proc.h>
     74 #include <sys/dirent.h>
     75 #include <sys/file.h>
     76 #include <sys/stat.h>
     77 #include <sys/filedesc.h>
     78 #include <sys/ioctl.h>
     79 #include <sys/kernel.h>
     80 #include <sys/malloc.h>
     81 #include <sys/mbuf.h>
     82 #include <sys/mman.h>
     83 #include <sys/mount.h>
     84 #include <sys/reboot.h>
     85 #include <sys/resource.h>
     86 #include <sys/resourcevar.h>
     87 #include <sys/signal.h>
     88 #include <sys/signalvar.h>
     89 #include <sys/socket.h>
     90 #include <sys/time.h>
     91 #include <sys/times.h>
     92 #include <sys/vnode.h>
     93 #include <sys/uio.h>
     94 #include <sys/wait.h>
     95 #include <sys/utsname.h>
     96 #include <sys/unistd.h>
     97 #include <sys/swap.h>		/* for SWAP_ON */
     98 #include <sys/sysctl.h>		/* for KERN_DOMAINNAME */
     99 
    100 #include <sys/ptrace.h>
    101 #include <machine/ptrace.h>
    102 
    103 #include <sys/syscallargs.h>
    104 
    105 #include <compat/linux/common/linux_types.h>
    106 #include <compat/linux/common/linux_signal.h>
    107 
    108 #include <compat/linux/linux_syscallargs.h>
    109 
    110 #include <compat/linux/common/linux_fcntl.h>
    111 #include <compat/linux/common/linux_mmap.h>
    112 #include <compat/linux/common/linux_dirent.h>
    113 #include <compat/linux/common/linux_util.h>
    114 #include <compat/linux/common/linux_misc.h>
    115 #include <compat/linux/common/linux_ptrace.h>
    116 #include <compat/linux/common/linux_reboot.h>
    117 #include <compat/linux/common/linux_emuldata.h>
    118 
    119 const int linux_ptrace_request_map[] = {
    120 	LINUX_PTRACE_TRACEME,	PT_TRACE_ME,
    121 	LINUX_PTRACE_PEEKTEXT,	PT_READ_I,
    122 	LINUX_PTRACE_PEEKDATA,	PT_READ_D,
    123 	LINUX_PTRACE_POKETEXT,	PT_WRITE_I,
    124 	LINUX_PTRACE_POKEDATA,	PT_WRITE_D,
    125 	LINUX_PTRACE_CONT,	PT_CONTINUE,
    126 	LINUX_PTRACE_KILL,	PT_KILL,
    127 	LINUX_PTRACE_ATTACH,	PT_ATTACH,
    128 	LINUX_PTRACE_DETACH,	PT_DETACH,
    129 #ifdef PT_STEP
    130 	LINUX_PTRACE_SINGLESTEP,	PT_STEP,
    131 #endif
    132 	-1
    133 };
    134 
    135 const static struct mnttypes {
    136 	char *bsd;
    137 	int linux;
    138 } fstypes[] = {
    139 	{ MOUNT_FFS,		LINUX_DEFAULT_SUPER_MAGIC	},
    140 	{ MOUNT_NFS,		LINUX_NFS_SUPER_MAGIC 		},
    141 	{ MOUNT_MFS,		LINUX_DEFAULT_SUPER_MAGIC	},
    142 	{ MOUNT_MSDOS,		LINUX_MSDOS_SUPER_MAGIC		},
    143 	{ MOUNT_LFS,		LINUX_DEFAULT_SUPER_MAGIC	},
    144 	{ MOUNT_FDESC,		LINUX_DEFAULT_SUPER_MAGIC	},
    145 	{ MOUNT_PORTAL,		LINUX_DEFAULT_SUPER_MAGIC	},
    146 	{ MOUNT_NULL,		LINUX_DEFAULT_SUPER_MAGIC	},
    147 	{ MOUNT_OVERLAY,	LINUX_DEFAULT_SUPER_MAGIC	},
    148 	{ MOUNT_UMAP,		LINUX_DEFAULT_SUPER_MAGIC	},
    149 	{ MOUNT_KERNFS,		LINUX_DEFAULT_SUPER_MAGIC	},
    150 	{ MOUNT_PROCFS,		LINUX_PROC_SUPER_MAGIC		},
    151 	{ MOUNT_AFS,		LINUX_DEFAULT_SUPER_MAGIC	},
    152 	{ MOUNT_CD9660,		LINUX_ISOFS_SUPER_MAGIC		},
    153 	{ MOUNT_UNION,		LINUX_DEFAULT_SUPER_MAGIC	},
    154 	{ MOUNT_ADOSFS,		LINUX_ADFS_SUPER_MAGIC		},
    155 	{ MOUNT_EXT2FS,		LINUX_EXT2_SUPER_MAGIC		},
    156 	{ MOUNT_CFS,		LINUX_DEFAULT_SUPER_MAGIC	},
    157 	{ MOUNT_CODA,		LINUX_CODA_SUPER_MAGIC		},
    158 	{ MOUNT_FILECORE,	LINUX_DEFAULT_SUPER_MAGIC	},
    159 	{ MOUNT_NTFS,		LINUX_DEFAULT_SUPER_MAGIC	},
    160 	{ MOUNT_SMBFS,		LINUX_SMB_SUPER_MAGIC		}
    161 };
    162 #define FSTYPESSIZE (sizeof(fstypes) / sizeof(fstypes[0]))
    163 
    164 /* Local linux_misc.c functions: */
    165 static void bsd_to_linux_statfs __P((struct statfs *, struct linux_statfs *));
    166 static int linux_to_bsd_limit __P((int));
    167 
    168 /*
    169  * The information on a terminated (or stopped) process needs
    170  * to be converted in order for Linux binaries to get a valid signal
    171  * number out of it.
    172  */
    173 void
    174 bsd_to_linux_wstat(st)
    175 	int *st;
    176 {
    177 
    178 	int sig;
    179 
    180 	if (WIFSIGNALED(*st)) {
    181 		sig = WTERMSIG(*st);
    182 		if (sig >= 0 && sig < NSIG)
    183 			*st= (*st& ~0177) | native_to_linux_sig[sig];
    184 	} else if (WIFSTOPPED(*st)) {
    185 		sig = WSTOPSIG(*st);
    186 		if (sig >= 0 && sig < NSIG)
    187 			*st = (*st & ~0xff00) | (native_to_linux_sig[sig] << 8);
    188 	}
    189 }
    190 
    191 /*
    192  * This is very much the same as waitpid()
    193  */
    194 int
    195 linux_sys_wait4(l, v, retval)
    196 	struct lwp *l;
    197 	void *v;
    198 	register_t *retval;
    199 {
    200 	struct linux_sys_wait4_args /* {
    201 		syscallarg(int) pid;
    202 		syscallarg(int *) status;
    203 		syscallarg(int) options;
    204 		syscallarg(struct rusage *) rusage;
    205 	} */ *uap = v;
    206 	struct proc *p = l->l_proc;
    207 	struct sys_wait4_args w4a;
    208 	int error, *status, tstat, options, linux_options;
    209 	caddr_t sg;
    210 
    211 	if (SCARG(uap, status) != NULL) {
    212 		sg = stackgap_init(p->p_emul);
    213 		status = (int *) stackgap_alloc(&sg, sizeof *status);
    214 	} else
    215 		status = NULL;
    216 
    217 	linux_options = SCARG(uap, options);
    218 	options = 0;
    219 	if (linux_options &
    220 	    ~(LINUX_WAIT4_WNOHANG|LINUX_WAIT4_WUNTRACED|LINUX_WAIT4_WALL|
    221 	      LINUX_WAIT4_WCLONE))
    222 		return (EINVAL);
    223 
    224 	if (linux_options & LINUX_WAIT4_WNOHANG)
    225 		options |= WNOHANG;
    226 	if (linux_options & LINUX_WAIT4_WUNTRACED)
    227 		options |= WUNTRACED;
    228 	if (linux_options & LINUX_WAIT4_WALL)
    229 		options |= WALLSIG;
    230 	if (linux_options & LINUX_WAIT4_WCLONE)
    231 		options |= WALTSIG;
    232 
    233 	SCARG(&w4a, pid) = SCARG(uap, pid);
    234 	SCARG(&w4a, status) = status;
    235 	SCARG(&w4a, options) = options;
    236 	SCARG(&w4a, rusage) = SCARG(uap, rusage);
    237 
    238 	if ((error = sys_wait4(l, &w4a, retval)))
    239 		return error;
    240 
    241 	sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD);
    242 
    243 	if (status != NULL) {
    244 		if ((error = copyin(status, &tstat, sizeof tstat)))
    245 			return error;
    246 
    247 		bsd_to_linux_wstat(&tstat);
    248 		return copyout(&tstat, SCARG(uap, status), sizeof tstat);
    249 	}
    250 
    251 	return 0;
    252 }
    253 
    254 /*
    255  * Linux brk(2). The check if the new address is >= the old one is
    256  * done in the kernel in Linux. NetBSD does it in the library.
    257  */
    258 int
    259 linux_sys_brk(l, v, retval)
    260 	struct lwp *l;
    261 	void *v;
    262 	register_t *retval;
    263 {
    264 	struct linux_sys_brk_args /* {
    265 		syscallarg(char *) nsize;
    266 	} */ *uap = v;
    267 	struct proc *p = l->l_proc;
    268 	char *nbrk = SCARG(uap, nsize);
    269 	struct sys_obreak_args oba;
    270 	struct vmspace *vm = p->p_vmspace;
    271 	struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata;
    272 
    273 	SCARG(&oba, nsize) = nbrk;
    274 
    275 	if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(l, &oba, retval) == 0)
    276 		ed->p_break = (char *)nbrk;
    277 	else
    278 		nbrk = ed->p_break;
    279 
    280 	retval[0] = (register_t)nbrk;
    281 	return 0;
    282 }
    283 
    284 /*
    285  * Convert BSD statfs structure to Linux statfs structure.
    286  * The Linux structure has less fields, and it also wants
    287  * the length of a name in a dir entry in a field, which
    288  * we fake (probably the wrong way).
    289  */
    290 static void
    291 bsd_to_linux_statfs(bsp, lsp)
    292 	struct statfs *bsp;
    293 	struct linux_statfs *lsp;
    294 {
    295 	int i;
    296 
    297 	for (i = 0; i < FSTYPESSIZE; i++)
    298 		if (strcmp(bsp->f_fstypename, fstypes[i].bsd) == 0)
    299 			break;
    300 
    301 	if (i == FSTYPESSIZE) {
    302 #ifdef DIAGNOSTIC
    303 		printf("unhandled fstype in linux emulation: %s\n",
    304 		    bsp->f_fstypename);
    305 #endif
    306 		lsp->l_ftype = LINUX_DEFAULT_SUPER_MAGIC;
    307 	} else {
    308 		lsp->l_ftype = fstypes[i].linux;
    309 	}
    310 
    311 	lsp->l_fbsize = bsp->f_bsize;
    312 	lsp->l_fblocks = bsp->f_blocks;
    313 	lsp->l_fbfree = bsp->f_bfree;
    314 	lsp->l_fbavail = bsp->f_bavail;
    315 	lsp->l_ffiles = bsp->f_files;
    316 	lsp->l_fffree = bsp->f_ffree;
    317 	/* Linux sets the fsid to 0..., we don't */
    318 	lsp->l_ffsid.val[0] = bsp->f_fsid.val[0];
    319 	lsp->l_ffsid.val[1] = bsp->f_fsid.val[1];
    320 	lsp->l_fnamelen = MAXNAMLEN;	/* XXX */
    321 	(void)memset(lsp->l_fspare, 0, sizeof(lsp->l_fspare));
    322 }
    323 
    324 /*
    325  * Implement the fs stat functions. Straightforward.
    326  */
    327 int
    328 linux_sys_statfs(l, v, retval)
    329 	struct lwp *l;
    330 	void *v;
    331 	register_t *retval;
    332 {
    333 	struct linux_sys_statfs_args /* {
    334 		syscallarg(const char *) path;
    335 		syscallarg(struct linux_statfs *) sp;
    336 	} */ *uap = v;
    337 	struct proc *p = l->l_proc;
    338 	struct statfs btmp, *bsp;
    339 	struct linux_statfs ltmp;
    340 	struct sys_statfs_args bsa;
    341 	caddr_t sg;
    342 	int error;
    343 
    344 	sg = stackgap_init(p->p_emul);
    345 	bsp = (struct statfs *) stackgap_alloc(&sg, sizeof (struct statfs));
    346 
    347 	CHECK_ALT_EXIST(p, &sg, SCARG(uap, path));
    348 
    349 	SCARG(&bsa, path) = SCARG(uap, path);
    350 	SCARG(&bsa, buf) = bsp;
    351 
    352 	if ((error = sys_statfs(l, &bsa, retval)))
    353 		return error;
    354 
    355 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
    356 		return error;
    357 
    358 	bsd_to_linux_statfs(&btmp, &ltmp);
    359 
    360 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
    361 }
    362 
    363 int
    364 linux_sys_fstatfs(l, v, retval)
    365 	struct lwp *l;
    366 	void *v;
    367 	register_t *retval;
    368 {
    369 	struct linux_sys_fstatfs_args /* {
    370 		syscallarg(int) fd;
    371 		syscallarg(struct linux_statfs *) sp;
    372 	} */ *uap = v;
    373 	struct proc *p = l->l_proc;
    374 	struct statfs btmp, *bsp;
    375 	struct linux_statfs ltmp;
    376 	struct sys_fstatfs_args bsa;
    377 	caddr_t sg;
    378 	int error;
    379 
    380 	sg = stackgap_init(p->p_emul);
    381 	bsp = (struct statfs *) stackgap_alloc(&sg, sizeof (struct statfs));
    382 
    383 	SCARG(&bsa, fd) = SCARG(uap, fd);
    384 	SCARG(&bsa, buf) = bsp;
    385 
    386 	if ((error = sys_fstatfs(l, &bsa, retval)))
    387 		return error;
    388 
    389 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
    390 		return error;
    391 
    392 	bsd_to_linux_statfs(&btmp, &ltmp);
    393 
    394 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
    395 }
    396 
    397 /*
    398  * uname(). Just copy the info from the various strings stored in the
    399  * kernel, and put it in the Linux utsname structure. That structure
    400  * is almost the same as the NetBSD one, only it has fields 65 characters
    401  * long, and an extra domainname field.
    402  */
    403 int
    404 linux_sys_uname(l, v, retval)
    405 	struct lwp *l;
    406 	void *v;
    407 	register_t *retval;
    408 {
    409 	struct linux_sys_uname_args /* {
    410 		syscallarg(struct linux_utsname *) up;
    411 	} */ *uap = v;
    412 	struct linux_utsname luts;
    413 
    414 	strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
    415 	strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
    416 	strncpy(luts.l_release, linux_release, sizeof(luts.l_release));
    417 	strncpy(luts.l_version, linux_version, sizeof(luts.l_version));
    418 	strncpy(luts.l_machine, machine, sizeof(luts.l_machine));
    419 	strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
    420 
    421 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
    422 }
    423 
    424 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
    425 /* Used indirectly on: arm, i386, m68k */
    426 
    427 /*
    428  * New type Linux mmap call.
    429  * Only called directly on machines with >= 6 free regs.
    430  */
    431 int
    432 linux_sys_mmap(l, v, retval)
    433 	struct lwp *l;
    434 	void *v;
    435 	register_t *retval;
    436 {
    437 	struct linux_sys_mmap_args /* {
    438 		syscallarg(unsigned long) addr;
    439 		syscallarg(size_t) len;
    440 		syscallarg(int) prot;
    441 		syscallarg(int) flags;
    442 		syscallarg(int) fd;
    443 		syscallarg(linux_off_t) offset;
    444 	} */ *uap = v;
    445 	struct sys_mmap_args cma;
    446 	int flags;
    447 
    448 	flags = 0;
    449 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_SHARED, MAP_SHARED);
    450 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_PRIVATE, MAP_PRIVATE);
    451 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_FIXED, MAP_FIXED);
    452 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_ANON, MAP_ANON);
    453 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
    454 
    455 	SCARG(&cma,addr) = (void *)SCARG(uap, addr);
    456 	SCARG(&cma,len) = SCARG(uap, len);
    457 	SCARG(&cma,prot) = SCARG(uap, prot);
    458 	if (SCARG(&cma,prot) & VM_PROT_WRITE) /* XXX */
    459 		SCARG(&cma,prot) |= VM_PROT_READ;
    460 	SCARG(&cma,flags) = flags;
    461 	SCARG(&cma,fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
    462 	SCARG(&cma,pad) = 0;
    463 	SCARG(&cma,pos) = (off_t)SCARG(uap, offset);
    464 
    465 	return sys_mmap(l, &cma, retval);
    466 }
    467 
    468 /*
    469  * Newer type Linux mmap call.
    470  */
    471 int
    472 linux_sys_mmap2(l, v, retval)
    473 	struct lwp *l;
    474 	void *v;
    475 	register_t *retval;
    476 {
    477 	struct linux_sys_mmap2_args /* {
    478 		syscallarg(void *) addr;
    479 		syscallarg(size_t) len;
    480 		syscallarg(int) prot;
    481 		syscallarg(int) flags;
    482 		syscallarg(int) fd;
    483 		syscallarg(off_t) offset;
    484 	} */ *uap = v;
    485 	struct sys_mmap_args cma;
    486 	int flags;
    487 
    488 	flags = 0;
    489 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_SHARED, MAP_SHARED);
    490 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_PRIVATE, MAP_PRIVATE);
    491 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_FIXED, MAP_FIXED);
    492 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_ANON, MAP_ANON);
    493 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
    494 
    495 	SCARG(&cma,addr) = (void *)SCARG(uap, addr);
    496 	SCARG(&cma,len) = SCARG(uap, len);
    497 	SCARG(&cma,prot) = SCARG(uap, prot);
    498 	if (SCARG(&cma,prot) & VM_PROT_WRITE) /* XXX */
    499 		SCARG(&cma,prot) |= VM_PROT_READ;
    500 	SCARG(&cma,flags) = flags;
    501 	SCARG(&cma,fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
    502 	SCARG(&cma,pos) = (off_t)SCARG(uap, offset);
    503 
    504 	return sys_mmap(l, &cma, retval);
    505 }
    506 
    507 int
    508 linux_sys_mremap(l, v, retval)
    509 	struct lwp *l;
    510 	void *v;
    511 	register_t *retval;
    512 {
    513 	struct linux_sys_mremap_args /* {
    514 		syscallarg(void *) old_address;
    515 		syscallarg(size_t) old_size;
    516 		syscallarg(size_t) new_size;
    517 		syscallarg(u_long) flags;
    518 	} */ *uap = v;
    519 	struct sys_munmap_args mua;
    520 	size_t old_size, new_size;
    521 	int error;
    522 
    523 	old_size = round_page(SCARG(uap, old_size));
    524 	new_size = round_page(SCARG(uap, new_size));
    525 
    526 	/*
    527 	 * Growing mapped region.
    528 	 */
    529 	if (new_size > old_size) {
    530 		/*
    531 		 * XXX Implement me.  What we probably want to do is
    532 		 * XXX dig out the guts of the old mapping, mmap that
    533 		 * XXX object again with the new size, then munmap
    534 		 * XXX the old mapping.
    535 		 */
    536 		*retval = 0;
    537 		return (ENOMEM);
    538 	}
    539 
    540 	/*
    541 	 * Shrinking mapped region.
    542 	 */
    543 	if (new_size < old_size) {
    544 		SCARG(&mua, addr) = (caddr_t)SCARG(uap, old_address) +
    545 		    new_size;
    546 		SCARG(&mua, len) = old_size - new_size;
    547 		error = sys_munmap(l, &mua, retval);
    548 		*retval = error ? 0 : (register_t)SCARG(uap, old_address);
    549 		return (error);
    550 	}
    551 
    552 	/*
    553 	 * No change.
    554 	 */
    555 	*retval = (register_t)SCARG(uap, old_address);
    556 	return (0);
    557 }
    558 
    559 int
    560 linux_sys_msync(l, v, retval)
    561 	struct lwp *l;
    562 	void *v;
    563 	register_t *retval;
    564 {
    565 	struct linux_sys_msync_args /* {
    566 		syscallarg(caddr_t) addr;
    567 		syscallarg(int) len;
    568 		syscallarg(int) fl;
    569 	} */ *uap = v;
    570 
    571 	struct sys___msync13_args bma;
    572 
    573 	/* flags are ignored */
    574 	SCARG(&bma, addr) = SCARG(uap, addr);
    575 	SCARG(&bma, len) = SCARG(uap, len);
    576 	SCARG(&bma, flags) = SCARG(uap, fl);
    577 
    578 	return sys___msync13(l, &bma, retval);
    579 }
    580 
    581 /*
    582  * This code is partly stolen from src/lib/libc/compat-43/times.c
    583  * XXX - CLK_TCK isn't declared in /sys, just in <time.h>, done here
    584  */
    585 
    586 #define CLK_TCK 100
    587 #define	CONVTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
    588 
    589 int
    590 linux_sys_times(l, v, retval)
    591 	struct lwp *l;
    592 	void *v;
    593 	register_t *retval;
    594 {
    595 	struct linux_sys_times_args /* {
    596 		syscallarg(struct times *) tms;
    597 	} */ *uap = v;
    598 	struct proc *p = l->l_proc;
    599 	struct timeval t;
    600 	struct linux_tms ltms;
    601 	struct rusage ru;
    602 	int error, s;
    603 
    604 	calcru(p, &ru.ru_utime, &ru.ru_stime, NULL);
    605 	ltms.ltms_utime = CONVTCK(ru.ru_utime);
    606 	ltms.ltms_stime = CONVTCK(ru.ru_stime);
    607 
    608 	ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
    609 	ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
    610 
    611 	if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
    612 		return error;
    613 
    614 	s = splclock();
    615 	timersub(&time, &boottime, &t);
    616 	splx(s);
    617 
    618 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
    619 	return 0;
    620 }
    621 
    622 /*
    623  * Linux 'readdir' call. This code is mostly taken from the
    624  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
    625  * an attempt has been made to keep it a little cleaner (failing
    626  * miserably, because of the cruft needed if count 1 is passed).
    627  *
    628  * The d_off field should contain the offset of the next valid entry,
    629  * but in Linux it has the offset of the entry itself. We emulate
    630  * that bug here.
    631  *
    632  * Read in BSD-style entries, convert them, and copy them out.
    633  *
    634  * Note that this doesn't handle union-mounted filesystems.
    635  */
    636 int
    637 linux_sys_getdents(l, v, retval)
    638 	struct lwp *l;
    639 	void *v;
    640 	register_t *retval;
    641 {
    642 	struct linux_sys_getdents_args /* {
    643 		syscallarg(int) fd;
    644 		syscallarg(struct linux_dirent *) dent;
    645 		syscallarg(unsigned int) count;
    646 	} */ *uap = v;
    647 	struct proc *p = l->l_proc;
    648 	struct dirent *bdp;
    649 	struct vnode *vp;
    650 	caddr_t	inp, buf;		/* BSD-format */
    651 	int len, reclen;		/* BSD-format */
    652 	caddr_t outp;			/* Linux-format */
    653 	int resid, linux_reclen = 0;	/* Linux-format */
    654 	struct file *fp;
    655 	struct uio auio;
    656 	struct iovec aiov;
    657 	struct linux_dirent idb;
    658 	off_t off;		/* true file offset */
    659 	int buflen, error, eofflag, nbytes, oldcall;
    660 	struct vattr va;
    661 	off_t *cookiebuf = NULL, *cookie;
    662 	int ncookies;
    663 
    664 	/* getvnode() will use the descriptor for us */
    665 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
    666 		return (error);
    667 
    668 	if ((fp->f_flag & FREAD) == 0) {
    669 		error = EBADF;
    670 		goto out1;
    671 	}
    672 
    673 	vp = (struct vnode *)fp->f_data;
    674 	if (vp->v_type != VDIR) {
    675 		error = EINVAL;
    676 		goto out1;
    677 	}
    678 
    679 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)))
    680 		goto out1;
    681 
    682 	nbytes = SCARG(uap, count);
    683 	if (nbytes == 1) {	/* emulating old, broken behaviour */
    684 		nbytes = sizeof (struct linux_dirent);
    685 		buflen = max(va.va_blocksize, nbytes);
    686 		oldcall = 1;
    687 	} else {
    688 		buflen = min(MAXBSIZE, nbytes);
    689 		if (buflen < va.va_blocksize)
    690 			buflen = va.va_blocksize;
    691 		oldcall = 0;
    692 	}
    693 	buf = malloc(buflen, M_TEMP, M_WAITOK);
    694 
    695 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    696 	off = fp->f_offset;
    697 again:
    698 	aiov.iov_base = buf;
    699 	aiov.iov_len = buflen;
    700 	auio.uio_iov = &aiov;
    701 	auio.uio_iovcnt = 1;
    702 	auio.uio_rw = UIO_READ;
    703 	auio.uio_segflg = UIO_SYSSPACE;
    704 	auio.uio_procp = p;
    705 	auio.uio_resid = buflen;
    706 	auio.uio_offset = off;
    707 	/*
    708          * First we read into the malloc'ed buffer, then
    709          * we massage it into user space, one record at a time.
    710          */
    711 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
    712 	    &ncookies);
    713 	if (error)
    714 		goto out;
    715 
    716 	inp = buf;
    717 	outp = (caddr_t)SCARG(uap, dent);
    718 	resid = nbytes;
    719 	if ((len = buflen - auio.uio_resid) == 0)
    720 		goto eof;
    721 
    722 	for (cookie = cookiebuf; len > 0; len -= reclen) {
    723 		bdp = (struct dirent *)inp;
    724 		reclen = bdp->d_reclen;
    725 		if (reclen & 3)
    726 			panic("linux_readdir");
    727 		if (bdp->d_fileno == 0) {
    728 			inp += reclen;	/* it is a hole; squish it out */
    729 			off = *cookie++;
    730 			continue;
    731 		}
    732 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
    733 		if (reclen > len || resid < linux_reclen) {
    734 			/* entry too big for buffer, so just stop */
    735 			outp++;
    736 			break;
    737 		}
    738 		/*
    739 		 * Massage in place to make a Linux-shaped dirent (otherwise
    740 		 * we have to worry about touching user memory outside of
    741 		 * the copyout() call).
    742 		 */
    743 		idb.d_ino = (linux_ino_t)bdp->d_fileno;
    744 		/*
    745 		 * The old readdir() call misuses the offset and reclen fields.
    746 		 */
    747 		if (oldcall) {
    748 			idb.d_off = (linux_off_t)linux_reclen;
    749 			idb.d_reclen = (u_short)bdp->d_namlen;
    750 		} else {
    751 			if (sizeof (linux_off_t) < 4 && (off >> 32) != 0) {
    752 				compat_offseterr(vp, "linux_getdents");
    753 				error = EINVAL;
    754 				goto out;
    755 			}
    756 			idb.d_off = (linux_off_t)off;
    757 			idb.d_reclen = (u_short)linux_reclen;
    758 		}
    759 		strcpy(idb.d_name, bdp->d_name);
    760 		if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
    761 			goto out;
    762 		/* advance past this real entry */
    763 		inp += reclen;
    764 		off = *cookie++;	/* each entry points to itself */
    765 		/* advance output past Linux-shaped entry */
    766 		outp += linux_reclen;
    767 		resid -= linux_reclen;
    768 		if (oldcall)
    769 			break;
    770 	}
    771 
    772 	/* if we squished out the whole block, try again */
    773 	if (outp == (caddr_t)SCARG(uap, dent))
    774 		goto again;
    775 	fp->f_offset = off;	/* update the vnode offset */
    776 
    777 	if (oldcall)
    778 		nbytes = resid + linux_reclen;
    779 
    780 eof:
    781 	*retval = nbytes - resid;
    782 out:
    783 	VOP_UNLOCK(vp, 0);
    784 	if (cookiebuf)
    785 		free(cookiebuf, M_TEMP);
    786 	free(buf, M_TEMP);
    787  out1:
    788 	FILE_UNUSE(fp, p);
    789 	return error;
    790 }
    791 
    792 /*
    793  * Even when just using registers to pass arguments to syscalls you can
    794  * have 5 of them on the i386. So this newer version of select() does
    795  * this.
    796  */
    797 int
    798 linux_sys_select(l, v, retval)
    799 	struct lwp *l;
    800 	void *v;
    801 	register_t *retval;
    802 {
    803 	struct linux_sys_select_args /* {
    804 		syscallarg(int) nfds;
    805 		syscallarg(fd_set *) readfds;
    806 		syscallarg(fd_set *) writefds;
    807 		syscallarg(fd_set *) exceptfds;
    808 		syscallarg(struct timeval *) timeout;
    809 	} */ *uap = v;
    810 
    811 	return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds),
    812 	    SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout));
    813 }
    814 
    815 /*
    816  * Common code for the old and new versions of select(). A couple of
    817  * things are important:
    818  * 1) return the amount of time left in the 'timeout' parameter
    819  * 2) select never returns ERESTART on Linux, always return EINTR
    820  */
    821 int
    822 linux_select1(l, retval, nfds, readfds, writefds, exceptfds, timeout)
    823 	struct lwp *l;
    824 	register_t *retval;
    825 	int nfds;
    826 	fd_set *readfds, *writefds, *exceptfds;
    827 	struct timeval *timeout;
    828 {
    829 	struct sys_select_args bsa;
    830 	struct proc *p = l->l_proc;
    831 	struct timeval tv0, tv1, utv, *tvp;
    832 	caddr_t sg;
    833 	int error;
    834 
    835 	SCARG(&bsa, nd) = nfds;
    836 	SCARG(&bsa, in) = readfds;
    837 	SCARG(&bsa, ou) = writefds;
    838 	SCARG(&bsa, ex) = exceptfds;
    839 	SCARG(&bsa, tv) = timeout;
    840 
    841 	/*
    842 	 * Store current time for computation of the amount of
    843 	 * time left.
    844 	 */
    845 	if (timeout) {
    846 		if ((error = copyin(timeout, &utv, sizeof(utv))))
    847 			return error;
    848 		if (itimerfix(&utv)) {
    849 			/*
    850 			 * The timeval was invalid.  Convert it to something
    851 			 * valid that will act as it does under Linux.
    852 			 */
    853 			sg = stackgap_init(p->p_emul);
    854 			tvp = stackgap_alloc(&sg, sizeof(utv));
    855 			utv.tv_sec += utv.tv_usec / 1000000;
    856 			utv.tv_usec %= 1000000;
    857 			if (utv.tv_usec < 0) {
    858 				utv.tv_sec -= 1;
    859 				utv.tv_usec += 1000000;
    860 			}
    861 			if (utv.tv_sec < 0)
    862 				timerclear(&utv);
    863 			if ((error = copyout(&utv, tvp, sizeof(utv))))
    864 				return error;
    865 			SCARG(&bsa, tv) = tvp;
    866 		}
    867 		microtime(&tv0);
    868 	}
    869 
    870 	error = sys_select(l, &bsa, retval);
    871 	if (error) {
    872 		/*
    873 		 * See fs/select.c in the Linux kernel.  Without this,
    874 		 * Maelstrom doesn't work.
    875 		 */
    876 		if (error == ERESTART)
    877 			error = EINTR;
    878 		return error;
    879 	}
    880 
    881 	if (timeout) {
    882 		if (*retval) {
    883 			/*
    884 			 * Compute how much time was left of the timeout,
    885 			 * by subtracting the current time and the time
    886 			 * before we started the call, and subtracting
    887 			 * that result from the user-supplied value.
    888 			 */
    889 			microtime(&tv1);
    890 			timersub(&tv1, &tv0, &tv1);
    891 			timersub(&utv, &tv1, &utv);
    892 			if (utv.tv_sec < 0)
    893 				timerclear(&utv);
    894 		} else
    895 			timerclear(&utv);
    896 		if ((error = copyout(&utv, timeout, sizeof(utv))))
    897 			return error;
    898 	}
    899 
    900 	return 0;
    901 }
    902 
    903 /*
    904  * Get the process group of a certain process. Look it up
    905  * and return the value.
    906  */
    907 int
    908 linux_sys_getpgid(l, v, retval)
    909 	struct lwp *l;
    910 	void *v;
    911 	register_t *retval;
    912 {
    913 	struct linux_sys_getpgid_args /* {
    914 		syscallarg(int) pid;
    915 	} */ *uap = v;
    916 	struct proc *p = l->l_proc;
    917 	struct proc *targp;
    918 
    919 	if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) {
    920 		if ((targp = pfind(SCARG(uap, pid))) == 0)
    921 			return ESRCH;
    922 	}
    923 	else
    924 		targp = p;
    925 
    926 	retval[0] = targp->p_pgid;
    927 	return 0;
    928 }
    929 
    930 /*
    931  * Set the 'personality' (emulation mode) for the current process. Only
    932  * accept the Linux personality here (0). This call is needed because
    933  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
    934  * ELF binaries run in Linux mode, not SVR4 mode.
    935  */
    936 int
    937 linux_sys_personality(l, v, retval)
    938 	struct lwp *l;
    939 	void *v;
    940 	register_t *retval;
    941 {
    942 	struct linux_sys_personality_args /* {
    943 		syscallarg(int) per;
    944 	} */ *uap = v;
    945 
    946 	if (SCARG(uap, per) != 0)
    947 		return EINVAL;
    948 	retval[0] = 0;
    949 	return 0;
    950 }
    951 
    952 #if defined(__i386__) || defined(__m68k__)
    953 /*
    954  * The calls are here because of type conversions.
    955  */
    956 int
    957 linux_sys_setreuid16(l, v, retval)
    958 	struct lwp *l;
    959 	void *v;
    960 	register_t *retval;
    961 {
    962 	struct linux_sys_setreuid16_args /* {
    963 		syscallarg(int) ruid;
    964 		syscallarg(int) euid;
    965 	} */ *uap = v;
    966 	struct sys_setreuid_args bsa;
    967 
    968 	SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
    969 		(uid_t)-1 : SCARG(uap, ruid);
    970 	SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
    971 		(uid_t)-1 : SCARG(uap, euid);
    972 
    973 	return sys_setreuid(l, &bsa, retval);
    974 }
    975 
    976 int
    977 linux_sys_setregid16(l, v, retval)
    978 	struct lwp *l;
    979 	void *v;
    980 	register_t *retval;
    981 {
    982 	struct linux_sys_setregid16_args /* {
    983 		syscallarg(int) rgid;
    984 		syscallarg(int) egid;
    985 	} */ *uap = v;
    986 	struct sys_setregid_args bsa;
    987 
    988 	SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
    989 		(uid_t)-1 : SCARG(uap, rgid);
    990 	SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
    991 		(uid_t)-1 : SCARG(uap, egid);
    992 
    993 	return sys_setregid(l, &bsa, retval);
    994 }
    995 
    996 int
    997 linux_sys_setresuid16(l, v, retval)
    998 	struct lwp *l;
    999 	void *v;
   1000 	register_t *retval;
   1001 {
   1002 	struct linux_sys_setresuid16_args /* {
   1003 		syscallarg(uid_t) ruid;
   1004 		syscallarg(uid_t) euid;
   1005 		syscallarg(uid_t) suid;
   1006 	} */ *uap = v;
   1007 	struct linux_sys_setresuid16_args lsa;
   1008 
   1009 	SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
   1010 		(uid_t)-1 : SCARG(uap, ruid);
   1011 	SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
   1012 		(uid_t)-1 : SCARG(uap, euid);
   1013 	SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ?
   1014 		(uid_t)-1 : SCARG(uap, suid);
   1015 
   1016 	return linux_sys_setresuid(l, &lsa, retval);
   1017 }
   1018 
   1019 int
   1020 linux_sys_setresgid16(l, v, retval)
   1021 	struct lwp *l;
   1022 	void *v;
   1023 	register_t *retval;
   1024 {
   1025 	struct linux_sys_setresgid16_args /* {
   1026 		syscallarg(gid_t) rgid;
   1027 		syscallarg(gid_t) egid;
   1028 		syscallarg(gid_t) sgid;
   1029 	} */ *uap = v;
   1030 	struct linux_sys_setresgid16_args lsa;
   1031 
   1032 	SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
   1033 		(gid_t)-1 : SCARG(uap, rgid);
   1034 	SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
   1035 		(gid_t)-1 : SCARG(uap, egid);
   1036 	SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ?
   1037 		(gid_t)-1 : SCARG(uap, sgid);
   1038 
   1039 	return linux_sys_setresgid(l, &lsa, retval);
   1040 }
   1041 
   1042 int
   1043 linux_sys_getgroups16(l, v, retval)
   1044 	struct lwp *l;
   1045 	void *v;
   1046 	register_t *retval;
   1047 {
   1048 	struct linux_sys_getgroups16_args /* {
   1049 		syscallarg(int) gidsetsize;
   1050 		syscallarg(linux_gid_t *) gidset;
   1051 	} */ *uap = v;
   1052 	struct proc *p = l->l_proc;
   1053 	caddr_t sg;
   1054 	int n, error, i;
   1055 	struct sys_getgroups_args bsa;
   1056 	gid_t *bset, *kbset;
   1057 	linux_gid_t *lset;
   1058 	struct pcred *pc = p->p_cred;
   1059 
   1060 	n = SCARG(uap, gidsetsize);
   1061 	if (n < 0)
   1062 		return EINVAL;
   1063 	error = 0;
   1064 	bset = kbset = NULL;
   1065 	lset = NULL;
   1066 	if (n > 0) {
   1067 		n = min(pc->pc_ucred->cr_ngroups, n);
   1068 		sg = stackgap_init(p->p_emul);
   1069 		bset = stackgap_alloc(&sg, n * sizeof (gid_t));
   1070 		kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
   1071 		lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
   1072 		if (bset == NULL || kbset == NULL || lset == NULL)
   1073 			return ENOMEM;
   1074 		SCARG(&bsa, gidsetsize) = n;
   1075 		SCARG(&bsa, gidset) = bset;
   1076 		error = sys_getgroups(l, &bsa, retval);
   1077 		if (error != 0)
   1078 			goto out;
   1079 		error = copyin(bset, kbset, n * sizeof (gid_t));
   1080 		if (error != 0)
   1081 			goto out;
   1082 		for (i = 0; i < n; i++)
   1083 			lset[i] = (linux_gid_t)kbset[i];
   1084 		error = copyout(lset, SCARG(uap, gidset),
   1085 		    n * sizeof (linux_gid_t));
   1086 	} else
   1087 		*retval = pc->pc_ucred->cr_ngroups;
   1088 out:
   1089 	if (kbset != NULL)
   1090 		free(kbset, M_TEMP);
   1091 	if (lset != NULL)
   1092 		free(lset, M_TEMP);
   1093 	return error;
   1094 }
   1095 
   1096 int
   1097 linux_sys_setgroups16(l, v, retval)
   1098 	struct lwp *l;
   1099 	void *v;
   1100 	register_t *retval;
   1101 {
   1102 	struct linux_sys_setgroups16_args /* {
   1103 		syscallarg(int) gidsetsize;
   1104 		syscallarg(linux_gid_t *) gidset;
   1105 	} */ *uap = v;
   1106 	struct proc *p = l->l_proc;
   1107 	caddr_t sg;
   1108 	int n;
   1109 	int error, i;
   1110 	struct sys_setgroups_args bsa;
   1111 	gid_t *bset, *kbset;
   1112 	linux_gid_t *lset;
   1113 
   1114 	n = SCARG(uap, gidsetsize);
   1115 	if (n < 0 || n > NGROUPS)
   1116 		return EINVAL;
   1117 	sg = stackgap_init(p->p_emul);
   1118 	bset = stackgap_alloc(&sg, n * sizeof (gid_t));
   1119 	lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
   1120 	kbset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
   1121 	if (lset == NULL || bset == NULL)
   1122 		return ENOMEM;
   1123 	error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t));
   1124 	if (error != 0)
   1125 		goto out;
   1126 	for (i = 0; i < n; i++)
   1127 		kbset[i] = (gid_t)lset[i];
   1128 	error = copyout(kbset, bset, n * sizeof (gid_t));
   1129 	if (error != 0)
   1130 		goto out;
   1131 	SCARG(&bsa, gidsetsize) = n;
   1132 	SCARG(&bsa, gidset) = bset;
   1133 	error = sys_setgroups(l, &bsa, retval);
   1134 
   1135 out:
   1136 	if (lset != NULL)
   1137 		free(lset, M_TEMP);
   1138 	if (kbset != NULL)
   1139 		free(kbset, M_TEMP);
   1140 
   1141 	return error;
   1142 }
   1143 
   1144 #endif /* __i386__ || __m68k__ */
   1145 
   1146 /*
   1147  * We have nonexistent fsuid equal to uid.
   1148  * If modification is requested, refuse.
   1149  */
   1150 int
   1151 linux_sys_setfsuid(l, v, retval)
   1152 	 struct lwp *l;
   1153 	 void *v;
   1154 	 register_t *retval;
   1155 {
   1156 	 struct linux_sys_setfsuid_args /* {
   1157 		 syscallarg(uid_t) uid;
   1158 	 } */ *uap = v;
   1159 	 struct proc *p = l->l_proc;
   1160 	 uid_t uid;
   1161 
   1162 	 uid = SCARG(uap, uid);
   1163 	 if (p->p_cred->p_ruid != uid)
   1164 		 return sys_nosys(l, v, retval);
   1165 	 else
   1166 		 return (0);
   1167 }
   1168 
   1169 /* XXX XXX XXX */
   1170 #ifndef alpha
   1171 int
   1172 linux_sys_getfsuid(l, v, retval)
   1173 	struct lwp *l;
   1174 	void *v;
   1175 	register_t *retval;
   1176 {
   1177 	return sys_getuid(l, v, retval);
   1178 }
   1179 #endif
   1180 
   1181 int
   1182 linux_sys___sysctl(l, v, retval)
   1183 	struct lwp *l;
   1184 	void *v;
   1185 	register_t *retval;
   1186 {
   1187 	struct linux_sys___sysctl_args /* {
   1188 		syscallarg(struct linux___sysctl *) lsp;
   1189 	} */ *uap = v;
   1190 	struct linux___sysctl ls;
   1191 	struct sys___sysctl_args bsa;
   1192 	int error;
   1193 
   1194 	if ((error = copyin(SCARG(uap, lsp), &ls, sizeof ls)))
   1195 		return error;
   1196 	SCARG(&bsa, name) = ls.name;
   1197 	SCARG(&bsa, namelen) = ls.namelen;
   1198 	SCARG(&bsa, old) = ls.old;
   1199 	SCARG(&bsa, oldlenp) = ls.oldlenp;
   1200 	SCARG(&bsa, new) = ls.new;
   1201 	SCARG(&bsa, newlen) = ls.newlen;
   1202 
   1203 	return sys___sysctl(l, &bsa, retval);
   1204 }
   1205 
   1206 int
   1207 linux_sys_setresuid(l, v, retval)
   1208 	struct lwp *l;
   1209 	void *v;
   1210 	register_t *retval;
   1211 {
   1212 	struct linux_sys_setresuid_args /* {
   1213 		syscallarg(uid_t) ruid;
   1214 		syscallarg(uid_t) euid;
   1215 		syscallarg(uid_t) suid;
   1216 	} */ *uap = v;
   1217 	struct proc *p = l->l_proc;
   1218 	struct pcred *pc = p->p_cred;
   1219 	uid_t ruid, euid, suid;
   1220 	int error;
   1221 
   1222 	ruid = SCARG(uap, ruid);
   1223 	euid = SCARG(uap, euid);
   1224 	suid = SCARG(uap, suid);
   1225 
   1226 	/*
   1227 	 * Note: These checks are a little different than the NetBSD
   1228 	 * setreuid(2) call performs.  This precisely follows the
   1229 	 * behavior of the Linux kernel.
   1230 	 */
   1231 	if (ruid != (uid_t)-1 &&
   1232 	    ruid != pc->p_ruid &&
   1233 	    ruid != pc->pc_ucred->cr_uid &&
   1234 	    ruid != pc->p_svuid &&
   1235 	    (error = suser(pc->pc_ucred, &p->p_acflag)))
   1236 		return (error);
   1237 
   1238 	if (euid != (uid_t)-1 &&
   1239 	    euid != pc->p_ruid &&
   1240 	    euid != pc->pc_ucred->cr_uid &&
   1241 	    euid != pc->p_svuid &&
   1242 	    (error = suser(pc->pc_ucred, &p->p_acflag)))
   1243 		return (error);
   1244 
   1245 	if (suid != (uid_t)-1 &&
   1246 	    suid != pc->p_ruid &&
   1247 	    suid != pc->pc_ucred->cr_uid &&
   1248 	    suid != pc->p_svuid &&
   1249 	    (error = suser(pc->pc_ucred, &p->p_acflag)))
   1250 		return (error);
   1251 
   1252 	/*
   1253 	 * Now assign the new real, effective, and saved UIDs.
   1254 	 * Note that Linux, unlike NetBSD in setreuid(2), does not
   1255 	 * set the saved UID in this call unless the user specifies
   1256 	 * it.
   1257 	 */
   1258 	if (ruid != (uid_t)-1) {
   1259 		(void)chgproccnt(pc->p_ruid, -1);
   1260 		(void)chgproccnt(ruid, 1);
   1261 		pc->p_ruid = ruid;
   1262 	}
   1263 
   1264 	if (euid != (uid_t)-1) {
   1265 		pc->pc_ucred = crcopy(pc->pc_ucred);
   1266 		pc->pc_ucred->cr_uid = euid;
   1267 	}
   1268 
   1269 	if (suid != (uid_t)-1)
   1270 		pc->p_svuid = suid;
   1271 
   1272 	if (ruid != (uid_t)-1 && euid != (uid_t)-1 && suid != (uid_t)-1)
   1273 		p->p_flag |= P_SUGID;
   1274 	return (0);
   1275 }
   1276 
   1277 int
   1278 linux_sys_getresuid(l, v, retval)
   1279 	struct lwp *l;
   1280 	void *v;
   1281 	register_t *retval;
   1282 {
   1283 	struct linux_sys_getresuid_args /* {
   1284 		syscallarg(uid_t *) ruid;
   1285 		syscallarg(uid_t *) euid;
   1286 		syscallarg(uid_t *) suid;
   1287 	} */ *uap = v;
   1288 	struct proc *p = l->l_proc;
   1289 	struct pcred *pc = p->p_cred;
   1290 	int error;
   1291 
   1292 	/*
   1293 	 * Linux copies these values out to userspace like so:
   1294 	 *
   1295 	 *	1. Copy out ruid.
   1296 	 *	2. If that succeeds, copy out euid.
   1297 	 *	3. If both of those succeed, copy out suid.
   1298 	 */
   1299 	if ((error = copyout(&pc->p_ruid, SCARG(uap, ruid),
   1300 			     sizeof(uid_t))) != 0)
   1301 		return (error);
   1302 
   1303 	if ((error = copyout(&pc->pc_ucred->cr_uid, SCARG(uap, euid),
   1304 			     sizeof(uid_t))) != 0)
   1305 		return (error);
   1306 
   1307 	return (copyout(&pc->p_svuid, SCARG(uap, suid), sizeof(uid_t)));
   1308 }
   1309 
   1310 int
   1311 linux_sys_ptrace(l, v, retval)
   1312 	struct lwp *l;
   1313 	void *v;
   1314 	register_t *retval;
   1315 {
   1316 	struct linux_sys_ptrace_args /* {
   1317 		i386, m68k, powerpc: T=int
   1318 		alpha: T=long
   1319 		syscallarg(T) request;
   1320 		syscallarg(T) pid;
   1321 		syscallarg(T) addr;
   1322 		syscallarg(T) data;
   1323 	} */ *uap = v;
   1324 	const int *ptr;
   1325 	int request;
   1326 	int error;
   1327 
   1328 	ptr = linux_ptrace_request_map;
   1329 	request = SCARG(uap, request);
   1330 	while (*ptr != -1)
   1331 		if (*ptr++ == request) {
   1332 			struct sys_ptrace_args pta;
   1333 
   1334 			SCARG(&pta, req) = *ptr;
   1335 			SCARG(&pta, pid) = SCARG(uap, pid);
   1336 			SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr);
   1337 			SCARG(&pta, data) = SCARG(uap, data);
   1338 
   1339 			/*
   1340 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
   1341 			 * to continue where the process left off previously.
   1342 			 * The same thing is achieved by addr == (caddr_t) 1
   1343 			 * on NetBSD, so rewrite 'addr' appropriately.
   1344 			 */
   1345 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
   1346 				SCARG(&pta, addr) = (caddr_t) 1;
   1347 
   1348 			error = sys_ptrace(l, &pta, retval);
   1349 			if (error)
   1350 				return error;
   1351 			switch (request) {
   1352 			case LINUX_PTRACE_PEEKTEXT:
   1353 			case LINUX_PTRACE_PEEKDATA:
   1354 				error = copyout (retval,
   1355 				    (caddr_t)SCARG(uap, data), sizeof *retval);
   1356 				*retval = SCARG(uap, data);
   1357 				break;
   1358 			default:
   1359 				break;
   1360 			}
   1361 			return error;
   1362 		}
   1363 		else
   1364 			ptr++;
   1365 
   1366 	return LINUX_SYS_PTRACE_ARCH(l, uap, retval);
   1367 }
   1368 
   1369 int
   1370 linux_sys_reboot(struct lwp *l, void *v, register_t *retval)
   1371 {
   1372 	struct linux_sys_reboot_args /* {
   1373 		syscallarg(int) magic1;
   1374 		syscallarg(int) magic2;
   1375 		syscallarg(int) cmd;
   1376 		syscallarg(void *) arg;
   1377 	} */ *uap = v;
   1378 	struct sys_reboot_args /* {
   1379 		syscallarg(int) opt;
   1380 		syscallarg(char *) bootstr;
   1381 	} */ sra;
   1382 	struct proc *p = l->l_proc;
   1383 	int error;
   1384 
   1385 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
   1386 		return(error);
   1387 
   1388 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
   1389 		return(EINVAL);
   1390 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
   1391 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
   1392 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
   1393 		return(EINVAL);
   1394 
   1395 	switch (SCARG(uap, cmd)) {
   1396 	case LINUX_REBOOT_CMD_RESTART:
   1397 		SCARG(&sra, opt) = RB_AUTOBOOT;
   1398 		break;
   1399 	case LINUX_REBOOT_CMD_HALT:
   1400 		SCARG(&sra, opt) = RB_HALT;
   1401 		break;
   1402 	case LINUX_REBOOT_CMD_POWER_OFF:
   1403 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
   1404 		break;
   1405 	case LINUX_REBOOT_CMD_RESTART2:
   1406 		/* Reboot with an argument. */
   1407 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
   1408 		SCARG(&sra, bootstr) = SCARG(uap, arg);
   1409 		break;
   1410 	case LINUX_REBOOT_CMD_CAD_ON:
   1411 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
   1412 	case LINUX_REBOOT_CMD_CAD_OFF:
   1413 		return(0);
   1414 	default:
   1415 		return(EINVAL);
   1416 	}
   1417 
   1418 	return(sys_reboot(l, &sra, retval));
   1419 }
   1420 
   1421 /*
   1422  * Copy of compat_12_sys_swapon().
   1423  */
   1424 int
   1425 linux_sys_swapon(l, v, retval)
   1426 	struct lwp *l;
   1427 	void *v;
   1428 	register_t *retval;
   1429 {
   1430 	struct sys_swapctl_args ua;
   1431 	struct linux_sys_swapon_args /* {
   1432 		syscallarg(const char *) name;
   1433 	} */ *uap = v;
   1434 
   1435 	SCARG(&ua, cmd) = SWAP_ON;
   1436 	SCARG(&ua, arg) = (void *)SCARG(uap, name);
   1437 	SCARG(&ua, misc) = 0;	/* priority */
   1438 	return (sys_swapctl(l, &ua, retval));
   1439 }
   1440 
   1441 /*
   1442  * Stop swapping to the file or block device specified by path.
   1443  */
   1444 int
   1445 linux_sys_swapoff(l, v, retval)
   1446 	struct lwp *l;
   1447 	void *v;
   1448 	register_t *retval;
   1449 {
   1450 	struct sys_swapctl_args ua;
   1451 	struct linux_sys_swapoff_args /* {
   1452 		syscallarg(const char *) path;
   1453 	} */ *uap = v;
   1454 
   1455 	SCARG(&ua, cmd) = SWAP_OFF;
   1456 	SCARG(&ua, arg) = (void *)SCARG(uap, path);
   1457 	return (sys_swapctl(l, &ua, retval));
   1458 }
   1459 
   1460 /*
   1461  * Copy of compat_09_sys_setdomainname()
   1462  */
   1463 /* ARGSUSED */
   1464 int
   1465 linux_sys_setdomainname(l, v, retval)
   1466 	struct lwp *l;
   1467 	void *v;
   1468 	register_t *retval;
   1469 {
   1470 	struct linux_sys_setdomainname_args /* {
   1471 		syscallarg(char *) domainname;
   1472 		syscallarg(int) len;
   1473 	} */ *uap = v;
   1474 	struct proc *p = l->l_proc;
   1475 	int name;
   1476 	int error;
   1477 
   1478 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
   1479 		return (error);
   1480 	name = KERN_DOMAINNAME;
   1481 	return (kern_sysctl(&name, 1, 0, 0, SCARG(uap, domainname),
   1482 			    SCARG(uap, len), p));
   1483 }
   1484 
   1485 /*
   1486  * sysinfo()
   1487  */
   1488 /* ARGSUSED */
   1489 int
   1490 linux_sys_sysinfo(l, v, retval)
   1491 	struct lwp *l;
   1492 	void *v;
   1493 	register_t *retval;
   1494 {
   1495 	struct linux_sys_sysinfo_args /* {
   1496 		syscallarg(struct linux_sysinfo *) arg;
   1497 	} */ *uap = v;
   1498 	struct linux_sysinfo si;
   1499 	struct loadavg *la;
   1500 
   1501 	si.uptime = time.tv_sec - boottime.tv_sec;
   1502 	la = &averunnable;
   1503 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
   1504 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
   1505 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
   1506 	si.totalram = ctob(physmem);
   1507 	si.freeram = uvmexp.free * uvmexp.pagesize;
   1508 	si.sharedram = 0;	/* XXX */
   1509 	si.bufferram = uvmexp.filepages * uvmexp.pagesize;
   1510 	si.totalswap = uvmexp.swpages * uvmexp.pagesize;
   1511 	si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
   1512 	si.procs = nprocs;
   1513 
   1514 	/* The following are only present in newer Linux kernels. */
   1515 	si.totalbig = 0;
   1516 	si.freebig = 0;
   1517 	si.mem_unit = 1;
   1518 
   1519 	return (copyout(&si, SCARG(uap, arg), sizeof si));
   1520 }
   1521 
   1522 #define bsd_to_linux_rlimit1(l, b, f) \
   1523     (l)->f = ((b)->f == RLIM_INFINITY || ((b)->f & 0xffffffff00000000) != 0) ? \
   1524     LINUX_RLIM_INFINITY : (int32_t)(b)->f
   1525 #define bsd_to_linux_rlimit(l, b) \
   1526     bsd_to_linux_rlimit1(l, b, rlim_cur); \
   1527     bsd_to_linux_rlimit1(l, b, rlim_max)
   1528 
   1529 #define linux_to_bsd_rlimit1(b, l, f) \
   1530     (b)->f = (l)->f == LINUX_RLIM_INFINITY ? RLIM_INFINITY : (b)->f
   1531 #define linux_to_bsd_rlimit(b, l) \
   1532     linux_to_bsd_rlimit1(b, l, rlim_cur); \
   1533     linux_to_bsd_rlimit1(b, l, rlim_max)
   1534 
   1535 static int
   1536 linux_to_bsd_limit(lim)
   1537 	int lim;
   1538 {
   1539 	switch (lim) {
   1540 	case LINUX_RLIMIT_CPU:
   1541 		return RLIMIT_CPU;
   1542 	case LINUX_RLIMIT_FSIZE:
   1543 		return RLIMIT_FSIZE;
   1544 	case LINUX_RLIMIT_DATA:
   1545 		return RLIMIT_DATA;
   1546 	case LINUX_RLIMIT_STACK:
   1547 		return RLIMIT_STACK;
   1548 	case LINUX_RLIMIT_CORE:
   1549 		return RLIMIT_CORE;
   1550 	case LINUX_RLIMIT_RSS:
   1551 		return RLIMIT_RSS;
   1552 	case LINUX_RLIMIT_NPROC:
   1553 		return RLIMIT_NPROC;
   1554 	case LINUX_RLIMIT_NOFILE:
   1555 		return RLIMIT_NOFILE;
   1556 	case LINUX_RLIMIT_MEMLOCK:
   1557 		return RLIMIT_MEMLOCK;
   1558 	case LINUX_RLIMIT_AS:
   1559 	case LINUX_RLIMIT_LOCKS:
   1560 		return -EOPNOTSUPP;
   1561 	default:
   1562 		return -EINVAL;
   1563 	}
   1564 }
   1565 
   1566 
   1567 int
   1568 linux_sys_getrlimit(p, v, retval)
   1569 	struct proc *p;
   1570 	void *v;
   1571 	register_t *retval;
   1572 {
   1573 	struct linux_sys_getrlimit_args /* {
   1574 		syscallarg(int) which;
   1575 		syscallarg(struct orlimit *) rlp;
   1576 	} */ *uap = v;
   1577 	caddr_t sg = stackgap_init(p->p_emul);
   1578 	struct sys_getrlimit_args ap;
   1579 	struct rlimit rl;
   1580 	struct orlimit orl;
   1581 	int error;
   1582 
   1583 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
   1584 	if ((error = SCARG(&ap, which)) < 0)
   1585 		return -error;
   1586 	SCARG(&ap, rlp) = stackgap_alloc(&sg, sizeof rl);
   1587 	if ((error = sys_getrlimit(p, &ap, retval)) != 0)
   1588 		return error;
   1589 	if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0)
   1590 		return error;
   1591 	bsd_to_linux_rlimit(&orl, &rl);
   1592 	return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
   1593 }
   1594 
   1595 int
   1596 linux_sys_setrlimit(p, v, retval)
   1597 	struct proc *p;
   1598 	void *v;
   1599 	register_t *retval;
   1600 {
   1601 	struct linux_sys_setrlimit_args /* {
   1602 		syscallarg(int) which;
   1603 		syscallarg(struct orlimit *) rlp;
   1604 	} */ *uap = v;
   1605 	caddr_t sg = stackgap_init(p->p_emul);
   1606 	struct sys_setrlimit_args ap;
   1607 	struct rlimit rl;
   1608 	struct orlimit orl;
   1609 	int error;
   1610 
   1611 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
   1612 	SCARG(&ap, rlp) = stackgap_alloc(&sg, sizeof rl);
   1613 	if ((error = SCARG(&ap, which)) < 0)
   1614 		return -error;
   1615 	if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
   1616 		return error;
   1617 	linux_to_bsd_rlimit(&rl, &orl);
   1618 	/* XXX: alpha complains about this */
   1619 	if ((error = copyout(&rl, (void *)SCARG(&ap, rlp), sizeof(rl))) != 0)
   1620 		return error;
   1621 	return sys_setrlimit(p, &ap, retval);
   1622 }
   1623 
   1624 #ifndef __mips__
   1625 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
   1626 int
   1627 linux_sys_ugetrlimit(p, v, retval)
   1628 	struct proc *p;
   1629 	void *v;
   1630 	register_t *retval;
   1631 {
   1632 	return linux_sys_getrlimit(p, v, retval);
   1633 }
   1634 #endif
   1635 
   1636 /*
   1637  * This gets called for unsupported syscalls. The difference to sys_nosys()
   1638  * is that process does not get SIGSYS, the call just returns with ENOSYS.
   1639  * This is the way Linux does it and glibc depends on this behaviour.
   1640  */
   1641 int
   1642 linux_sys_nosys(l, v, retval)
   1643 	struct lwp *l;
   1644 	void *v;
   1645 	register_t *retval;
   1646 {
   1647 	return (ENOSYS);
   1648 }
   1649