No OneTemporary
Actions

Size

96 KB

Referenced Files

None

Subscribers

None

View Options

This document is not UTF8. It was detected as ISO-8859-1 (Latin 1) and converted to UTF8 for display.

	Index: stable/8/sys/amd64/include/xen
	===================================================================
	--- stable/8/sys/amd64/include/xen (revision 219197)
	+++ stable/8/sys/amd64/include/xen (revision 219198)

	Property changes on: stable/8/sys/amd64/include/xen
	___________________________________________________________________
	Modified: svn:mergeinfo
	## -0,0 +0,1 ##
	Merged /head/sys/amd64/include/xen:r218658,218668
	Index: stable/8/sys/amd64/linux32/linux32_sysvec.c
	===================================================================
	--- stable/8/sys/amd64/linux32/linux32_sysvec.c (revision 219197)
	+++ stable/8/sys/amd64/linux32/linux32_sysvec.c (revision 219198)
	@@ -1,1222 +1,1222 @@
	/*-
	* Copyright (c) 2004 Tim J. Robbins
	* Copyright (c) 2003 Peter Wemm
	* Copyright (c) 2002 Doug Rabson
	* Copyright (c) 1998-1999 Andrew Gallatin
	* Copyright (c) 1994-1996 Søren Schmidt
	* All rights reserved.
	*
	* Redistribution and use in source and binary forms, with or without
	* modification, are permitted provided that the following conditions
	* are met:
	* 1. Redistributions of source code must retain the above copyright
	* notice, this list of conditions and the following disclaimer
	* in this position and unchanged.
	* 2. Redistributions in binary form must reproduce the above copyright
	* notice, this list of conditions and the following disclaimer in the
	* documentation and/or other materials provided with the distribution.
	* 3. The name of the author may not be used to endorse or promote products
	* derived from this software without specific prior written permission
	*
	* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
	* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
	* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
	* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
	* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
	* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
	* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
	* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
	* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
	* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
	*/

	#include <sys/cdefs.h>
	__FBSDID("$FreeBSD$");
	#include "opt_compat.h"

	#ifndef COMPAT_FREEBSD32
	#error "Unable to compile Linux-emulator due to missing COMPAT_FREEBSD32 option!"
	#endif

	#define __ELF_WORD_SIZE 32

	#include <sys/param.h>
	#include <sys/systm.h>
	#include <sys/exec.h>
	#include <sys/fcntl.h>
	#include <sys/imgact.h>
	#include <sys/imgact_elf.h>
	#include <sys/kernel.h>
	#include <sys/lock.h>
	#include <sys/malloc.h>
	#include <sys/module.h>
	#include <sys/mutex.h>
	#include <sys/proc.h>
	#include <sys/resourcevar.h>
	#include <sys/signalvar.h>
	#include <sys/sysctl.h>
	#include <sys/syscallsubr.h>
	#include <sys/sysent.h>
	#include <sys/sysproto.h>
	#include <sys/vnode.h>
	#include <sys/eventhandler.h>

	#include <vm/vm.h>
	#include <vm/pmap.h>
	#include <vm/vm_extern.h>
	#include <vm/vm_map.h>
	#include <vm/vm_object.h>
	#include <vm/vm_page.h>
	#include <vm/vm_param.h>

	#include <machine/cpu.h>
	#include <machine/md_var.h>
	#include <machine/pcb.h>
	#include <machine/specialreg.h>

	#include <amd64/linux32/linux.h>
	#include <amd64/linux32/linux32_proto.h>
	-#include <compat/linux/linux_futex.h>
	#include <compat/linux/linux_emul.h>
	+#include <compat/linux/linux_futex.h>
	#include <compat/linux/linux_mib.h>
	#include <compat/linux/linux_misc.h>
	#include <compat/linux/linux_signal.h>
	#include <compat/linux/linux_util.h>

	MODULE_VERSION(linux, 1);

	MALLOC_DEFINE(M_LINUX, "linux", "Linux mode structures");

	#define AUXARGS_ENTRY_32(pos, id, val) \
	do { \
	suword32(pos++, id); \
	suword32(pos++, val); \
	} while (0)

	#if BYTE_ORDER == LITTLE_ENDIAN
	#define SHELLMAGIC 0x2123 /* #! */
	#else
	#define SHELLMAGIC 0x2321
	#endif

	/*
	* Allow the sendsig functions to use the ldebug() facility
	* even though they are not syscalls themselves. Map them
	* to syscall 0. This is slightly less bogus than using
	* ldebug(sigreturn).
	*/
	#define LINUX_SYS_linux_rt_sendsig 0
	#define LINUX_SYS_linux_sendsig 0

	const char *linux_platform = "i686";
	static int linux_szplatform;
	extern char linux_sigcode[];
	extern int linux_szsigcode;

	extern struct sysent linux_sysent[LINUX_SYS_MAXSYSCALL];

	SET_DECLARE(linux_ioctl_handler_set, struct linux_ioctl_handler);
	SET_DECLARE(linux_device_handler_set, struct linux_device_handler);

	static int elf_linux_fixup(register_t **stack_base,
	struct image_params *iparams);
	static register_t linux_copyout_strings(struct image_params imgp);
	static void linux_sendsig(sig_t catcher, ksiginfo_t ksi, sigset_t mask);
	static void exec_linux_setregs(struct thread *td, u_long entry,
	u_long stack, u_long ps_strings);
	static void linux32_fixlimit(struct rlimit *rl, int which);
	static boolean_t linux32_trans_osrel(const Elf_Note note, int32_t osrel);

	static eventhandler_tag linux_exit_tag;
	static eventhandler_tag linux_schedtail_tag;
	static eventhandler_tag linux_exec_tag;

	/*
	* Linux syscalls return negative errno's, we do positive and map them
	* Reference:
	* FreeBSD: src/sys/sys/errno.h
	* Linux: linux-2.6.17.8/include/asm-generic/errno-base.h
	* linux-2.6.17.8/include/asm-generic/errno.h
	*/
	static int bsd_to_linux_errno[ELAST + 1] = {
	-0, -1, -2, -3, -4, -5, -6, -7, -8, -9,
	-10, -35, -12, -13, -14, -15, -16, -17, -18, -19,
	-20, -21, -22, -23, -24, -25, -26, -27, -28, -29,
	-30, -31, -32, -33, -34, -11,-115,-114, -88, -89,
	-90, -91, -92, -93, -94, -95, -96, -97, -98, -99,
	-100,-101,-102,-103,-104,-105,-106,-107,-108,-109,
	-110,-111, -40, -36,-112,-113, -39, -11, -87,-122,
	-116, -66, -6, -6, -6, -6, -6, -37, -38, -9,
	-6, -6, -43, -42, -75,-125, -84, -95, -16, -74,
	-72, -67, -71
	};

	int bsd_to_linux_signal[LINUX_SIGTBLSZ] = {
	LINUX_SIGHUP, LINUX_SIGINT, LINUX_SIGQUIT, LINUX_SIGILL,
	LINUX_SIGTRAP, LINUX_SIGABRT, 0, LINUX_SIGFPE,
	LINUX_SIGKILL, LINUX_SIGBUS, LINUX_SIGSEGV, LINUX_SIGSYS,
	LINUX_SIGPIPE, LINUX_SIGALRM, LINUX_SIGTERM, LINUX_SIGURG,
	LINUX_SIGSTOP, LINUX_SIGTSTP, LINUX_SIGCONT, LINUX_SIGCHLD,
	LINUX_SIGTTIN, LINUX_SIGTTOU, LINUX_SIGIO, LINUX_SIGXCPU,
	LINUX_SIGXFSZ, LINUX_SIGVTALRM, LINUX_SIGPROF, LINUX_SIGWINCH,
	0, LINUX_SIGUSR1, LINUX_SIGUSR2
	};

	int linux_to_bsd_signal[LINUX_SIGTBLSZ] = {
	SIGHUP, SIGINT, SIGQUIT, SIGILL,
	SIGTRAP, SIGABRT, SIGBUS, SIGFPE,
	SIGKILL, SIGUSR1, SIGSEGV, SIGUSR2,
	SIGPIPE, SIGALRM, SIGTERM, SIGBUS,
	SIGCHLD, SIGCONT, SIGSTOP, SIGTSTP,
	SIGTTIN, SIGTTOU, SIGURG, SIGXCPU,
	SIGXFSZ, SIGVTALRM, SIGPROF, SIGWINCH,
	SIGIO, SIGURG, SIGSYS
	};

	#define LINUX_T_UNKNOWN 255
	static int _bsd_to_linux_trapcode[] = {
	LINUX_T_UNKNOWN, /* 0 */
	6, /* 1 T_PRIVINFLT */
	LINUX_T_UNKNOWN, /* 2 */
	3, /* 3 T_BPTFLT */
	LINUX_T_UNKNOWN, /* 4 */
	LINUX_T_UNKNOWN, /* 5 */
	16, /* 6 T_ARITHTRAP */
	254, /* 7 T_ASTFLT */
	LINUX_T_UNKNOWN, /* 8 */
	13, /* 9 T_PROTFLT */
	1, /* 10 T_TRCTRAP */
	LINUX_T_UNKNOWN, /* 11 */
	14, /* 12 T_PAGEFLT */
	LINUX_T_UNKNOWN, /* 13 */
	17, /* 14 T_ALIGNFLT */
	LINUX_T_UNKNOWN, /* 15 */
	LINUX_T_UNKNOWN, /* 16 */
	LINUX_T_UNKNOWN, /* 17 */
	0, /* 18 T_DIVIDE */
	2, /* 19 T_NMI */
	4, /* 20 T_OFLOW */
	5, /* 21 T_BOUND */
	7, /* 22 T_DNA */
	8, /* 23 T_DOUBLEFLT */
	9, /* 24 T_FPOPFLT */
	10, /* 25 T_TSSFLT */
	11, /* 26 T_SEGNPFLT */
	12, /* 27 T_STKFLT */
	18, /* 28 T_MCHK */
	19, /* 29 T_XMMFLT */
	15 /* 30 T_RESERVED */
	};
	#define bsd_to_linux_trapcode(code) \
	((code)<sizeof(_bsd_to_linux_trapcode)/sizeof(*_bsd_to_linux_trapcode)? \
	_bsd_to_linux_trapcode[(code)]: \
	LINUX_T_UNKNOWN)

	struct linux32_ps_strings {
	u_int32_t ps_argvstr; /* first of 0 or more argument strings */
	u_int ps_nargvstr; /* the number of argument strings */
	u_int32_t ps_envstr; /* first of 0 or more environment strings */
	u_int ps_nenvstr; /* the number of environment strings */
	};

	/*
	* If FreeBSD & Linux have a difference of opinion about what a trap
	* means, deal with it here.
	*
	* MPSAFE
	*/
	static int
	translate_traps(int signal, int trap_code)
	{
	if (signal != SIGBUS)
	return signal;
	switch (trap_code) {
	case T_PROTFLT:
	case T_TSSFLT:
	case T_DOUBLEFLT:
	case T_PAGEFLT:
	return SIGSEGV;
	default:
	return signal;
	}
	}

	static int
	elf_linux_fixup(register_t *stack_base, struct image_params imgp)
	{
	Elf32_Auxargs *args;
	Elf32_Addr *base;
	Elf32_Addr pos, uplatform;
	struct linux32_ps_strings *arginfo;

	arginfo = (struct linux32_ps_strings *)LINUX32_PS_STRINGS;
	uplatform = (Elf32_Addr *)((caddr_t)arginfo - linux_szsigcode -
	linux_szplatform);

	KASSERT(curthread->td_proc == imgp->proc,
	("unsafe elf_linux_fixup(), should be curproc"));
	base = (Elf32_Addr )stack_base;
	args = (Elf32_Auxargs *)imgp->auxargs;
	pos = base + (imgp->args->argc + imgp->args->envc + 2);

	AUXARGS_ENTRY_32(pos, LINUX_AT_HWCAP, cpu_feature);

	/*
	* Do not export AT_CLKTCK when emulating Linux kernel prior to 2.4.0,
	* as it has appeared in the 2.4.0-rc7 first time.
	* Being exported, AT_CLKTCK is returned by sysconf(_SC_CLK_TCK),
	* glibc falls back to the hard-coded CLK_TCK value when aux entry
	* is not present.
	* Also see linux_times() implementation.
	*/
	if (linux_kernver(curthread) >= LINUX_KERNVER_2004000)
	AUXARGS_ENTRY_32(pos, LINUX_AT_CLKTCK, stclohz);
	AUXARGS_ENTRY_32(pos, AT_PHDR, args->phdr);
	AUXARGS_ENTRY_32(pos, AT_PHENT, args->phent);
	AUXARGS_ENTRY_32(pos, AT_PHNUM, args->phnum);
	AUXARGS_ENTRY_32(pos, AT_PAGESZ, args->pagesz);
	AUXARGS_ENTRY_32(pos, AT_FLAGS, args->flags);
	AUXARGS_ENTRY_32(pos, AT_ENTRY, args->entry);
	AUXARGS_ENTRY_32(pos, AT_BASE, args->base);
	AUXARGS_ENTRY_32(pos, LINUX_AT_SECURE, 0);
	AUXARGS_ENTRY_32(pos, AT_UID, imgp->proc->p_ucred->cr_ruid);
	AUXARGS_ENTRY_32(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid);
	AUXARGS_ENTRY_32(pos, AT_GID, imgp->proc->p_ucred->cr_rgid);
	AUXARGS_ENTRY_32(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid);
	AUXARGS_ENTRY_32(pos, LINUX_AT_PLATFORM, PTROUT(uplatform));
	if (args->execfd != -1)
	AUXARGS_ENTRY_32(pos, AT_EXECFD, args->execfd);
	AUXARGS_ENTRY_32(pos, AT_NULL, 0);

	free(imgp->auxargs, M_TEMP);
	imgp->auxargs = NULL;

	base--;
	suword32(base, (uint32_t)imgp->args->argc);
	stack_base = (register_t )base;
	return 0;
	}

	extern unsigned long linux_sznonrtsigcode;

	static void
	linux_rt_sendsig(sig_t catcher, ksiginfo_t ksi, sigset_t mask)
	{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct sigacts *psp;
	struct trapframe *regs;
	struct l_rt_sigframe *fp, frame;
	int oonstack;
	int sig;
	int code;

	sig = ksi->ksi_signo;
	code = ksi->ksi_code;
	PROC_LOCK_ASSERT(p, MA_OWNED);
	psp = p->p_sigacts;
	mtx_assert(&psp->ps_mtx, MA_OWNED);
	regs = td->td_frame;
	oonstack = sigonstack(regs->tf_rsp);

	#ifdef DEBUG
	if (ldebug(rt_sendsig))
	printf(ARGS(rt_sendsig, "%p, %d, %p, %u"),
	catcher, sig, (void*)mask, code);
	#endif
	/*
	* Allocate space for the signal handler context.
	*/
	if ((td->td_pflags & TDP_ALTSTACK) && !oonstack &&
	SIGISMEMBER(psp->ps_sigonstack, sig)) {
	fp = (struct l_rt_sigframe *)(td->td_sigstk.ss_sp +
	td->td_sigstk.ss_size - sizeof(struct l_rt_sigframe));
	} else
	fp = (struct l_rt_sigframe *)regs->tf_rsp - 1;
	mtx_unlock(&psp->ps_mtx);

	/*
	* Build the argument list for the signal handler.
	*/
	if (p->p_sysent->sv_sigtbl)
	if (sig <= p->p_sysent->sv_sigsize)
	sig = p->p_sysent->sv_sigtbl[_SIG_IDX(sig)];

	bzero(&frame, sizeof(frame));

	frame.sf_handler = PTROUT(catcher);
	frame.sf_sig = sig;
	frame.sf_siginfo = PTROUT(&fp->sf_si);
	frame.sf_ucontext = PTROUT(&fp->sf_sc);

	/* Fill in POSIX parts */
	ksiginfo_to_lsiginfo(ksi, &frame.sf_si, sig);

	/*
	* Build the signal context to be used by sigreturn.
	*/
	frame.sf_sc.uc_flags = 0; /* XXX ??? */
	frame.sf_sc.uc_link = 0; /* XXX ??? */

	frame.sf_sc.uc_stack.ss_sp = PTROUT(td->td_sigstk.ss_sp);
	frame.sf_sc.uc_stack.ss_size = td->td_sigstk.ss_size;
	frame.sf_sc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK)
	? ((oonstack) ? LINUX_SS_ONSTACK : 0) : LINUX_SS_DISABLE;
	PROC_UNLOCK(p);

	bsd_to_linux_sigset(mask, &frame.sf_sc.uc_sigmask);

	frame.sf_sc.uc_mcontext.sc_mask = frame.sf_sc.uc_sigmask.__bits[0];
	frame.sf_sc.uc_mcontext.sc_edi = regs->tf_rdi;
	frame.sf_sc.uc_mcontext.sc_esi = regs->tf_rsi;
	frame.sf_sc.uc_mcontext.sc_ebp = regs->tf_rbp;
	frame.sf_sc.uc_mcontext.sc_ebx = regs->tf_rbx;
	frame.sf_sc.uc_mcontext.sc_edx = regs->tf_rdx;
	frame.sf_sc.uc_mcontext.sc_ecx = regs->tf_rcx;
	frame.sf_sc.uc_mcontext.sc_eax = regs->tf_rax;
	frame.sf_sc.uc_mcontext.sc_eip = regs->tf_rip;
	frame.sf_sc.uc_mcontext.sc_cs = regs->tf_cs;
	frame.sf_sc.uc_mcontext.sc_gs = regs->tf_gs;
	frame.sf_sc.uc_mcontext.sc_fs = regs->tf_fs;
	frame.sf_sc.uc_mcontext.sc_es = regs->tf_es;
	frame.sf_sc.uc_mcontext.sc_ds = regs->tf_ds;
	frame.sf_sc.uc_mcontext.sc_eflags = regs->tf_rflags;
	frame.sf_sc.uc_mcontext.sc_esp_at_signal = regs->tf_rsp;
	frame.sf_sc.uc_mcontext.sc_ss = regs->tf_ss;
	frame.sf_sc.uc_mcontext.sc_err = regs->tf_err;
	frame.sf_sc.uc_mcontext.sc_cr2 = (u_int32_t)(uintptr_t)ksi->ksi_addr;
	frame.sf_sc.uc_mcontext.sc_trapno = bsd_to_linux_trapcode(code);

	#ifdef DEBUG
	if (ldebug(rt_sendsig))
	printf(LMSG("rt_sendsig flags: 0x%x, sp: %p, ss: 0x%lx, mask: 0x%x"),
	frame.sf_sc.uc_stack.ss_flags, td->td_sigstk.ss_sp,
	td->td_sigstk.ss_size, frame.sf_sc.uc_mcontext.sc_mask);
	#endif

	if (copyout(&frame, fp, sizeof(frame)) != 0) {
	/*
	* Process has trashed its stack; give it an illegal
	* instruction to halt it in its tracks.
	*/
	#ifdef DEBUG
	if (ldebug(rt_sendsig))
	printf(LMSG("rt_sendsig: bad stack %p, oonstack=%x"),
	fp, oonstack);
	#endif
	PROC_LOCK(p);
	sigexit(td, SIGILL);
	}

	/*
	* Build context to run handler in.
	*/
	regs->tf_rsp = PTROUT(fp);
	regs->tf_rip = LINUX32_PS_STRINGS - *(p->p_sysent->sv_szsigcode) +
	linux_sznonrtsigcode;
	regs->tf_rflags &= ~(PSL_T \| PSL_D);
	regs->tf_cs = _ucode32sel;
	regs->tf_ss = _udatasel;
	regs->tf_ds = _udatasel;
	regs->tf_es = _udatasel;
	regs->tf_fs = _ufssel;
	regs->tf_gs = _ugssel;
	regs->tf_flags = TF_HASSEGS;
	set_pcb_flags(td->td_pcb, PCB_FULL_IRET);
	PROC_LOCK(p);
	mtx_lock(&psp->ps_mtx);
	}


	/*
	* Send an interrupt to process.
	*
	* Stack is set up to allow sigcode stored
	* in u. to call routine, followed by kcall
	* to sigreturn routine below. After sigreturn
	* resets the signal mask, the stack, and the
	* frame pointer, it returns to the user
	* specified pc, psl.
	*/
	static void
	linux_sendsig(sig_t catcher, ksiginfo_t ksi, sigset_t mask)
	{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct sigacts *psp;
	struct trapframe *regs;
	struct l_sigframe *fp, frame;
	l_sigset_t lmask;
	int oonstack, i;
	int sig, code;

	sig = ksi->ksi_signo;
	code = ksi->ksi_code;
	PROC_LOCK_ASSERT(p, MA_OWNED);
	psp = p->p_sigacts;
	mtx_assert(&psp->ps_mtx, MA_OWNED);
	if (SIGISMEMBER(psp->ps_siginfo, sig)) {
	/* Signal handler installed with SA_SIGINFO. */
	linux_rt_sendsig(catcher, ksi, mask);
	return;
	}

	regs = td->td_frame;
	oonstack = sigonstack(regs->tf_rsp);

	#ifdef DEBUG
	if (ldebug(sendsig))
	printf(ARGS(sendsig, "%p, %d, %p, %u"),
	catcher, sig, (void*)mask, code);
	#endif

	/*
	* Allocate space for the signal handler context.
	*/
	if ((td->td_pflags & TDP_ALTSTACK) && !oonstack &&
	SIGISMEMBER(psp->ps_sigonstack, sig)) {
	fp = (struct l_sigframe *)(td->td_sigstk.ss_sp +
	td->td_sigstk.ss_size - sizeof(struct l_sigframe));
	} else
	fp = (struct l_sigframe *)regs->tf_rsp - 1;
	mtx_unlock(&psp->ps_mtx);
	PROC_UNLOCK(p);

	/*
	* Build the argument list for the signal handler.
	*/
	if (p->p_sysent->sv_sigtbl)
	if (sig <= p->p_sysent->sv_sigsize)
	sig = p->p_sysent->sv_sigtbl[_SIG_IDX(sig)];

	bzero(&frame, sizeof(frame));

	frame.sf_handler = PTROUT(catcher);
	frame.sf_sig = sig;

	bsd_to_linux_sigset(mask, &lmask);

	/*
	* Build the signal context to be used by sigreturn.
	*/
	frame.sf_sc.sc_mask = lmask.__bits[0];
	frame.sf_sc.sc_gs = regs->tf_gs;
	frame.sf_sc.sc_fs = regs->tf_fs;
	frame.sf_sc.sc_es = regs->tf_es;
	frame.sf_sc.sc_ds = regs->tf_ds;
	frame.sf_sc.sc_edi = regs->tf_rdi;
	frame.sf_sc.sc_esi = regs->tf_rsi;
	frame.sf_sc.sc_ebp = regs->tf_rbp;
	frame.sf_sc.sc_ebx = regs->tf_rbx;
	frame.sf_sc.sc_edx = regs->tf_rdx;
	frame.sf_sc.sc_ecx = regs->tf_rcx;
	frame.sf_sc.sc_eax = regs->tf_rax;
	frame.sf_sc.sc_eip = regs->tf_rip;
	frame.sf_sc.sc_cs = regs->tf_cs;
	frame.sf_sc.sc_eflags = regs->tf_rflags;
	frame.sf_sc.sc_esp_at_signal = regs->tf_rsp;
	frame.sf_sc.sc_ss = regs->tf_ss;
	frame.sf_sc.sc_err = regs->tf_err;
	frame.sf_sc.sc_cr2 = (u_int32_t)(uintptr_t)ksi->ksi_addr;
	frame.sf_sc.sc_trapno = bsd_to_linux_trapcode(code);

	for (i = 0; i < (LINUX_NSIG_WORDS-1); i++)
	frame.sf_extramask[i] = lmask.__bits[i+1];

	if (copyout(&frame, fp, sizeof(frame)) != 0) {
	/*
	* Process has trashed its stack; give it an illegal
	* instruction to halt it in its tracks.
	*/
	PROC_LOCK(p);
	sigexit(td, SIGILL);
	}

	/*
	* Build context to run handler in.
	*/
	regs->tf_rsp = PTROUT(fp);
	regs->tf_rip = LINUX32_PS_STRINGS - *(p->p_sysent->sv_szsigcode);
	regs->tf_rflags &= ~(PSL_T \| PSL_D);
	regs->tf_cs = _ucode32sel;
	regs->tf_ss = _udatasel;
	regs->tf_ds = _udatasel;
	regs->tf_es = _udatasel;
	regs->tf_fs = _ufssel;
	regs->tf_gs = _ugssel;
	regs->tf_flags = TF_HASSEGS;
	set_pcb_flags(td->td_pcb, PCB_FULL_IRET);
	PROC_LOCK(p);
	mtx_lock(&psp->ps_mtx);
	}

	/*
	* System call to cleanup state after a signal
	* has been taken. Reset signal mask and
	* stack state from context left by sendsig (above).
	* Return to previous pc and psl as specified by
	* context left by sendsig. Check carefully to
	* make sure that the user has not modified the
	* psl to gain improper privileges or to cause
	* a machine fault.
	*/
	int
	linux_sigreturn(struct thread td, struct linux_sigreturn_args args)
	{
	struct l_sigframe frame;
	struct trapframe *regs;
	sigset_t bmask;
	l_sigset_t lmask;
	int eflags, i;
	ksiginfo_t ksi;

	regs = td->td_frame;

	#ifdef DEBUG
	if (ldebug(sigreturn))
	printf(ARGS(sigreturn, "%p"), (void *)args->sfp);
	#endif
	/*
	* The trampoline code hands us the sigframe.
	* It is unsafe to keep track of it ourselves, in the event that a
	* program jumps out of a signal handler.
	*/
	if (copyin(args->sfp, &frame, sizeof(frame)) != 0)
	return (EFAULT);

	/*
	* Check for security violations.
	*/
	#define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0)
	eflags = frame.sf_sc.sc_eflags;
	/*
	* XXX do allow users to change the privileged flag PSL_RF. The
	* cpu sets PSL_RF in tf_eflags for faults. Debuggers should
	* sometimes set it there too. tf_eflags is kept in the signal
	* context during signal handling and there is no other place
	* to remember it, so the PSL_RF bit may be corrupted by the
	* signal handler without us knowing. Corruption of the PSL_RF
	* bit at worst causes one more or one less debugger trap, so
	* allowing it is fairly harmless.
	*/
	if (!EFLAGS_SECURE(eflags & ~PSL_RF, regs->tf_rflags & ~PSL_RF))
	return(EINVAL);

	/*
	* Don't allow users to load a valid privileged %cs. Let the
	* hardware check for invalid selectors, excess privilege in
	* other selectors, invalid %eip's and invalid %esp's.
	*/
	#define CS_SECURE(cs) (ISPL(cs) == SEL_UPL)
	if (!CS_SECURE(frame.sf_sc.sc_cs)) {
	ksiginfo_init_trap(&ksi);
	ksi.ksi_signo = SIGBUS;
	ksi.ksi_code = BUS_OBJERR;
	ksi.ksi_trapno = T_PROTFLT;
	ksi.ksi_addr = (void *)regs->tf_rip;
	trapsignal(td, &ksi);
	return(EINVAL);
	}

	lmask.__bits[0] = frame.sf_sc.sc_mask;
	for (i = 0; i < (LINUX_NSIG_WORDS-1); i++)
	lmask.__bits[i+1] = frame.sf_extramask[i];
	linux_to_bsd_sigset(&lmask, &bmask);
	kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0);

	/*
	* Restore signal context.
	*/
	regs->tf_rdi = frame.sf_sc.sc_edi;
	regs->tf_rsi = frame.sf_sc.sc_esi;
	regs->tf_rbp = frame.sf_sc.sc_ebp;
	regs->tf_rbx = frame.sf_sc.sc_ebx;
	regs->tf_rdx = frame.sf_sc.sc_edx;
	regs->tf_rcx = frame.sf_sc.sc_ecx;
	regs->tf_rax = frame.sf_sc.sc_eax;
	regs->tf_rip = frame.sf_sc.sc_eip;
	regs->tf_cs = frame.sf_sc.sc_cs;
	regs->tf_ds = frame.sf_sc.sc_ds;
	regs->tf_es = frame.sf_sc.sc_es;
	regs->tf_fs = frame.sf_sc.sc_fs;
	regs->tf_gs = frame.sf_sc.sc_gs;
	regs->tf_rflags = eflags;
	regs->tf_rsp = frame.sf_sc.sc_esp_at_signal;
	regs->tf_ss = frame.sf_sc.sc_ss;
	set_pcb_flags(td->td_pcb, PCB_FULL_IRET);

	return (EJUSTRETURN);
	}

	/*
	* System call to cleanup state after a signal
	* has been taken. Reset signal mask and
	* stack state from context left by rt_sendsig (above).
	* Return to previous pc and psl as specified by
	* context left by sendsig. Check carefully to
	* make sure that the user has not modified the
	* psl to gain improper privileges or to cause
	* a machine fault.
	*/
	int
	linux_rt_sigreturn(struct thread td, struct linux_rt_sigreturn_args args)
	{
	struct l_ucontext uc;
	struct l_sigcontext *context;
	sigset_t bmask;
	l_stack_t *lss;
	stack_t ss;
	struct trapframe *regs;
	int eflags;
	ksiginfo_t ksi;

	regs = td->td_frame;

	#ifdef DEBUG
	if (ldebug(rt_sigreturn))
	printf(ARGS(rt_sigreturn, "%p"), (void *)args->ucp);
	#endif
	/*
	* The trampoline code hands us the ucontext.
	* It is unsafe to keep track of it ourselves, in the event that a
	* program jumps out of a signal handler.
	*/
	if (copyin(args->ucp, &uc, sizeof(uc)) != 0)
	return (EFAULT);

	context = &uc.uc_mcontext;

	/*
	* Check for security violations.
	*/
	#define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0)
	eflags = context->sc_eflags;
	/*
	* XXX do allow users to change the privileged flag PSL_RF. The
	* cpu sets PSL_RF in tf_eflags for faults. Debuggers should
	* sometimes set it there too. tf_eflags is kept in the signal
	* context during signal handling and there is no other place
	* to remember it, so the PSL_RF bit may be corrupted by the
	* signal handler without us knowing. Corruption of the PSL_RF
	* bit at worst causes one more or one less debugger trap, so
	* allowing it is fairly harmless.
	*/
	if (!EFLAGS_SECURE(eflags & ~PSL_RF, regs->tf_rflags & ~PSL_RF))
	return(EINVAL);

	/*
	* Don't allow users to load a valid privileged %cs. Let the
	* hardware check for invalid selectors, excess privilege in
	* other selectors, invalid %eip's and invalid %esp's.
	*/
	#define CS_SECURE(cs) (ISPL(cs) == SEL_UPL)
	if (!CS_SECURE(context->sc_cs)) {
	ksiginfo_init_trap(&ksi);
	ksi.ksi_signo = SIGBUS;
	ksi.ksi_code = BUS_OBJERR;
	ksi.ksi_trapno = T_PROTFLT;
	ksi.ksi_addr = (void *)regs->tf_rip;
	trapsignal(td, &ksi);
	return(EINVAL);
	}

	linux_to_bsd_sigset(&uc.uc_sigmask, &bmask);
	kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0);

	/*
	* Restore signal context
	*/
	regs->tf_gs = context->sc_gs;
	regs->tf_fs = context->sc_fs;
	regs->tf_es = context->sc_es;
	regs->tf_ds = context->sc_ds;
	regs->tf_rdi = context->sc_edi;
	regs->tf_rsi = context->sc_esi;
	regs->tf_rbp = context->sc_ebp;
	regs->tf_rbx = context->sc_ebx;
	regs->tf_rdx = context->sc_edx;
	regs->tf_rcx = context->sc_ecx;
	regs->tf_rax = context->sc_eax;
	regs->tf_rip = context->sc_eip;
	regs->tf_cs = context->sc_cs;
	regs->tf_rflags = eflags;
	regs->tf_rsp = context->sc_esp_at_signal;
	regs->tf_ss = context->sc_ss;
	set_pcb_flags(td->td_pcb, PCB_FULL_IRET);

	/*
	* call sigaltstack & ignore results..
	*/
	lss = &uc.uc_stack;
	ss.ss_sp = PTRIN(lss->ss_sp);
	ss.ss_size = lss->ss_size;
	ss.ss_flags = linux_to_bsd_sigaltstack(lss->ss_flags);

	#ifdef DEBUG
	if (ldebug(rt_sigreturn))
	printf(LMSG("rt_sigret flags: 0x%x, sp: %p, ss: 0x%lx, mask: 0x%x"),
	ss.ss_flags, ss.ss_sp, ss.ss_size, context->sc_mask);
	#endif
	(void)kern_sigaltstack(td, &ss, NULL);

	return (EJUSTRETURN);
	}

	static int
	linux32_fetch_syscall_args(struct thread td, struct syscall_args sa)
	{
	struct proc *p;
	struct trapframe *frame;

	p = td->td_proc;
	frame = td->td_frame;

	sa->args[0] = frame->tf_rbx;
	sa->args[1] = frame->tf_rcx;
	sa->args[2] = frame->tf_rdx;
	sa->args[3] = frame->tf_rsi;
	sa->args[4] = frame->tf_rdi;
	sa->args[5] = frame->tf_rbp; /* Unconfirmed */
	sa->code = frame->tf_rax;

	if (sa->code >= p->p_sysent->sv_size)
	sa->callp = &p->p_sysent->sv_table[0];
	else
	sa->callp = &p->p_sysent->sv_table[sa->code];
	sa->narg = sa->callp->sy_narg;

	td->td_retval[0] = 0;
	td->td_retval[1] = frame->tf_rdx;

	return (0);
	}

	/*
	* If a linux binary is exec'ing something, try this image activator
	* first. We override standard shell script execution in order to
	* be able to modify the interpreter path. We only do this if a linux
	* binary is doing the exec, so we do not create an EXEC module for it.
	*/
	static int exec_linux_imgact_try(struct image_params *iparams);

	static int
	exec_linux_imgact_try(struct image_params *imgp)
	{
	const char head = (const char )imgp->image_header;
	char *rpath;
	int error = -1, len;

	/*
	* The interpreter for shell scripts run from a linux binary needs
	* to be located in /compat/linux if possible in order to recursively
	* maintain linux path emulation.
	*/
	if (((const short *)head)[0] == SHELLMAGIC) {
	/*
	* Run our normal shell image activator. If it succeeds attempt
	* to use the alternate path for the interpreter. If an
	* alternate * path is found, use our stringspace to store it.
	*/
	if ((error = exec_shell_imgact(imgp)) == 0) {
	linux_emul_convpath(FIRST_THREAD_IN_PROC(imgp->proc),
	imgp->interpreter_name, UIO_SYSSPACE, &rpath, 0,
	AT_FDCWD);
	if (rpath != NULL) {
	len = strlen(rpath) + 1;

	if (len <= MAXSHELLCMDLEN) {
	memcpy(imgp->interpreter_name, rpath,
	len);
	}
	free(rpath, M_TEMP);
	}
	}
	}
	return(error);
	}

	/*
	* Clear registers on exec
	* XXX copied from ia32_signal.c.
	*/
	static void
	exec_linux_setregs(td, entry, stack, ps_strings)
	struct thread *td;
	u_long entry;
	u_long stack;
	u_long ps_strings;
	{
	struct trapframe *regs = td->td_frame;
	struct pcb *pcb = td->td_pcb;

	mtx_lock(&dt_lock);
	if (td->td_proc->p_md.md_ldt != NULL)
	user_ldt_free(td);
	else
	mtx_unlock(&dt_lock);

	critical_enter();
	wrmsr(MSR_FSBASE, 0);
	wrmsr(MSR_KGSBASE, 0); /* User value while we're in the kernel */
	pcb->pcb_fsbase = 0;
	pcb->pcb_gsbase = 0;
	critical_exit();
	pcb->pcb_initial_fpucw = __LINUX_NPXCW__;

	bzero((char *)regs, sizeof(struct trapframe));
	regs->tf_rip = entry;
	regs->tf_rsp = stack;
	regs->tf_rflags = PSL_USER \| (regs->tf_rflags & PSL_T);
	regs->tf_gs = _ugssel;
	regs->tf_fs = _ufssel;
	regs->tf_es = _udatasel;
	regs->tf_ds = _udatasel;
	regs->tf_ss = _udatasel;
	regs->tf_flags = TF_HASSEGS;
	regs->tf_cs = _ucode32sel;
	regs->tf_rbx = ps_strings;

	fpstate_drop(td);

	/* Do full restore on return so that we can change to a different %cs */
	set_pcb_flags(pcb, PCB_32BIT \| PCB_FULL_IRET);
	clear_pcb_flags(pcb, PCB_GS32BIT);
	td->td_retval[1] = 0;
	}

	/*
	* XXX copied from ia32_sysvec.c.
	*/
	static register_t *
	linux_copyout_strings(struct image_params *imgp)
	{
	int argc, envc;
	u_int32_t *vectp;
	char stringp, destp;
	u_int32_t *stack_base;
	struct linux32_ps_strings *arginfo;

	/*
	* Calculate string base and vector table pointers.
	* Also deal with signal trampoline code for this exec type.
	*/
	arginfo = (struct linux32_ps_strings *)LINUX32_PS_STRINGS;
	destp = (caddr_t)arginfo - linux_szsigcode - SPARE_USRSPACE -
	linux_szplatform - roundup((ARG_MAX - imgp->args->stringspace),
	sizeof(char *));

	/*
	* install sigcode
	*/
	copyout(imgp->proc->p_sysent->sv_sigcode,
	((caddr_t)arginfo - linux_szsigcode), linux_szsigcode);

	/*
	* Install LINUX_PLATFORM
	*/
	copyout(linux_platform, ((caddr_t)arginfo - linux_szsigcode -
	linux_szplatform), linux_szplatform);

	/*
	* If we have a valid auxargs ptr, prepare some room
	* on the stack.
	*/
	if (imgp->auxargs) {
	/*
	* 'AT_COUNT*2' is size for the ELF Auxargs data. This is for
	* lower compatibility.
	*/
	imgp->auxarg_size = (imgp->auxarg_size) ? imgp->auxarg_size :
	(LINUX_AT_COUNT * 2);
	/*
	* The '+ 2' is for the null pointers at the end of each of
	* the arg and env vector sets,and imgp->auxarg_size is room
	* for argument of Runtime loader.
	*/
	vectp = (u_int32_t *) (destp - (imgp->args->argc +
	imgp->args->envc + 2 + imgp->auxarg_size) *
	sizeof(u_int32_t));

	} else
	/*
	* The '+ 2' is for the null pointers at the end of each of
	* the arg and env vector sets
	*/
	vectp = (u_int32_t *)(destp - (imgp->args->argc +
	imgp->args->envc + 2) * sizeof(u_int32_t));

	/*
	* vectp also becomes our initial stack base
	*/
	stack_base = vectp;

	stringp = imgp->args->begin_argv;
	argc = imgp->args->argc;
	envc = imgp->args->envc;
	/*
	* Copy out strings - arguments and environment.
	*/
	copyout(stringp, destp, ARG_MAX - imgp->args->stringspace);

	/*
	* Fill in "ps_strings" struct for ps, w, etc.
	*/
	suword32(&arginfo->ps_argvstr, (uint32_t)(intptr_t)vectp);
	suword32(&arginfo->ps_nargvstr, argc);

	/*
	* Fill in argument portion of vector table.
	*/
	for (; argc > 0; --argc) {
	suword32(vectp++, (uint32_t)(intptr_t)destp);
	while (*stringp++ != 0)
	destp++;
	destp++;
	}

	/* a null vector table pointer separates the argp's from the envp's */
	suword32(vectp++, 0);

	suword32(&arginfo->ps_envstr, (uint32_t)(intptr_t)vectp);
	suword32(&arginfo->ps_nenvstr, envc);

	/*
	* Fill in environment portion of vector table.
	*/
	for (; envc > 0; --envc) {
	suword32(vectp++, (uint32_t)(intptr_t)destp);
	while (*stringp++ != 0)
	destp++;
	destp++;
	}

	/* end of vector table is a null pointer */
	suword32(vectp, 0);

	return ((register_t *)stack_base);
	}

	SYSCTL_NODE(_compat, OID_AUTO, linux32, CTLFLAG_RW, 0,
	"32-bit Linux emulation");

	static u_long linux32_maxdsiz = LINUX32_MAXDSIZ;
	SYSCTL_ULONG(_compat_linux32, OID_AUTO, maxdsiz, CTLFLAG_RW,
	&linux32_maxdsiz, 0, "");
	static u_long linux32_maxssiz = LINUX32_MAXSSIZ;
	SYSCTL_ULONG(_compat_linux32, OID_AUTO, maxssiz, CTLFLAG_RW,
	&linux32_maxssiz, 0, "");
	static u_long linux32_maxvmem = LINUX32_MAXVMEM;
	SYSCTL_ULONG(_compat_linux32, OID_AUTO, maxvmem, CTLFLAG_RW,
	&linux32_maxvmem, 0, "");

	static void
	linux32_fixlimit(struct rlimit *rl, int which)
	{

	switch (which) {
	case RLIMIT_DATA:
	if (linux32_maxdsiz != 0) {
	if (rl->rlim_cur > linux32_maxdsiz)
	rl->rlim_cur = linux32_maxdsiz;
	if (rl->rlim_max > linux32_maxdsiz)
	rl->rlim_max = linux32_maxdsiz;
	}
	break;
	case RLIMIT_STACK:
	if (linux32_maxssiz != 0) {
	if (rl->rlim_cur > linux32_maxssiz)
	rl->rlim_cur = linux32_maxssiz;
	if (rl->rlim_max > linux32_maxssiz)
	rl->rlim_max = linux32_maxssiz;
	}
	break;
	case RLIMIT_VMEM:
	if (linux32_maxvmem != 0) {
	if (rl->rlim_cur > linux32_maxvmem)
	rl->rlim_cur = linux32_maxvmem;
	if (rl->rlim_max > linux32_maxvmem)
	rl->rlim_max = linux32_maxvmem;
	}
	break;
	}
	}

	struct sysentvec elf_linux_sysvec = {
	.sv_size = LINUX_SYS_MAXSYSCALL,
	.sv_table = linux_sysent,
	.sv_mask = 0,
	.sv_sigsize = LINUX_SIGTBLSZ,
	.sv_sigtbl = bsd_to_linux_signal,
	.sv_errsize = ELAST + 1,
	.sv_errtbl = bsd_to_linux_errno,
	.sv_transtrap = translate_traps,
	.sv_fixup = elf_linux_fixup,
	.sv_sendsig = linux_sendsig,
	.sv_sigcode = linux_sigcode,
	.sv_szsigcode = &linux_szsigcode,
	.sv_prepsyscall = NULL,
	.sv_name = "Linux ELF32",
	.sv_coredump = elf32_coredump,
	.sv_imgact_try = exec_linux_imgact_try,
	.sv_minsigstksz = LINUX_MINSIGSTKSZ,
	.sv_pagesize = PAGE_SIZE,
	.sv_minuser = VM_MIN_ADDRESS,
	.sv_maxuser = LINUX32_USRSTACK,
	.sv_usrstack = LINUX32_USRSTACK,
	.sv_psstrings = LINUX32_PS_STRINGS,
	.sv_stackprot = VM_PROT_ALL,
	.sv_copyout_strings = linux_copyout_strings,
	.sv_setregs = exec_linux_setregs,
	.sv_fixlimit = linux32_fixlimit,
	.sv_maxssiz = &linux32_maxssiz,
	.sv_flags = SV_ABI_LINUX \| SV_ILP32 \| SV_IA32,
	.sv_set_syscall_retval = cpu_set_syscall_retval,
	.sv_fetch_syscall_args = linux32_fetch_syscall_args,
	.sv_syscallnames = NULL,
	};

	static char GNU_ABI_VENDOR[] = "GNU";
	static int GNULINUX_ABI_DESC = 0;

	static boolean_t
	linux32_trans_osrel(const Elf_Note note, int32_t osrel)
	{
	const Elf32_Word *desc;
	uintptr_t p;

	p = (uintptr_t)(note + 1);
	p += roundup2(note->n_namesz, sizeof(Elf32_Addr));

	desc = (const Elf32_Word *)p;
	if (desc[0] != GNULINUX_ABI_DESC)
	return (FALSE);

	/*
	* For linux we encode osrel as follows (see linux_mib.c):
	* VVVMMMIII (version, major, minor), see linux_mib.c.
	*/
	osrel = desc[1] 1000000 + desc[2] * 1000 + desc[3];

	return (TRUE);
	}

	static Elf_Brandnote linux32_brandnote = {
	.hdr.n_namesz = sizeof(GNU_ABI_VENDOR),
	.hdr.n_descsz = 16, /* XXX at least 16 */
	.hdr.n_type = 1,
	.vendor = GNU_ABI_VENDOR,
	.flags = BN_TRANSLATE_OSREL,
	.trans_osrel = linux32_trans_osrel
	};

	static Elf32_Brandinfo linux_brand = {
	.brand = ELFOSABI_LINUX,
	.machine = EM_386,
	.compat_3_brand = "Linux",
	.emul_path = "/compat/linux",
	.interp_path = "/lib/ld-linux.so.1",
	.sysvec = &elf_linux_sysvec,
	.interp_newpath = NULL,
	.brand_note = &linux32_brandnote,
	.flags = BI_CAN_EXEC_DYN \| BI_BRAND_NOTE
	};

	static Elf32_Brandinfo linux_glibc2brand = {
	.brand = ELFOSABI_LINUX,
	.machine = EM_386,
	.compat_3_brand = "Linux",
	.emul_path = "/compat/linux",
	.interp_path = "/lib/ld-linux.so.2",
	.sysvec = &elf_linux_sysvec,
	.interp_newpath = NULL,
	.brand_note = &linux32_brandnote,
	.flags = BI_CAN_EXEC_DYN \| BI_BRAND_NOTE
	};

	Elf32_Brandinfo *linux_brandlist[] = {
	&linux_brand,
	&linux_glibc2brand,
	NULL
	};

	static int
	linux_elf_modevent(module_t mod, int type, void *data)
	{
	Elf32_Brandinfo **brandinfo;
	int error;
	struct linux_ioctl_handler **lihp;
	struct linux_device_handler **ldhp;

	error = 0;

	switch(type) {
	case MOD_LOAD:
	for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
	++brandinfo)
	if (elf32_insert_brand_entry(*brandinfo) < 0)
	error = EINVAL;
	if (error == 0) {
	SET_FOREACH(lihp, linux_ioctl_handler_set)
	linux_ioctl_register_handler(*lihp);
	SET_FOREACH(ldhp, linux_device_handler_set)
	linux_device_register_handler(*ldhp);
	mtx_init(&emul_lock, "emuldata lock", NULL, MTX_DEF);
	sx_init(&emul_shared_lock, "emuldata->shared lock");
	LIST_INIT(&futex_list);
	mtx_init(&futex_mtx, "ftllk", NULL, MTX_DEF);
	linux_exit_tag = EVENTHANDLER_REGISTER(process_exit,
	linux_proc_exit, NULL, 1000);
	linux_schedtail_tag = EVENTHANDLER_REGISTER(schedtail,
	linux_schedtail, NULL, 1000);
	linux_exec_tag = EVENTHANDLER_REGISTER(process_exec,
	linux_proc_exec, NULL, 1000);
	linux_szplatform = roundup(strlen(linux_platform) + 1,
	sizeof(char *));
	linux_osd_jail_register();
	stclohz = (stathz ? stathz : hz);
	if (bootverbose)
	printf("Linux ELF exec handler installed\n");
	} else
	printf("cannot insert Linux ELF brand handler\n");
	break;
	case MOD_UNLOAD:
	for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
	++brandinfo)
	if (elf32_brand_inuse(*brandinfo))
	error = EBUSY;
	if (error == 0) {
	for (brandinfo = &linux_brandlist[0];
	*brandinfo != NULL; ++brandinfo)
	if (elf32_remove_brand_entry(*brandinfo) < 0)
	error = EINVAL;
	}
	if (error == 0) {
	SET_FOREACH(lihp, linux_ioctl_handler_set)
	linux_ioctl_unregister_handler(*lihp);
	SET_FOREACH(ldhp, linux_device_handler_set)
	linux_device_unregister_handler(*ldhp);
	mtx_destroy(&emul_lock);
	sx_destroy(&emul_shared_lock);
	mtx_destroy(&futex_mtx);
	EVENTHANDLER_DEREGISTER(process_exit, linux_exit_tag);
	EVENTHANDLER_DEREGISTER(schedtail, linux_schedtail_tag);
	EVENTHANDLER_DEREGISTER(process_exec, linux_exec_tag);
	linux_osd_jail_deregister();
	if (bootverbose)
	printf("Linux ELF exec handler removed\n");
	} else
	printf("Could not deinstall ELF interpreter entry\n");
	break;
	default:
	return EOPNOTSUPP;
	}
	return error;
	}

	static moduledata_t linux_elf_mod = {
	"linuxelf",
	linux_elf_modevent,
	0
	};

	DECLARE_MODULE_TIED(linuxelf, linux_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY);
	Index: stable/8/sys/cddl/contrib/opensolaris
	===================================================================
	--- stable/8/sys/cddl/contrib/opensolaris (revision 219197)
	+++ stable/8/sys/cddl/contrib/opensolaris (revision 219198)

	Property changes on: stable/8/sys/cddl/contrib/opensolaris
	___________________________________________________________________
	Modified: svn:mergeinfo
	## -0,0 +0,1 ##
	Merged /head/sys/cddl/contrib/opensolaris:r218658,218668
	Index: stable/8/sys/compat/linux/linux_futex.c
	===================================================================
	--- stable/8/sys/compat/linux/linux_futex.c (revision 219197)
	+++ stable/8/sys/compat/linux/linux_futex.c (revision 219198)
	@@ -1,886 +1,886 @@
	/* $NetBSD: linux_futex.c,v 1.7 2006/07/24 19:01:49 manu Exp $ */

	/*-
	* Copyright (c) 2005 Emmanuel Dreyfus, all rights reserved.
	*
	* Redistribution and use in source and binary forms, with or without
	* modification, are permitted provided that the following conditions
	* are met:
	* 1. Redistributions of source code must retain the above copyright
	* notice, this list of conditions and the following disclaimer.
	* 2. Redistributions in binary form must reproduce the above copyright
	* notice, this list of conditions and the following disclaimer in the
	* documentation and/or other materials provided with the distribution.
	* 3. All advertising materials mentioning features or use of this software
	* must display the following acknowledgement:
	* This product includes software developed by Emmanuel Dreyfus
	* 4. The name of the author may not be used to endorse or promote
	* products derived from this software without specific prior written
	* permission.
	*
	* THIS SOFTWARE IS PROVIDED BY THE THE AUTHOR AND CONTRIBUTORS ``AS IS''
	* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
	* THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
	* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS
	* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
	* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
	* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
	* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
	* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
	* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
	* POSSIBILITY OF SUCH DAMAGE.
	*/

	#include <sys/cdefs.h>
	__FBSDID("$FreeBSD$");
	#if 0
	__KERNEL_RCSID(1, "$NetBSD: linux_futex.c,v 1.7 2006/07/24 19:01:49 manu Exp $");
	#endif

	#include "opt_compat.h"

	#include <sys/param.h>
	#include <sys/systm.h>
	#include <sys/imgact.h>
	#include <sys/kernel.h>
	#include <sys/ktr.h>
	#include <sys/lock.h>
	#include <sys/malloc.h>
	#include <sys/mutex.h>
	#include <sys/priv.h>
	#include <sys/proc.h>
	#include <sys/queue.h>
	#include <sys/sched.h>
	#include <sys/sx.h>

	#ifdef COMPAT_LINUX32
	#include <machine/../linux32/linux.h>
	#include <machine/../linux32/linux32_proto.h>
	#else
	#include <machine/../linux/linux.h>
	#include <machine/../linux/linux_proto.h>
	#endif
	-#include <compat/linux/linux_futex.h>
	#include <compat/linux/linux_emul.h>
	+#include <compat/linux/linux_futex.h>
	#include <compat/linux/linux_util.h>

	MALLOC_DEFINE(M_FUTEX, "futex", "Linux futexes");
	MALLOC_DEFINE(M_FUTEX_WP, "futex wp", "Linux futexes wp");

	struct futex;

	struct waiting_proc {
	uint32_t wp_flags;
	struct futex *wp_futex;
	TAILQ_ENTRY(waiting_proc) wp_list;
	};

	struct futex {
	struct sx f_lck;
	uint32_t *f_uaddr;
	uint32_t f_refcount;
	uint32_t f_bitset;
	LIST_ENTRY(futex) f_list;
	TAILQ_HEAD(lf_waiting_proc, waiting_proc) f_waiting_proc;
	};

	struct futex_list futex_list;

	#define FUTEX_LOCK(f) sx_xlock(&(f)->f_lck)
	#define FUTEX_UNLOCK(f) sx_xunlock(&(f)->f_lck)
	#define FUTEX_INIT(f) sx_init_flags(&(f)->f_lck, "ftlk", SX_DUPOK)
	#define FUTEX_DESTROY(f) sx_destroy(&(f)->f_lck)
	#define FUTEX_ASSERT_LOCKED(f) sx_assert(&(f)->f_lck, SA_XLOCKED)

	struct mtx futex_mtx; /* protects the futex list */
	#define FUTEXES_LOCK mtx_lock(&futex_mtx)
	#define FUTEXES_UNLOCK mtx_unlock(&futex_mtx)

	/* flags for futex_get() */
	#define FUTEX_CREATE_WP 0x1 /* create waiting_proc */
	#define FUTEX_DONTCREATE 0x2 /* don't create futex if not exists */
	#define FUTEX_DONTEXISTS 0x4 /* return EINVAL if futex exists */

	/* wp_flags */
	#define FUTEX_WP_REQUEUED 0x1 /* wp requeued - wp moved from wp_list
	* of futex where thread sleep to wp_list
	* of another futex.
	*/
	#define FUTEX_WP_REMOVED 0x2 /* wp is woken up and removed from futex
	* wp_list to prevent double wakeup.
	*/

	/* support.s */
	int futex_xchgl(int oparg, uint32_t uaddr, int oldval);
	int futex_addl(int oparg, uint32_t uaddr, int oldval);
	int futex_orl(int oparg, uint32_t uaddr, int oldval);
	int futex_andl(int oparg, uint32_t uaddr, int oldval);
	int futex_xorl(int oparg, uint32_t uaddr, int oldval);

	static void
	futex_put(struct futex f, struct waiting_proc wp)
	{

	FUTEX_ASSERT_LOCKED(f);
	if (wp != NULL) {
	if ((wp->wp_flags & FUTEX_WP_REMOVED) == 0)
	TAILQ_REMOVE(&f->f_waiting_proc, wp, wp_list);
	free(wp, M_FUTEX_WP);
	}

	FUTEXES_LOCK;
	if (--f->f_refcount == 0) {
	LIST_REMOVE(f, f_list);
	FUTEXES_UNLOCK;
	FUTEX_UNLOCK(f);

	LINUX_CTR2(sys_futex, "futex_put destroy uaddr %p ref %d",
	f->f_uaddr, f->f_refcount);
	FUTEX_DESTROY(f);
	free(f, M_FUTEX);
	return;
	}

	LINUX_CTR2(sys_futex, "futex_put uaddr %p ref %d",
	f->f_uaddr, f->f_refcount);
	FUTEXES_UNLOCK;
	FUTEX_UNLOCK(f);
	}

	static int
	futex_get0(uint32_t uaddr, struct futex *newf, uint32_t flags)
	{
	struct futex f, tmpf;

	*newf = tmpf = NULL;

	retry:
	FUTEXES_LOCK;
	LIST_FOREACH(f, &futex_list, f_list) {
	if (f->f_uaddr == uaddr) {
	if (tmpf != NULL) {
	FUTEX_UNLOCK(tmpf);
	FUTEX_DESTROY(tmpf);
	free(tmpf, M_FUTEX);
	}
	if (flags & FUTEX_DONTEXISTS) {
	FUTEXES_UNLOCK;
	return (EINVAL);
	}

	/*
	* Increment refcount of the found futex to
	* prevent it from deallocation before FUTEX_LOCK()
	*/
	++f->f_refcount;
	FUTEXES_UNLOCK;

	FUTEX_LOCK(f);
	*newf = f;
	LINUX_CTR2(sys_futex, "futex_get uaddr %p ref %d",
	uaddr, f->f_refcount);
	return (0);
	}
	}

	if (flags & FUTEX_DONTCREATE) {
	FUTEXES_UNLOCK;
	LINUX_CTR1(sys_futex, "futex_get uaddr %p null", uaddr);
	return (0);
	}

	if (tmpf == NULL) {
	FUTEXES_UNLOCK;
	tmpf = malloc(sizeof(*tmpf), M_FUTEX, M_WAITOK \| M_ZERO);
	tmpf->f_uaddr = uaddr;
	tmpf->f_refcount = 1;
	tmpf->f_bitset = FUTEX_BITSET_MATCH_ANY;
	FUTEX_INIT(tmpf);
	TAILQ_INIT(&tmpf->f_waiting_proc);

	/*
	* Lock the new futex before an insert into the futex_list
	* to prevent futex usage by other.
	*/
	FUTEX_LOCK(tmpf);
	goto retry;
	}

	LIST_INSERT_HEAD(&futex_list, tmpf, f_list);
	FUTEXES_UNLOCK;

	LINUX_CTR2(sys_futex, "futex_get uaddr %p ref %d new",
	uaddr, tmpf->f_refcount);
	*newf = tmpf;
	return (0);
	}

	static int
	futex_get(uint32_t uaddr, struct waiting_proc wp, struct futex *f,
	uint32_t flags)
	{
	int error;

	if (flags & FUTEX_CREATE_WP) {
	*wp = malloc(sizeof(struct waiting_proc), M_FUTEX_WP, M_WAITOK);
	(*wp)->wp_flags = 0;
	}
	error = futex_get0(uaddr, f, flags);
	if (error) {
	if (flags & FUTEX_CREATE_WP)
	free(*wp, M_FUTEX_WP);
	return (error);
	}
	if (flags & FUTEX_CREATE_WP) {
	TAILQ_INSERT_HEAD(&(f)->f_waiting_proc, wp, wp_list);
	(wp)->wp_futex = f;
	}

	return (error);
	}

	static int
	futex_sleep(struct futex f, struct waiting_proc wp, unsigned long timeout)
	{
	int error;

	FUTEX_ASSERT_LOCKED(f);
	LINUX_CTR4(sys_futex, "futex_sleep enter uaddr %p wp %p timo %ld ref %d",
	f->f_uaddr, wp, timeout, f->f_refcount);
	error = sx_sleep(wp, &f->f_lck, PCATCH, "futex", timeout);
	if (wp->wp_flags & FUTEX_WP_REQUEUED) {
	KASSERT(f != wp->wp_futex, ("futex != wp_futex"));
	LINUX_CTR5(sys_futex, "futex_sleep out error %d uaddr %p w"
	" %p requeued uaddr %p ref %d",
	error, f->f_uaddr, wp, wp->wp_futex->f_uaddr,
	wp->wp_futex->f_refcount);
	futex_put(f, NULL);
	f = wp->wp_futex;
	FUTEX_LOCK(f);
	} else
	LINUX_CTR3(sys_futex, "futex_sleep out error %d uaddr %p wp %p",
	error, f->f_uaddr, wp);

	futex_put(f, wp);
	return (error);
	}

	static int
	futex_wake(struct futex *f, int n, uint32_t bitset)
	{
	struct waiting_proc wp, wpt;
	int count = 0;

	if (bitset == 0)
	return (EINVAL);

	FUTEX_ASSERT_LOCKED(f);
	TAILQ_FOREACH_SAFE(wp, &f->f_waiting_proc, wp_list, wpt) {
	LINUX_CTR3(sys_futex, "futex_wake uaddr %p wp %p ref %d",
	f->f_uaddr, wp, f->f_refcount);
	/*
	* Unless we find a matching bit in
	* the bitset, continue searching.
	*/
	if (!(wp->wp_futex->f_bitset & bitset))
	continue;

	wp->wp_flags \|= FUTEX_WP_REMOVED;
	TAILQ_REMOVE(&f->f_waiting_proc, wp, wp_list);
	wakeup_one(wp);
	if (++count == n)
	break;
	}

	return (count);
	}

	static int
	futex_requeue(struct futex f, int n, struct futex f2, int n2)
	{
	struct waiting_proc wp, wpt;
	int count = 0;

	FUTEX_ASSERT_LOCKED(f);
	FUTEX_ASSERT_LOCKED(f2);

	TAILQ_FOREACH_SAFE(wp, &f->f_waiting_proc, wp_list, wpt) {
	if (++count <= n) {
	LINUX_CTR2(sys_futex, "futex_req_wake uaddr %p wp %p",
	f->f_uaddr, wp);
	wp->wp_flags \|= FUTEX_WP_REMOVED;
	TAILQ_REMOVE(&f->f_waiting_proc, wp, wp_list);
	wakeup_one(wp);
	} else {
	LINUX_CTR3(sys_futex, "futex_requeue uaddr %p wp %p to %p",
	f->f_uaddr, wp, f2->f_uaddr);
	wp->wp_flags \|= FUTEX_WP_REQUEUED;
	/* Move wp to wp_list of f2 futex */
	TAILQ_REMOVE(&f->f_waiting_proc, wp, wp_list);
	TAILQ_INSERT_HEAD(&f2->f_waiting_proc, wp, wp_list);

	/*
	* Thread which sleeps on wp after waking should
	* acquire f2 lock, so increment refcount of f2 to
	* prevent it from premature deallocation.
	*/
	wp->wp_futex = f2;
	FUTEXES_LOCK;
	++f2->f_refcount;
	FUTEXES_UNLOCK;
	if (count - n >= n2)
	break;
	}
	}

	return (count);
	}

	static int
	futex_wait(struct futex f, struct waiting_proc wp, struct l_timespec *ts,
	uint32_t bitset)
	{
	struct l_timespec timeout = {0, 0};
	struct timeval tv = {0, 0};
	int timeout_hz;
	int error;

	if (bitset == 0)
	return (EINVAL);
	f->f_bitset = bitset;

	if (ts != NULL) {
	error = copyin(ts, &timeout, sizeof(timeout));
	if (error)
	return (error);
	}

	tv.tv_usec = timeout.tv_sec * 1000000 + timeout.tv_nsec / 1000;
	timeout_hz = tvtohz(&tv);

	if (timeout.tv_sec == 0 && timeout.tv_nsec == 0)
	timeout_hz = 0;

	/*
	* If the user process requests a non null timeout,
	* make sure we do not turn it into an infinite
	* timeout because timeout_hz gets null.
	*
	* We use a minimal timeout of 1/hz. Maybe it would
	* make sense to just return ETIMEDOUT without sleeping.
	*/
	if (((timeout.tv_sec != 0) \|\| (timeout.tv_nsec != 0)) &&
	(timeout_hz == 0))
	timeout_hz = 1;

	error = futex_sleep(f, wp, timeout_hz);
	if (error == EWOULDBLOCK)
	error = ETIMEDOUT;

	return (error);
	}

	static int
	futex_atomic_op(struct thread td, int encoded_op, uint32_t uaddr)
	{
	int op = (encoded_op >> 28) & 7;
	int cmp = (encoded_op >> 24) & 15;
	int oparg = (encoded_op << 8) >> 20;
	int cmparg = (encoded_op << 20) >> 20;
	int oldval = 0, ret;

	if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
	oparg = 1 << oparg;

	#ifdef DEBUG
	if (ldebug(sys_futex))
	printf("futex_atomic_op: op = %d, cmp = %d, oparg = %x, "
	"cmparg = %x, uaddr = %p\n",
	op, cmp, oparg, cmparg, uaddr);
	#endif
	/* XXX: linux verifies access here and returns EFAULT */

	switch (op) {
	case FUTEX_OP_SET:
	ret = futex_xchgl(oparg, uaddr, &oldval);
	break;
	case FUTEX_OP_ADD:
	ret = futex_addl(oparg, uaddr, &oldval);
	break;
	case FUTEX_OP_OR:
	ret = futex_orl(oparg, uaddr, &oldval);
	break;
	case FUTEX_OP_ANDN:
	ret = futex_andl(~oparg, uaddr, &oldval);
	break;
	case FUTEX_OP_XOR:
	ret = futex_xorl(oparg, uaddr, &oldval);
	break;
	default:
	ret = -ENOSYS;
	break;
	}

	if (ret)
	return (ret);

	switch (cmp) {
	case FUTEX_OP_CMP_EQ:
	return (oldval == cmparg);
	case FUTEX_OP_CMP_NE:
	return (oldval != cmparg);
	case FUTEX_OP_CMP_LT:
	return (oldval < cmparg);
	case FUTEX_OP_CMP_GE:
	return (oldval >= cmparg);
	case FUTEX_OP_CMP_LE:
	return (oldval <= cmparg);
	case FUTEX_OP_CMP_GT:
	return (oldval > cmparg);
	default:
	return (-ENOSYS);
	}
	}

	int
	linux_sys_futex(struct thread td, struct linux_sys_futex_args args)
	{
	int clockrt, nrwake, op_ret, ret, val;
	struct linux_emuldata *em;
	struct waiting_proc *wp;
	struct futex f, f2 = NULL;
	int error = 0;

	/*
	* Our implementation provides only privates futexes. Most of the apps
	* should use private futexes but don't claim so. Therefore we treat
	* all futexes as private by clearing the FUTEX_PRIVATE_FLAG. It works
	* in most cases (ie. when futexes are not shared on file descriptor
	* or between different processes.).
	*/
	args->op = args->op & ~LINUX_FUTEX_PRIVATE_FLAG;

	/*
	* Currently support for switching between CLOCK_MONOTONIC and
	* CLOCK_REALTIME is not present. However Linux forbids the use of
	* FUTEX_CLOCK_REALTIME with any op except FUTEX_WAIT_BITSET and
	* FUTEX_WAIT_REQUEUE_PI.
	*/
	clockrt = args->op & LINUX_FUTEX_CLOCK_REALTIME;
	args->op = args->op & ~LINUX_FUTEX_CLOCK_REALTIME;
	if (clockrt && args->op != LINUX_FUTEX_WAIT_BITSET &&
	args->op != LINUX_FUTEX_WAIT_REQUEUE_PI)
	return (ENOSYS);

	switch (args->op) {
	case LINUX_FUTEX_WAIT:
	args->val3 = FUTEX_BITSET_MATCH_ANY;
	/* FALLTHROUGH */

	case LINUX_FUTEX_WAIT_BITSET:

	LINUX_CTR3(sys_futex, "WAIT uaddr %p val %d val3 %d",
	args->uaddr, args->val, args->val3);
	#ifdef DEBUG
	if (ldebug(sys_futex))
	printf(ARGS(sys_futex,
	"futex_wait uaddr %p val %d val3 %d"),
	args->uaddr, args->val, args->val3);
	#endif
	error = futex_get(args->uaddr, &wp, &f, FUTEX_CREATE_WP);
	if (error)
	return (error);
	error = copyin(args->uaddr, &val, sizeof(val));
	if (error) {
	LINUX_CTR1(sys_futex, "WAIT copyin failed %d",
	error);
	futex_put(f, wp);
	return (error);
	}
	if (val != args->val) {
	LINUX_CTR4(sys_futex,
	"WAIT uaddr %p val %d != uval %d val3 %d",
	args->uaddr, args->val, val, args->val3);
	futex_put(f, wp);
	return (EWOULDBLOCK);
	}

	error = futex_wait(f, wp, args->timeout, args->val3);
	break;

	case LINUX_FUTEX_WAKE:
	args->val3 = FUTEX_BITSET_MATCH_ANY;
	/* FALLTHROUGH */

	case LINUX_FUTEX_WAKE_BITSET:

	LINUX_CTR3(sys_futex, "WAKE uaddr %p val % d val3 %d",
	args->uaddr, args->val, args->val3);

	/*
	* XXX: Linux is able to cope with different addresses
	* corresponding to the same mapped memory in the sleeping
	* and waker process(es).
	*/
	#ifdef DEBUG
	if (ldebug(sys_futex))
	printf(ARGS(sys_futex, "futex_wake uaddr %p val %d val3 %d"),
	args->uaddr, args->val, args->val3);
	#endif
	error = futex_get(args->uaddr, NULL, &f, FUTEX_DONTCREATE);
	if (error)
	return (error);
	if (f == NULL) {
	td->td_retval[0] = 0;
	return (error);
	}
	td->td_retval[0] = futex_wake(f, args->val, args->val3);
	futex_put(f, NULL);
	break;

	case LINUX_FUTEX_CMP_REQUEUE:

	LINUX_CTR5(sys_futex, "CMP_REQUEUE uaddr %p "
	"val %d val3 %d uaddr2 %p val2 %d",
	args->uaddr, args->val, args->val3, args->uaddr2,
	(int)(unsigned long)args->timeout);

	#ifdef DEBUG
	if (ldebug(sys_futex))
	printf(ARGS(sys_futex, "futex_cmp_requeue uaddr %p "
	"val %d val3 %d uaddr2 %p val2 %d"),
	args->uaddr, args->val, args->val3, args->uaddr2,
	(int)(unsigned long)args->timeout);
	#endif

	/*
	* Linux allows this, we would not, it is an incorrect
	* usage of declared ABI, so return EINVAL.
	*/
	if (args->uaddr == args->uaddr2)
	return (EINVAL);
	error = futex_get0(args->uaddr, &f, 0);
	if (error)
	return (error);

	/*
	* To avoid deadlocks return EINVAL if second futex
	* exists at this time.
	*
	* Glibc fall back to FUTEX_WAKE in case of any error
	* returned by FUTEX_CMP_REQUEUE.
	*/
	error = futex_get0(args->uaddr2, &f2, FUTEX_DONTEXISTS);
	if (error) {
	futex_put(f, NULL);
	return (error);
	}
	error = copyin(args->uaddr, &val, sizeof(val));
	if (error) {
	LINUX_CTR1(sys_futex, "CMP_REQUEUE copyin failed %d",
	error);
	futex_put(f2, NULL);
	futex_put(f, NULL);
	return (error);
	}
	if (val != args->val3) {
	LINUX_CTR2(sys_futex, "CMP_REQUEUE val %d != uval %d",
	args->val, val);
	futex_put(f2, NULL);
	futex_put(f, NULL);
	return (EAGAIN);
	}

	nrwake = (int)(unsigned long)args->timeout;
	td->td_retval[0] = futex_requeue(f, args->val, f2, nrwake);
	futex_put(f2, NULL);
	futex_put(f, NULL);
	break;

	case LINUX_FUTEX_WAKE_OP:

	LINUX_CTR5(sys_futex, "WAKE_OP "
	"uaddr %p op %d val %x uaddr2 %p val3 %x",
	args->uaddr, args->op, args->val,
	args->uaddr2, args->val3);

	#ifdef DEBUG
	if (ldebug(sys_futex))
	printf(ARGS(sys_futex, "futex_wake_op "
	"uaddr %p op %d val %x uaddr2 %p val3 %x"),
	args->uaddr, args->op, args->val,
	args->uaddr2, args->val3);
	#endif
	error = futex_get0(args->uaddr, &f, 0);
	if (error)
	return (error);
	if (args->uaddr != args->uaddr2)
	error = futex_get0(args->uaddr2, &f2, 0);
	if (error) {
	futex_put(f, NULL);
	return (error);
	}

	/*
	* This function returns positive number as results and
	* negative as errors
	*/
	op_ret = futex_atomic_op(td, args->val3, args->uaddr2);

	if (op_ret < 0) {
	/* XXX: We don't handle the EFAULT yet. */
	if (op_ret != -EFAULT) {
	if (f2 != NULL)
	futex_put(f2, NULL);
	futex_put(f, NULL);
	return (-op_ret);
	}
	if (f2 != NULL)
	futex_put(f2, NULL);
	futex_put(f, NULL);
	return (EFAULT);
	}

	ret = futex_wake(f, args->val, args->val3);

	if (op_ret > 0) {
	op_ret = 0;
	nrwake = (int)(unsigned long)args->timeout;

	if (f2 != NULL)
	op_ret += futex_wake(f2, nrwake, args->val3);
	else
	op_ret += futex_wake(f, nrwake, args->val3);
	ret += op_ret;

	}
	if (f2 != NULL)
	futex_put(f2, NULL);
	futex_put(f, NULL);
	td->td_retval[0] = ret;
	break;

	case LINUX_FUTEX_LOCK_PI:
	/* not yet implemented */
	linux_msg(td,
	"linux_sys_futex: "
	"op LINUX_FUTEX_LOCK_PI not implemented\n");
	return (ENOSYS);

	case LINUX_FUTEX_UNLOCK_PI:
	/* not yet implemented */
	linux_msg(td,
	"linux_sys_futex: "
	"op LINUX_FUTEX_UNLOCK_PI not implemented\n");
	return (ENOSYS);

	case LINUX_FUTEX_TRYLOCK_PI:
	/* not yet implemented */
	linux_msg(td,
	"linux_sys_futex: "
	"op LINUX_FUTEX_TRYLOCK_PI not implemented\n");
	return (ENOSYS);

	case LINUX_FUTEX_REQUEUE:

	/*
	* Glibc does not use this operation since version 2.3.3,
	* as it is racy and replaced by FUTEX_CMP_REQUEUE operation.
	* Glibc versions prior to 2.3.3 fall back to FUTEX_WAKE when
	* FUTEX_REQUEUE returned EINVAL.
	*/
	em = em_find(td->td_proc, EMUL_DONTLOCK);
	if ((em->flags & LINUX_XDEPR_REQUEUEOP) == 0) {
	linux_msg(td,
	"linux_sys_futex: "
	"unsupported futex_requeue op\n");
	em->flags \|= LINUX_XDEPR_REQUEUEOP;
	}
	return (EINVAL);

	case LINUX_FUTEX_WAIT_REQUEUE_PI:
	/* not yet implemented */
	linux_msg(td,
	"linux_sys_futex: "
	"op FUTEX_WAIT_REQUEUE_PI not implemented\n");
	return (ENOSYS);

	case LINUX_FUTEX_CMP_REQUEUE_PI:
	/* not yet implemented */
	linux_msg(td,
	"linux_sys_futex: "
	"op LINUX_FUTEX_CMP_REQUEUE_PI not implemented\n");
	return (ENOSYS);

	default:
	linux_msg(td,
	"linux_sys_futex: unknown op %d\n", args->op);
	return (ENOSYS);
	}

	return (error);
	}

	int
	linux_set_robust_list(struct thread td, struct linux_set_robust_list_args args)
	{
	struct linux_emuldata *em;

	#ifdef DEBUG
	if (ldebug(set_robust_list))
	printf(ARGS(set_robust_list, "head %p len %d"),
	args->head, args->len);
	#endif

	if (args->len != sizeof(struct linux_robust_list_head))
	return (EINVAL);

	em = em_find(td->td_proc, EMUL_DOLOCK);
	em->robust_futexes = args->head;
	EMUL_UNLOCK(&emul_lock);

	return (0);
	}

	int
	linux_get_robust_list(struct thread td, struct linux_get_robust_list_args args)
	{
	struct linux_emuldata *em;
	struct linux_robust_list_head *head;
	l_size_t len = sizeof(struct linux_robust_list_head);
	int error = 0;

	#ifdef DEBUG
	if (ldebug(get_robust_list))
	printf(ARGS(get_robust_list, ""));
	#endif

	if (!args->pid) {
	em = em_find(td->td_proc, EMUL_DONTLOCK);
	head = em->robust_futexes;
	} else {
	struct proc *p;

	p = pfind(args->pid);
	if (p == NULL)
	return (ESRCH);

	em = em_find(p, EMUL_DONTLOCK);
	/* XXX: ptrace? */
	if (priv_check(td, PRIV_CRED_SETUID) \|\|
	priv_check(td, PRIV_CRED_SETEUID) \|\|
	p_candebug(td, p)) {
	PROC_UNLOCK(p);
	return (EPERM);
	}
	head = em->robust_futexes;

	PROC_UNLOCK(p);
	}

	error = copyout(&len, args->len, sizeof(l_size_t));
	if (error)
	return (EFAULT);

	error = copyout(head, args->head, sizeof(struct linux_robust_list_head));

	return (error);
	}

	static int
	handle_futex_death(struct proc p, uint32_t uaddr, int pi)
	{
	uint32_t uval, nval, mval;
	struct futex *f;
	int error;

	retry:
	if (copyin(uaddr, &uval, 4))
	return (EFAULT);
	if ((uval & FUTEX_TID_MASK) == p->p_pid) {
	mval = (uval & FUTEX_WAITERS) \| FUTEX_OWNER_DIED;
	nval = casuword32(uaddr, uval, mval);

	if (nval == -1)
	return (EFAULT);

	if (nval != uval)
	goto retry;

	if (!pi && (uval & FUTEX_WAITERS)) {
	error = futex_get(uaddr, NULL, &f,
	FUTEX_DONTCREATE);
	if (error)
	return (error);
	if (f != NULL) {
	futex_wake(f, 1, FUTEX_BITSET_MATCH_ANY);
	futex_put(f, NULL);
	}
	}
	}

	return (0);
	}

	static int
	fetch_robust_entry(struct linux_robust_list **entry,
	struct linux_robust_list *head, int pi)
	{
	l_ulong uentry;

	if (copyin((const void *)head, &uentry, sizeof(l_ulong)))
	return (EFAULT);

	entry = (void )(uentry & ~1UL);
	*pi = uentry & 1;

	return (0);
	}

	/* This walks the list of robust futexes releasing them. */
	void
	release_futexes(struct proc *p)
	{
	struct linux_robust_list_head *head = NULL;
	struct linux_robust_list entry, next_entry, *pending;
	unsigned int limit = 2048, pi, next_pi, pip;
	struct linux_emuldata *em;
	l_long futex_offset;
	int rc;

	em = em_find(p, EMUL_DONTLOCK);
	head = em->robust_futexes;

	if (head == NULL)
	return;

	if (fetch_robust_entry(&entry, PTRIN(&head->list.next), &pi))
	return;

	if (copyin(&head->futex_offset, &futex_offset, sizeof(futex_offset)))
	return;

	if (fetch_robust_entry(&pending, PTRIN(&head->pending_list), &pip))
	return;

	while (entry != &head->list) {
	rc = fetch_robust_entry(&next_entry, PTRIN(&entry->next), &next_pi);

	if (entry != pending)
	if (handle_futex_death(p, (uint32_t *)entry + futex_offset, pi))
	return;
	if (rc)
	return;

	entry = next_entry;
	pi = next_pi;

	if (!--limit)
	break;

	sched_relinquish(curthread);
	}

	if (pending)
	handle_futex_death(p, (uint32_t *)pending + futex_offset, pip);
	}
	Index: stable/8/sys/contrib/dev/acpica
	===================================================================
	--- stable/8/sys/contrib/dev/acpica (revision 219197)
	+++ stable/8/sys/contrib/dev/acpica (revision 219198)

	Property changes on: stable/8/sys/contrib/dev/acpica
	___________________________________________________________________
	Modified: svn:mergeinfo
	## -0,0 +0,1 ##
	Merged /head/sys/contrib/dev/acpica:r218658,218668
	Index: stable/8/sys/contrib/pf
	===================================================================
	--- stable/8/sys/contrib/pf (revision 219197)
	+++ stable/8/sys/contrib/pf (revision 219198)

	Property changes on: stable/8/sys/contrib/pf
	___________________________________________________________________
	Modified: svn:mergeinfo
	## -0,0 +0,1 ##
	Merged /head/sys/contrib/pf:r218658,218668
	Index: stable/8/sys/i386/linux/linux_sysvec.c
	===================================================================
	--- stable/8/sys/i386/linux/linux_sysvec.c (revision 219197)
	+++ stable/8/sys/i386/linux/linux_sysvec.c (revision 219198)
	@@ -1,1188 +1,1188 @@
	/*-
	* Copyright (c) 1994-1996 Søren Schmidt
	* All rights reserved.
	*
	* Redistribution and use in source and binary forms, with or without
	* modification, are permitted provided that the following conditions
	* are met:
	* 1. Redistributions of source code must retain the above copyright
	* notice, this list of conditions and the following disclaimer
	* in this position and unchanged.
	* 2. Redistributions in binary form must reproduce the above copyright
	* notice, this list of conditions and the following disclaimer in the
	* documentation and/or other materials provided with the distribution.
	* 3. The name of the author may not be used to endorse or promote products
	* derived from this software without specific prior written permission
	*
	* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
	* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
	* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
	* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
	* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
	* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
	* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
	* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
	* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
	* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
	*/

	#include <sys/cdefs.h>
	__FBSDID("$FreeBSD$");

	#include <sys/param.h>
	#include <sys/systm.h>
	#include <sys/exec.h>
	#include <sys/fcntl.h>
	#include <sys/imgact.h>
	#include <sys/imgact_aout.h>
	#include <sys/imgact_elf.h>
	#include <sys/kernel.h>
	#include <sys/lock.h>
	#include <sys/malloc.h>
	#include <sys/module.h>
	#include <sys/mutex.h>
	#include <sys/proc.h>
	#include <sys/signalvar.h>
	#include <sys/syscallsubr.h>
	#include <sys/sysent.h>
	#include <sys/sysproto.h>
	#include <sys/vnode.h>
	#include <sys/eventhandler.h>

	#include <vm/vm.h>
	#include <vm/pmap.h>
	#include <vm/vm_extern.h>
	#include <vm/vm_map.h>
	#include <vm/vm_object.h>
	#include <vm/vm_page.h>
	#include <vm/vm_param.h>

	#include <machine/cpu.h>
	#include <machine/cputypes.h>
	#include <machine/md_var.h>
	#include <machine/pcb.h>

	#include <i386/linux/linux.h>
	#include <i386/linux/linux_proto.h>
	-#include <compat/linux/linux_futex.h>
	#include <compat/linux/linux_emul.h>
	+#include <compat/linux/linux_futex.h>
	#include <compat/linux/linux_mib.h>
	#include <compat/linux/linux_misc.h>
	#include <compat/linux/linux_signal.h>
	#include <compat/linux/linux_util.h>

	MODULE_VERSION(linux, 1);

	MALLOC_DEFINE(M_LINUX, "linux", "Linux mode structures");

	#if BYTE_ORDER == LITTLE_ENDIAN
	#define SHELLMAGIC 0x2123 /* #! */
	#else
	#define SHELLMAGIC 0x2321
	#endif

	/*
	* Allow the sendsig functions to use the ldebug() facility
	* even though they are not syscalls themselves. Map them
	* to syscall 0. This is slightly less bogus than using
	* ldebug(sigreturn).
	*/
	#define LINUX_SYS_linux_rt_sendsig 0
	#define LINUX_SYS_linux_sendsig 0

	extern char linux_sigcode[];
	extern int linux_szsigcode;

	extern struct sysent linux_sysent[LINUX_SYS_MAXSYSCALL];

	SET_DECLARE(linux_ioctl_handler_set, struct linux_ioctl_handler);
	SET_DECLARE(linux_device_handler_set, struct linux_device_handler);

	static int linux_fixup(register_t **stack_base,
	struct image_params *iparams);
	static int elf_linux_fixup(register_t **stack_base,
	struct image_params *iparams);
	static void linux_sendsig(sig_t catcher, ksiginfo_t ksi, sigset_t mask);
	static void exec_linux_setregs(struct thread *td, u_long entry,
	u_long stack, u_long ps_strings);
	static register_t linux_copyout_strings(struct image_params imgp);
	static boolean_t linux_trans_osrel(const Elf_Note note, int32_t osrel);

	static int linux_szplatform;
	const char *linux_platform;

	static eventhandler_tag linux_exit_tag;
	static eventhandler_tag linux_schedtail_tag;
	static eventhandler_tag linux_exec_tag;

	/*
	* Linux syscalls return negative errno's, we do positive and map them
	* Reference:
	* FreeBSD: src/sys/sys/errno.h
	* Linux: linux-2.6.17.8/include/asm-generic/errno-base.h
	* linux-2.6.17.8/include/asm-generic/errno.h
	*/
	static int bsd_to_linux_errno[ELAST + 1] = {
	-0, -1, -2, -3, -4, -5, -6, -7, -8, -9,
	-10, -35, -12, -13, -14, -15, -16, -17, -18, -19,
	-20, -21, -22, -23, -24, -25, -26, -27, -28, -29,
	-30, -31, -32, -33, -34, -11,-115,-114, -88, -89,
	-90, -91, -92, -93, -94, -95, -96, -97, -98, -99,
	-100,-101,-102,-103,-104,-105,-106,-107,-108,-109,
	-110,-111, -40, -36,-112,-113, -39, -11, -87,-122,
	-116, -66, -6, -6, -6, -6, -6, -37, -38, -9,
	-6, -6, -43, -42, -75,-125, -84, -95, -16, -74,
	-72, -67, -71
	};

	int bsd_to_linux_signal[LINUX_SIGTBLSZ] = {
	LINUX_SIGHUP, LINUX_SIGINT, LINUX_SIGQUIT, LINUX_SIGILL,
	LINUX_SIGTRAP, LINUX_SIGABRT, 0, LINUX_SIGFPE,
	LINUX_SIGKILL, LINUX_SIGBUS, LINUX_SIGSEGV, LINUX_SIGSYS,
	LINUX_SIGPIPE, LINUX_SIGALRM, LINUX_SIGTERM, LINUX_SIGURG,
	LINUX_SIGSTOP, LINUX_SIGTSTP, LINUX_SIGCONT, LINUX_SIGCHLD,
	LINUX_SIGTTIN, LINUX_SIGTTOU, LINUX_SIGIO, LINUX_SIGXCPU,
	LINUX_SIGXFSZ, LINUX_SIGVTALRM, LINUX_SIGPROF, LINUX_SIGWINCH,
	0, LINUX_SIGUSR1, LINUX_SIGUSR2
	};

	int linux_to_bsd_signal[LINUX_SIGTBLSZ] = {
	SIGHUP, SIGINT, SIGQUIT, SIGILL,
	SIGTRAP, SIGABRT, SIGBUS, SIGFPE,
	SIGKILL, SIGUSR1, SIGSEGV, SIGUSR2,
	SIGPIPE, SIGALRM, SIGTERM, SIGBUS,
	SIGCHLD, SIGCONT, SIGSTOP, SIGTSTP,
	SIGTTIN, SIGTTOU, SIGURG, SIGXCPU,
	SIGXFSZ, SIGVTALRM, SIGPROF, SIGWINCH,
	SIGIO, SIGURG, SIGSYS
	};

	#define LINUX_T_UNKNOWN 255
	static int _bsd_to_linux_trapcode[] = {
	LINUX_T_UNKNOWN, /* 0 */
	6, /* 1 T_PRIVINFLT */
	LINUX_T_UNKNOWN, /* 2 */
	3, /* 3 T_BPTFLT */
	LINUX_T_UNKNOWN, /* 4 */
	LINUX_T_UNKNOWN, /* 5 */
	16, /* 6 T_ARITHTRAP */
	254, /* 7 T_ASTFLT */
	LINUX_T_UNKNOWN, /* 8 */
	13, /* 9 T_PROTFLT */
	1, /* 10 T_TRCTRAP */
	LINUX_T_UNKNOWN, /* 11 */
	14, /* 12 T_PAGEFLT */
	LINUX_T_UNKNOWN, /* 13 */
	17, /* 14 T_ALIGNFLT */
	LINUX_T_UNKNOWN, /* 15 */
	LINUX_T_UNKNOWN, /* 16 */
	LINUX_T_UNKNOWN, /* 17 */
	0, /* 18 T_DIVIDE */
	2, /* 19 T_NMI */
	4, /* 20 T_OFLOW */
	5, /* 21 T_BOUND */
	7, /* 22 T_DNA */
	8, /* 23 T_DOUBLEFLT */
	9, /* 24 T_FPOPFLT */
	10, /* 25 T_TSSFLT */
	11, /* 26 T_SEGNPFLT */
	12, /* 27 T_STKFLT */
	18, /* 28 T_MCHK */
	19, /* 29 T_XMMFLT */
	15 /* 30 T_RESERVED */
	};
	#define bsd_to_linux_trapcode(code) \
	((code)<sizeof(_bsd_to_linux_trapcode)/sizeof(*_bsd_to_linux_trapcode)? \
	_bsd_to_linux_trapcode[(code)]: \
	LINUX_T_UNKNOWN)

	/*
	* If FreeBSD & Linux have a difference of opinion about what a trap
	* means, deal with it here.
	*
	* MPSAFE
	*/
	static int
	translate_traps(int signal, int trap_code)
	{
	if (signal != SIGBUS)
	return signal;
	switch (trap_code) {
	case T_PROTFLT:
	case T_TSSFLT:
	case T_DOUBLEFLT:
	case T_PAGEFLT:
	return SIGSEGV;
	default:
	return signal;
	}
	}

	static int
	linux_fixup(register_t *stack_base, struct image_params imgp)
	{
	register_t argv, envp;

	argv = *stack_base;
	envp = *stack_base + (imgp->args->argc + 1);
	(*stack_base)--;
	*stack_base = (intptr_t)(void )envp;
	(*stack_base)--;
	*stack_base = (intptr_t)(void )argv;
	(*stack_base)--;
	**stack_base = imgp->args->argc;
	return (0);
	}

	static int
	elf_linux_fixup(register_t *stack_base, struct image_params imgp)
	{
	struct proc *p;
	Elf32_Auxargs *args;
	Elf32_Addr *uplatform;
	struct ps_strings *arginfo;
	register_t *pos;

	KASSERT(curthread->td_proc == imgp->proc,
	("unsafe elf_linux_fixup(), should be curproc"));

	p = imgp->proc;
	arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings;
	uplatform = (Elf32_Addr *)((caddr_t)arginfo - linux_szsigcode -
	linux_szplatform);
	args = (Elf32_Auxargs *)imgp->auxargs;
	pos = *stack_base + (imgp->args->argc + imgp->args->envc + 2);

	AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, cpu_feature);

	/*
	* Do not export AT_CLKTCK when emulating Linux kernel prior to 2.4.0,
	* as it has appeared in the 2.4.0-rc7 first time.
	* Being exported, AT_CLKTCK is returned by sysconf(_SC_CLK_TCK),
	* glibc falls back to the hard-coded CLK_TCK value when aux entry
	* is not present.
	* Also see linux_times() implementation.
	*/
	if (linux_kernver(curthread) >= LINUX_KERNVER_2004000)
	AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, stclohz);
	AUXARGS_ENTRY(pos, AT_PHDR, args->phdr);
	AUXARGS_ENTRY(pos, AT_PHENT, args->phent);
	AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum);
	AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz);
	AUXARGS_ENTRY(pos, AT_FLAGS, args->flags);
	AUXARGS_ENTRY(pos, AT_ENTRY, args->entry);
	AUXARGS_ENTRY(pos, AT_BASE, args->base);
	AUXARGS_ENTRY(pos, LINUX_AT_SECURE, 0);
	AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid);
	AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid);
	AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid);
	AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid);
	AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(uplatform));
	if (args->execfd != -1)
	AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd);
	AUXARGS_ENTRY(pos, AT_NULL, 0);

	free(imgp->auxargs, M_TEMP);
	imgp->auxargs = NULL;

	(*stack_base)--;
	**stack_base = (register_t)imgp->args->argc;
	return (0);
	}

	/*
	* Copied from kern/kern_exec.c
	*/
	static register_t *
	linux_copyout_strings(struct image_params *imgp)
	{
	int argc, envc;
	char **vectp;
	char stringp, destp;
	register_t *stack_base;
	struct ps_strings *arginfo;
	struct proc *p;

	/*
	* Calculate string base and vector table pointers.
	* Also deal with signal trampoline code for this exec type.
	*/
	p = imgp->proc;
	arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings;
	destp = (caddr_t)arginfo - linux_szsigcode - SPARE_USRSPACE -
	linux_szplatform - roundup((ARG_MAX - imgp->args->stringspace),
	sizeof(char *));

	/*
	* install sigcode
	*/
	copyout(p->p_sysent->sv_sigcode, ((caddr_t)arginfo -
	linux_szsigcode), linux_szsigcode);

	/*
	* install LINUX_PLATFORM
	*/
	copyout(linux_platform, ((caddr_t)arginfo - linux_szsigcode -
	linux_szplatform), linux_szplatform);

	/*
	* If we have a valid auxargs ptr, prepare some room
	* on the stack.
	*/
	if (imgp->auxargs) {
	/*
	* 'AT_COUNT*2' is size for the ELF Auxargs data. This is for
	* lower compatibility.
	*/
	imgp->auxarg_size = (imgp->auxarg_size) ? imgp->auxarg_size :
	(LINUX_AT_COUNT * 2);
	/*
	* The '+ 2' is for the null pointers at the end of each of
	* the arg and env vector sets,and imgp->auxarg_size is room
	* for argument of Runtime loader.
	*/
	vectp = (char **)(destp - (imgp->args->argc +
	imgp->args->envc + 2 + imgp->auxarg_size) * sizeof(char *));
	} else {
	/*
	* The '+ 2' is for the null pointers at the end of each of
	* the arg and env vector sets
	*/
	vectp = (char *)(destp - (imgp->args->argc + imgp->args->envc + 2)
	sizeof(char *));
	}

	/*
	* vectp also becomes our initial stack base
	*/
	stack_base = (register_t *)vectp;

	stringp = imgp->args->begin_argv;
	argc = imgp->args->argc;
	envc = imgp->args->envc;

	/*
	* Copy out strings - arguments and environment.
	*/
	copyout(stringp, destp, ARG_MAX - imgp->args->stringspace);

	/*
	* Fill in "ps_strings" struct for ps, w, etc.
	*/
	suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
	suword(&arginfo->ps_nargvstr, argc);

	/*
	* Fill in argument portion of vector table.
	*/
	for (; argc > 0; --argc) {
	suword(vectp++, (long)(intptr_t)destp);
	while (*stringp++ != 0)
	destp++;
	destp++;
	}

	/* a null vector table pointer separates the argp's from the envp's */
	suword(vectp++, 0);

	suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
	suword(&arginfo->ps_nenvstr, envc);

	/*
	* Fill in environment portion of vector table.
	*/
	for (; envc > 0; --envc) {
	suword(vectp++, (long)(intptr_t)destp);
	while (*stringp++ != 0)
	destp++;
	destp++;
	}

	/* end of vector table is a null pointer */
	suword(vectp, 0);

	return (stack_base);
	}



	extern int _ucodesel, _udatasel;
	extern unsigned long linux_sznonrtsigcode;

	static void
	linux_rt_sendsig(sig_t catcher, ksiginfo_t ksi, sigset_t mask)
	{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct sigacts *psp;
	struct trapframe *regs;
	struct l_rt_sigframe *fp, frame;
	int sig, code;
	int oonstack;

	sig = ksi->ksi_signo;
	code = ksi->ksi_code;
	PROC_LOCK_ASSERT(p, MA_OWNED);
	psp = p->p_sigacts;
	mtx_assert(&psp->ps_mtx, MA_OWNED);
	regs = td->td_frame;
	oonstack = sigonstack(regs->tf_esp);

	#ifdef DEBUG
	if (ldebug(rt_sendsig))
	printf(ARGS(rt_sendsig, "%p, %d, %p, %u"),
	catcher, sig, (void*)mask, code);
	#endif
	/*
	* Allocate space for the signal handler context.
	*/
	if ((td->td_pflags & TDP_ALTSTACK) && !oonstack &&
	SIGISMEMBER(psp->ps_sigonstack, sig)) {
	fp = (struct l_rt_sigframe *)(td->td_sigstk.ss_sp +
	td->td_sigstk.ss_size - sizeof(struct l_rt_sigframe));
	} else
	fp = (struct l_rt_sigframe *)regs->tf_esp - 1;
	mtx_unlock(&psp->ps_mtx);

	/*
	* Build the argument list for the signal handler.
	*/
	if (p->p_sysent->sv_sigtbl)
	if (sig <= p->p_sysent->sv_sigsize)
	sig = p->p_sysent->sv_sigtbl[_SIG_IDX(sig)];

	bzero(&frame, sizeof(frame));

	frame.sf_handler = catcher;
	frame.sf_sig = sig;
	frame.sf_siginfo = &fp->sf_si;
	frame.sf_ucontext = &fp->sf_sc;

	/* Fill in POSIX parts */
	ksiginfo_to_lsiginfo(ksi, &frame.sf_si, sig);

	/*
	* Build the signal context to be used by sigreturn.
	*/
	frame.sf_sc.uc_flags = 0; /* XXX ??? */
	frame.sf_sc.uc_link = NULL; /* XXX ??? */

	frame.sf_sc.uc_stack.ss_sp = td->td_sigstk.ss_sp;
	frame.sf_sc.uc_stack.ss_size = td->td_sigstk.ss_size;
	frame.sf_sc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK)
	? ((oonstack) ? LINUX_SS_ONSTACK : 0) : LINUX_SS_DISABLE;
	PROC_UNLOCK(p);

	bsd_to_linux_sigset(mask, &frame.sf_sc.uc_sigmask);

	frame.sf_sc.uc_mcontext.sc_mask = frame.sf_sc.uc_sigmask.__bits[0];
	frame.sf_sc.uc_mcontext.sc_gs = rgs();
	frame.sf_sc.uc_mcontext.sc_fs = regs->tf_fs;
	frame.sf_sc.uc_mcontext.sc_es = regs->tf_es;
	frame.sf_sc.uc_mcontext.sc_ds = regs->tf_ds;
	frame.sf_sc.uc_mcontext.sc_edi = regs->tf_edi;
	frame.sf_sc.uc_mcontext.sc_esi = regs->tf_esi;
	frame.sf_sc.uc_mcontext.sc_ebp = regs->tf_ebp;
	frame.sf_sc.uc_mcontext.sc_ebx = regs->tf_ebx;
	frame.sf_sc.uc_mcontext.sc_edx = regs->tf_edx;
	frame.sf_sc.uc_mcontext.sc_ecx = regs->tf_ecx;
	frame.sf_sc.uc_mcontext.sc_eax = regs->tf_eax;
	frame.sf_sc.uc_mcontext.sc_eip = regs->tf_eip;
	frame.sf_sc.uc_mcontext.sc_cs = regs->tf_cs;
	frame.sf_sc.uc_mcontext.sc_eflags = regs->tf_eflags;
	frame.sf_sc.uc_mcontext.sc_esp_at_signal = regs->tf_esp;
	frame.sf_sc.uc_mcontext.sc_ss = regs->tf_ss;
	frame.sf_sc.uc_mcontext.sc_err = regs->tf_err;
	frame.sf_sc.uc_mcontext.sc_cr2 = (register_t)ksi->ksi_addr;
	frame.sf_sc.uc_mcontext.sc_trapno = bsd_to_linux_trapcode(code);

	#ifdef DEBUG
	if (ldebug(rt_sendsig))
	printf(LMSG("rt_sendsig flags: 0x%x, sp: %p, ss: 0x%x, mask: 0x%x"),
	frame.sf_sc.uc_stack.ss_flags, td->td_sigstk.ss_sp,
	td->td_sigstk.ss_size, frame.sf_sc.uc_mcontext.sc_mask);
	#endif

	if (copyout(&frame, fp, sizeof(frame)) != 0) {
	/*
	* Process has trashed its stack; give it an illegal
	* instruction to halt it in its tracks.
	*/
	#ifdef DEBUG
	if (ldebug(rt_sendsig))
	printf(LMSG("rt_sendsig: bad stack %p, oonstack=%x"),
	fp, oonstack);
	#endif
	PROC_LOCK(p);
	sigexit(td, SIGILL);
	}

	/*
	* Build context to run handler in.
	*/
	regs->tf_esp = (int)fp;
	regs->tf_eip = PS_STRINGS - *(p->p_sysent->sv_szsigcode) +
	linux_sznonrtsigcode;
	regs->tf_eflags &= ~(PSL_T \| PSL_VM \| PSL_D);
	regs->tf_cs = _ucodesel;
	regs->tf_ds = _udatasel;
	regs->tf_es = _udatasel;
	regs->tf_fs = _udatasel;
	regs->tf_ss = _udatasel;
	PROC_LOCK(p);
	mtx_lock(&psp->ps_mtx);
	}


	/*
	* Send an interrupt to process.
	*
	* Stack is set up to allow sigcode stored
	* in u. to call routine, followed by kcall
	* to sigreturn routine below. After sigreturn
	* resets the signal mask, the stack, and the
	* frame pointer, it returns to the user
	* specified pc, psl.
	*/
	static void
	linux_sendsig(sig_t catcher, ksiginfo_t ksi, sigset_t mask)
	{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct sigacts *psp;
	struct trapframe *regs;
	struct l_sigframe *fp, frame;
	l_sigset_t lmask;
	int sig, code;
	int oonstack, i;

	PROC_LOCK_ASSERT(p, MA_OWNED);
	psp = p->p_sigacts;
	sig = ksi->ksi_signo;
	code = ksi->ksi_code;
	mtx_assert(&psp->ps_mtx, MA_OWNED);
	if (SIGISMEMBER(psp->ps_siginfo, sig)) {
	/* Signal handler installed with SA_SIGINFO. */
	linux_rt_sendsig(catcher, ksi, mask);
	return;
	}
	regs = td->td_frame;
	oonstack = sigonstack(regs->tf_esp);

	#ifdef DEBUG
	if (ldebug(sendsig))
	printf(ARGS(sendsig, "%p, %d, %p, %u"),
	catcher, sig, (void*)mask, code);
	#endif

	/*
	* Allocate space for the signal handler context.
	*/
	if ((td->td_pflags & TDP_ALTSTACK) && !oonstack &&
	SIGISMEMBER(psp->ps_sigonstack, sig)) {
	fp = (struct l_sigframe *)(td->td_sigstk.ss_sp +
	td->td_sigstk.ss_size - sizeof(struct l_sigframe));
	} else
	fp = (struct l_sigframe *)regs->tf_esp - 1;
	mtx_unlock(&psp->ps_mtx);
	PROC_UNLOCK(p);

	/*
	* Build the argument list for the signal handler.
	*/
	if (p->p_sysent->sv_sigtbl)
	if (sig <= p->p_sysent->sv_sigsize)
	sig = p->p_sysent->sv_sigtbl[_SIG_IDX(sig)];

	bzero(&frame, sizeof(frame));

	frame.sf_handler = catcher;
	frame.sf_sig = sig;

	bsd_to_linux_sigset(mask, &lmask);

	/*
	* Build the signal context to be used by sigreturn.
	*/
	frame.sf_sc.sc_mask = lmask.__bits[0];
	frame.sf_sc.sc_gs = rgs();
	frame.sf_sc.sc_fs = regs->tf_fs;
	frame.sf_sc.sc_es = regs->tf_es;
	frame.sf_sc.sc_ds = regs->tf_ds;
	frame.sf_sc.sc_edi = regs->tf_edi;
	frame.sf_sc.sc_esi = regs->tf_esi;
	frame.sf_sc.sc_ebp = regs->tf_ebp;
	frame.sf_sc.sc_ebx = regs->tf_ebx;
	frame.sf_sc.sc_edx = regs->tf_edx;
	frame.sf_sc.sc_ecx = regs->tf_ecx;
	frame.sf_sc.sc_eax = regs->tf_eax;
	frame.sf_sc.sc_eip = regs->tf_eip;
	frame.sf_sc.sc_cs = regs->tf_cs;
	frame.sf_sc.sc_eflags = regs->tf_eflags;
	frame.sf_sc.sc_esp_at_signal = regs->tf_esp;
	frame.sf_sc.sc_ss = regs->tf_ss;
	frame.sf_sc.sc_err = regs->tf_err;
	frame.sf_sc.sc_cr2 = (register_t)ksi->ksi_addr;
	frame.sf_sc.sc_trapno = bsd_to_linux_trapcode(ksi->ksi_trapno);

	for (i = 0; i < (LINUX_NSIG_WORDS-1); i++)
	frame.sf_extramask[i] = lmask.__bits[i+1];

	if (copyout(&frame, fp, sizeof(frame)) != 0) {
	/*
	* Process has trashed its stack; give it an illegal
	* instruction to halt it in its tracks.
	*/
	PROC_LOCK(p);
	sigexit(td, SIGILL);
	}

	/*
	* Build context to run handler in.
	*/
	regs->tf_esp = (int)fp;
	regs->tf_eip = PS_STRINGS - *(p->p_sysent->sv_szsigcode);
	regs->tf_eflags &= ~(PSL_T \| PSL_VM \| PSL_D);
	regs->tf_cs = _ucodesel;
	regs->tf_ds = _udatasel;
	regs->tf_es = _udatasel;
	regs->tf_fs = _udatasel;
	regs->tf_ss = _udatasel;
	PROC_LOCK(p);
	mtx_lock(&psp->ps_mtx);
	}

	/*
	* System call to cleanup state after a signal
	* has been taken. Reset signal mask and
	* stack state from context left by sendsig (above).
	* Return to previous pc and psl as specified by
	* context left by sendsig. Check carefully to
	* make sure that the user has not modified the
	* psl to gain improper privileges or to cause
	* a machine fault.
	*/
	int
	linux_sigreturn(struct thread td, struct linux_sigreturn_args args)
	{
	struct l_sigframe frame;
	struct trapframe *regs;
	l_sigset_t lmask;
	sigset_t bmask;
	int eflags, i;
	ksiginfo_t ksi;

	regs = td->td_frame;

	#ifdef DEBUG
	if (ldebug(sigreturn))
	printf(ARGS(sigreturn, "%p"), (void *)args->sfp);
	#endif
	/*
	* The trampoline code hands us the sigframe.
	* It is unsafe to keep track of it ourselves, in the event that a
	* program jumps out of a signal handler.
	*/
	if (copyin(args->sfp, &frame, sizeof(frame)) != 0)
	return (EFAULT);

	/*
	* Check for security violations.
	*/
	#define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0)
	eflags = frame.sf_sc.sc_eflags;
	/*
	* XXX do allow users to change the privileged flag PSL_RF. The
	* cpu sets PSL_RF in tf_eflags for faults. Debuggers should
	* sometimes set it there too. tf_eflags is kept in the signal
	* context during signal handling and there is no other place
	* to remember it, so the PSL_RF bit may be corrupted by the
	* signal handler without us knowing. Corruption of the PSL_RF
	* bit at worst causes one more or one less debugger trap, so
	* allowing it is fairly harmless.
	*/
	if (!EFLAGS_SECURE(eflags & ~PSL_RF, regs->tf_eflags & ~PSL_RF))
	return(EINVAL);

	/*
	* Don't allow users to load a valid privileged %cs. Let the
	* hardware check for invalid selectors, excess privilege in
	* other selectors, invalid %eip's and invalid %esp's.
	*/
	#define CS_SECURE(cs) (ISPL(cs) == SEL_UPL)
	if (!CS_SECURE(frame.sf_sc.sc_cs)) {
	ksiginfo_init_trap(&ksi);
	ksi.ksi_signo = SIGBUS;
	ksi.ksi_code = BUS_OBJERR;
	ksi.ksi_trapno = T_PROTFLT;
	ksi.ksi_addr = (void *)regs->tf_eip;
	trapsignal(td, &ksi);
	return(EINVAL);
	}

	lmask.__bits[0] = frame.sf_sc.sc_mask;
	for (i = 0; i < (LINUX_NSIG_WORDS-1); i++)
	lmask.__bits[i+1] = frame.sf_extramask[i];
	linux_to_bsd_sigset(&lmask, &bmask);
	kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0);

	/*
	* Restore signal context.
	*/
	/* %gs was restored by the trampoline. */
	regs->tf_fs = frame.sf_sc.sc_fs;
	regs->tf_es = frame.sf_sc.sc_es;
	regs->tf_ds = frame.sf_sc.sc_ds;
	regs->tf_edi = frame.sf_sc.sc_edi;
	regs->tf_esi = frame.sf_sc.sc_esi;
	regs->tf_ebp = frame.sf_sc.sc_ebp;
	regs->tf_ebx = frame.sf_sc.sc_ebx;
	regs->tf_edx = frame.sf_sc.sc_edx;
	regs->tf_ecx = frame.sf_sc.sc_ecx;
	regs->tf_eax = frame.sf_sc.sc_eax;
	regs->tf_eip = frame.sf_sc.sc_eip;
	regs->tf_cs = frame.sf_sc.sc_cs;
	regs->tf_eflags = eflags;
	regs->tf_esp = frame.sf_sc.sc_esp_at_signal;
	regs->tf_ss = frame.sf_sc.sc_ss;

	return (EJUSTRETURN);
	}

	/*
	* System call to cleanup state after a signal
	* has been taken. Reset signal mask and
	* stack state from context left by rt_sendsig (above).
	* Return to previous pc and psl as specified by
	* context left by sendsig. Check carefully to
	* make sure that the user has not modified the
	* psl to gain improper privileges or to cause
	* a machine fault.
	*/
	int
	linux_rt_sigreturn(struct thread td, struct linux_rt_sigreturn_args args)
	{
	struct l_ucontext uc;
	struct l_sigcontext *context;
	sigset_t bmask;
	l_stack_t *lss;
	stack_t ss;
	struct trapframe *regs;
	int eflags;
	ksiginfo_t ksi;

	regs = td->td_frame;

	#ifdef DEBUG
	if (ldebug(rt_sigreturn))
	printf(ARGS(rt_sigreturn, "%p"), (void *)args->ucp);
	#endif
	/*
	* The trampoline code hands us the ucontext.
	* It is unsafe to keep track of it ourselves, in the event that a
	* program jumps out of a signal handler.
	*/
	if (copyin(args->ucp, &uc, sizeof(uc)) != 0)
	return (EFAULT);

	context = &uc.uc_mcontext;

	/*
	* Check for security violations.
	*/
	#define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0)
	eflags = context->sc_eflags;
	/*
	* XXX do allow users to change the privileged flag PSL_RF. The
	* cpu sets PSL_RF in tf_eflags for faults. Debuggers should
	* sometimes set it there too. tf_eflags is kept in the signal
	* context during signal handling and there is no other place
	* to remember it, so the PSL_RF bit may be corrupted by the
	* signal handler without us knowing. Corruption of the PSL_RF
	* bit at worst causes one more or one less debugger trap, so
	* allowing it is fairly harmless.
	*/
	if (!EFLAGS_SECURE(eflags & ~PSL_RF, regs->tf_eflags & ~PSL_RF))
	return(EINVAL);

	/*
	* Don't allow users to load a valid privileged %cs. Let the
	* hardware check for invalid selectors, excess privilege in
	* other selectors, invalid %eip's and invalid %esp's.
	*/
	#define CS_SECURE(cs) (ISPL(cs) == SEL_UPL)
	if (!CS_SECURE(context->sc_cs)) {
	ksiginfo_init_trap(&ksi);
	ksi.ksi_signo = SIGBUS;
	ksi.ksi_code = BUS_OBJERR;
	ksi.ksi_trapno = T_PROTFLT;
	ksi.ksi_addr = (void *)regs->tf_eip;
	trapsignal(td, &ksi);
	return(EINVAL);
	}

	linux_to_bsd_sigset(&uc.uc_sigmask, &bmask);
	kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0);

	/*
	* Restore signal context
	*/
	/* %gs was restored by the trampoline. */
	regs->tf_fs = context->sc_fs;
	regs->tf_es = context->sc_es;
	regs->tf_ds = context->sc_ds;
	regs->tf_edi = context->sc_edi;
	regs->tf_esi = context->sc_esi;
	regs->tf_ebp = context->sc_ebp;
	regs->tf_ebx = context->sc_ebx;
	regs->tf_edx = context->sc_edx;
	regs->tf_ecx = context->sc_ecx;
	regs->tf_eax = context->sc_eax;
	regs->tf_eip = context->sc_eip;
	regs->tf_cs = context->sc_cs;
	regs->tf_eflags = eflags;
	regs->tf_esp = context->sc_esp_at_signal;
	regs->tf_ss = context->sc_ss;

	/*
	* call sigaltstack & ignore results..
	*/
	lss = &uc.uc_stack;
	ss.ss_sp = lss->ss_sp;
	ss.ss_size = lss->ss_size;
	ss.ss_flags = linux_to_bsd_sigaltstack(lss->ss_flags);

	#ifdef DEBUG
	if (ldebug(rt_sigreturn))
	printf(LMSG("rt_sigret flags: 0x%x, sp: %p, ss: 0x%x, mask: 0x%x"),
	ss.ss_flags, ss.ss_sp, ss.ss_size, context->sc_mask);
	#endif
	(void)kern_sigaltstack(td, &ss, NULL);

	return (EJUSTRETURN);
	}

	static int
	linux_fetch_syscall_args(struct thread td, struct syscall_args sa)
	{
	struct proc *p;
	struct trapframe *frame;

	p = td->td_proc;
	frame = td->td_frame;

	sa->code = frame->tf_eax;
	sa->args[0] = frame->tf_ebx;
	sa->args[1] = frame->tf_ecx;
	sa->args[2] = frame->tf_edx;
	sa->args[3] = frame->tf_esi;
	sa->args[4] = frame->tf_edi;
	sa->args[5] = frame->tf_ebp; /* Unconfirmed */

	if (sa->code >= p->p_sysent->sv_size)
	sa->callp = &p->p_sysent->sv_table[0];
	else
	sa->callp = &p->p_sysent->sv_table[sa->code];
	sa->narg = sa->callp->sy_narg;

	td->td_retval[0] = 0;
	td->td_retval[1] = frame->tf_edx;

	return (0);
	}

	/*
	* If a linux binary is exec'ing something, try this image activator
	* first. We override standard shell script execution in order to
	* be able to modify the interpreter path. We only do this if a linux
	* binary is doing the exec, so we do not create an EXEC module for it.
	*/
	static int exec_linux_imgact_try(struct image_params *iparams);

	static int
	exec_linux_imgact_try(struct image_params *imgp)
	{
	const char head = (const char )imgp->image_header;
	char *rpath;
	int error = -1, len;

	/*
	* The interpreter for shell scripts run from a linux binary needs
	* to be located in /compat/linux if possible in order to recursively
	* maintain linux path emulation.
	*/
	if (((const short *)head)[0] == SHELLMAGIC) {
	/*
	* Run our normal shell image activator. If it succeeds attempt
	* to use the alternate path for the interpreter. If an alternate
	* path is found, use our stringspace to store it.
	*/
	if ((error = exec_shell_imgact(imgp)) == 0) {
	linux_emul_convpath(FIRST_THREAD_IN_PROC(imgp->proc),
	imgp->interpreter_name, UIO_SYSSPACE, &rpath, 0, AT_FDCWD);
	if (rpath != NULL) {
	len = strlen(rpath) + 1;

	if (len <= MAXSHELLCMDLEN) {
	memcpy(imgp->interpreter_name, rpath, len);
	}
	free(rpath, M_TEMP);
	}
	}
	}
	return(error);
	}

	/*
	* exec_setregs may initialize some registers differently than Linux
	* does, thus potentially confusing Linux binaries. If necessary, we
	* override the exec_setregs default(s) here.
	*/
	static void
	exec_linux_setregs(struct thread *td, u_long entry,
	u_long stack, u_long ps_strings)
	{
	struct pcb *pcb = td->td_pcb;

	exec_setregs(td, entry, stack, ps_strings);

	/* Linux sets %gs to 0, we default to _udatasel */
	pcb->pcb_gs = 0;
	load_gs(0);

	pcb->pcb_initial_npxcw = __LINUX_NPXCW__;
	}

	static void
	linux_get_machine(const char **dst)
	{

	switch (cpu_class) {
	case CPUCLASS_686:
	*dst = "i686";
	break;
	case CPUCLASS_586:
	*dst = "i586";
	break;
	case CPUCLASS_486:
	*dst = "i486";
	break;
	default:
	*dst = "i386";
	}
	}

	struct sysentvec linux_sysvec = {
	.sv_size = LINUX_SYS_MAXSYSCALL,
	.sv_table = linux_sysent,
	.sv_mask = 0,
	.sv_sigsize = LINUX_SIGTBLSZ,
	.sv_sigtbl = bsd_to_linux_signal,
	.sv_errsize = ELAST + 1,
	.sv_errtbl = bsd_to_linux_errno,
	.sv_transtrap = translate_traps,
	.sv_fixup = linux_fixup,
	.sv_sendsig = linux_sendsig,
	.sv_sigcode = linux_sigcode,
	.sv_szsigcode = &linux_szsigcode,
	.sv_prepsyscall = NULL,
	.sv_name = "Linux a.out",
	.sv_coredump = NULL,
	.sv_imgact_try = exec_linux_imgact_try,
	.sv_minsigstksz = LINUX_MINSIGSTKSZ,
	.sv_pagesize = PAGE_SIZE,
	.sv_minuser = VM_MIN_ADDRESS,
	.sv_maxuser = VM_MAXUSER_ADDRESS,
	.sv_usrstack = USRSTACK,
	.sv_psstrings = PS_STRINGS,
	.sv_stackprot = VM_PROT_ALL,
	.sv_copyout_strings = exec_copyout_strings,
	.sv_setregs = exec_linux_setregs,
	.sv_fixlimit = NULL,
	.sv_maxssiz = NULL,
	.sv_flags = SV_ABI_LINUX \| SV_AOUT \| SV_IA32 \| SV_ILP32,
	.sv_set_syscall_retval = cpu_set_syscall_retval,
	.sv_fetch_syscall_args = linux_fetch_syscall_args,
	.sv_syscallnames = NULL,
	};

	struct sysentvec elf_linux_sysvec = {
	.sv_size = LINUX_SYS_MAXSYSCALL,
	.sv_table = linux_sysent,
	.sv_mask = 0,
	.sv_sigsize = LINUX_SIGTBLSZ,
	.sv_sigtbl = bsd_to_linux_signal,
	.sv_errsize = ELAST + 1,
	.sv_errtbl = bsd_to_linux_errno,
	.sv_transtrap = translate_traps,
	.sv_fixup = elf_linux_fixup,
	.sv_sendsig = linux_sendsig,
	.sv_sigcode = linux_sigcode,
	.sv_szsigcode = &linux_szsigcode,
	.sv_prepsyscall = NULL,
	.sv_name = "Linux ELF",
	.sv_coredump = elf32_coredump,
	.sv_imgact_try = exec_linux_imgact_try,
	.sv_minsigstksz = LINUX_MINSIGSTKSZ,
	.sv_pagesize = PAGE_SIZE,
	.sv_minuser = VM_MIN_ADDRESS,
	.sv_maxuser = VM_MAXUSER_ADDRESS,
	.sv_usrstack = USRSTACK,
	.sv_psstrings = PS_STRINGS,
	.sv_stackprot = VM_PROT_ALL,
	.sv_copyout_strings = linux_copyout_strings,
	.sv_setregs = exec_linux_setregs,
	.sv_fixlimit = NULL,
	.sv_maxssiz = NULL,
	.sv_flags = SV_ABI_LINUX \| SV_IA32 \| SV_ILP32,
	.sv_set_syscall_retval = cpu_set_syscall_retval,
	.sv_fetch_syscall_args = linux_fetch_syscall_args,
	.sv_syscallnames = NULL,
	};

	static char GNU_ABI_VENDOR[] = "GNU";
	static int GNULINUX_ABI_DESC = 0;

	static boolean_t
	linux_trans_osrel(const Elf_Note note, int32_t osrel)
	{
	const Elf32_Word *desc;
	uintptr_t p;

	p = (uintptr_t)(note + 1);
	p += roundup2(note->n_namesz, sizeof(Elf32_Addr));

	desc = (const Elf32_Word *)p;
	if (desc[0] != GNULINUX_ABI_DESC)
	return (FALSE);

	/*
	* For linux we encode osrel as follows (see linux_mib.c):
	* VVVMMMIII (version, major, minor), see linux_mib.c.
	*/
	osrel = desc[1] 1000000 + desc[2] * 1000 + desc[3];

	return (TRUE);
	}

	static Elf_Brandnote linux_brandnote = {
	.hdr.n_namesz = sizeof(GNU_ABI_VENDOR),
	.hdr.n_descsz = 16, /* XXX at least 16 */
	.hdr.n_type = 1,
	.vendor = GNU_ABI_VENDOR,
	.flags = BN_TRANSLATE_OSREL,
	.trans_osrel = linux_trans_osrel
	};

	static Elf32_Brandinfo linux_brand = {
	.brand = ELFOSABI_LINUX,
	.machine = EM_386,
	.compat_3_brand = "Linux",
	.emul_path = "/compat/linux",
	.interp_path = "/lib/ld-linux.so.1",
	.sysvec = &elf_linux_sysvec,
	.interp_newpath = NULL,
	.brand_note = &linux_brandnote,
	.flags = BI_CAN_EXEC_DYN \| BI_BRAND_NOTE
	};

	static Elf32_Brandinfo linux_glibc2brand = {
	.brand = ELFOSABI_LINUX,
	.machine = EM_386,
	.compat_3_brand = "Linux",
	.emul_path = "/compat/linux",
	.interp_path = "/lib/ld-linux.so.2",
	.sysvec = &elf_linux_sysvec,
	.interp_newpath = NULL,
	.brand_note = &linux_brandnote,
	.flags = BI_CAN_EXEC_DYN \| BI_BRAND_NOTE
	};

	Elf32_Brandinfo *linux_brandlist[] = {
	&linux_brand,
	&linux_glibc2brand,
	NULL
	};

	static int
	linux_elf_modevent(module_t mod, int type, void *data)
	{
	Elf32_Brandinfo **brandinfo;
	int error;
	struct linux_ioctl_handler **lihp;
	struct linux_device_handler **ldhp;

	error = 0;

	switch(type) {
	case MOD_LOAD:
	for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
	++brandinfo)
	if (elf32_insert_brand_entry(*brandinfo) < 0)
	error = EINVAL;
	if (error == 0) {
	SET_FOREACH(lihp, linux_ioctl_handler_set)
	linux_ioctl_register_handler(*lihp);
	SET_FOREACH(ldhp, linux_device_handler_set)
	linux_device_register_handler(*ldhp);
	mtx_init(&emul_lock, "emuldata lock", NULL, MTX_DEF);
	sx_init(&emul_shared_lock, "emuldata->shared lock");
	LIST_INIT(&futex_list);
	mtx_init(&futex_mtx, "ftllk", NULL, MTX_DEF);
	linux_exit_tag = EVENTHANDLER_REGISTER(process_exit, linux_proc_exit,
	NULL, 1000);
	linux_schedtail_tag = EVENTHANDLER_REGISTER(schedtail, linux_schedtail,
	NULL, 1000);
	linux_exec_tag = EVENTHANDLER_REGISTER(process_exec, linux_proc_exec,
	NULL, 1000);
	linux_get_machine(&linux_platform);
	linux_szplatform = roundup(strlen(linux_platform) + 1,
	sizeof(char *));
	linux_osd_jail_register();
	stclohz = (stathz ? stathz : hz);
	if (bootverbose)
	printf("Linux ELF exec handler installed\n");
	} else
	printf("cannot insert Linux ELF brand handler\n");
	break;
	case MOD_UNLOAD:
	for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
	++brandinfo)
	if (elf32_brand_inuse(*brandinfo))
	error = EBUSY;
	if (error == 0) {
	for (brandinfo = &linux_brandlist[0];
	*brandinfo != NULL; ++brandinfo)
	if (elf32_remove_brand_entry(*brandinfo) < 0)
	error = EINVAL;
	}
	if (error == 0) {
	SET_FOREACH(lihp, linux_ioctl_handler_set)
	linux_ioctl_unregister_handler(*lihp);
	SET_FOREACH(ldhp, linux_device_handler_set)
	linux_device_unregister_handler(*ldhp);
	mtx_destroy(&emul_lock);
	sx_destroy(&emul_shared_lock);
	mtx_destroy(&futex_mtx);
	EVENTHANDLER_DEREGISTER(process_exit, linux_exit_tag);
	EVENTHANDLER_DEREGISTER(schedtail, linux_schedtail_tag);
	EVENTHANDLER_DEREGISTER(process_exec, linux_exec_tag);
	linux_osd_jail_deregister();
	if (bootverbose)
	printf("Linux ELF exec handler removed\n");
	} else
	printf("Could not deinstall ELF interpreter entry\n");
	break;
	default:
	return EOPNOTSUPP;
	}
	return error;
	}

	static moduledata_t linux_elf_mod = {
	"linuxelf",
	linux_elf_modevent,
	0
	};

	DECLARE_MODULE_TIED(linuxelf, linux_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY);
	Index: stable/8/sys
	===================================================================
	--- stable/8/sys (revision 219197)
	+++ stable/8/sys (revision 219198)

	Property changes on: stable/8/sys
	___________________________________________________________________
	Modified: svn:mergeinfo
	## -0,0 +0,1 ##
	Merged /head/sys:r218658,218668

File Metadata

Mime Type: text/x-c
Expires: Sat, Oct 18, 3:35 AM (2 d)
Storage Engine: blob
Storage Format: Raw Data
Storage Handle: 23816105
Default Alt Text: (96 KB)

No OneTemporaryActions

View Options

File Metadata

Event Timeline

No OneTemporary
Actions