kernel - Move grow_stack code in fault path to improve fault performance

[dragonfly.git] / sys / platform / vkernel / i386 / trap.c

/*-
 * Copyright (C) 1994, David Greenman
 * Copyright (c) 1990, 1993
 *      The Regents of the University of California.  All rights reserved.
 *
 * This code is derived from software contributed to Berkeley by
 * the University of Utah, and William Jolitz.
 *
 * 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 the University of
 *      California, Berkeley and its contributors.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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.
 *
 *      from: @(#)trap.c        7.4 (Berkeley) 5/13/91
 * $FreeBSD: src/sys/i386/i386/trap.c,v 1.147.2.11 2003/02/27 19:09:59 luoqi Exp $
 * $DragonFly: src/sys/platform/vkernel/i386/trap.c,v 1.35 2008/09/09 04:06:19 dillon Exp $
 */

/*
 * 386 Trap and System call handling
 */

#include "use_isa.h"
#include "use_npx.h"

#include "opt_ddb.h"
#include "opt_ktrace.h"

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/proc.h>
#include <sys/pioctl.h>
#include <sys/kernel.h>
#include <sys/resourcevar.h>
#include <sys/signalvar.h>
#include <sys/signal2.h>
#include <sys/syscall.h>
#include <sys/sysctl.h>
#include <sys/sysent.h>
#include <sys/uio.h>
#include <sys/vmmeter.h>
#include <sys/malloc.h>
#ifdef KTRACE
#include <sys/ktrace.h>
#endif
#include <sys/ktr.h>
#include <sys/upcall.h>
#include <sys/vkernel.h>
#include <sys/sysproto.h>
#include <sys/sysunion.h>
#include <sys/vmspace.h>

#include <vm/vm.h>
#include <vm/vm_param.h>
#include <sys/lock.h>
#include <vm/pmap.h>
#include <vm/vm_kern.h>
#include <vm/vm_map.h>
#include <vm/vm_page.h>
#include <vm/vm_extern.h>

#include <machine/cpu.h>
#include <machine/md_var.h>
#include <machine/pcb.h>
#include <machine/smp.h>
#include <machine/tss.h>
#include <machine/globaldata.h>

#include <machine/vm86.h>

#include <ddb/ddb.h>

#include <sys/msgport2.h>
#include <sys/thread2.h>
#include <sys/mplock2.h>

#ifdef SMP

#define MAKEMPSAFE(have_mplock)                 \
        if (have_mplock == 0) {                 \
                get_mplock();                   \
                have_mplock = 1;                \
        }

#else

#define MAKEMPSAFE(have_mplock)

#endif

int (*pmath_emulate) (struct trapframe *);

static int trap_pfault (struct trapframe *, int, vm_offset_t);
static void trap_fatal (struct trapframe *, int, vm_offset_t);
void dblfault_handler (void);

#if 0
extern inthand_t IDTVEC(syscall);
#endif

#define MAX_TRAP_MSG            28
static char *trap_msg[] = {
        "",                                     /*  0 unused */
        "privileged instruction fault",         /*  1 T_PRIVINFLT */
        "",                                     /*  2 unused */
        "breakpoint instruction fault",         /*  3 T_BPTFLT */
        "",                                     /*  4 unused */
        "",                                     /*  5 unused */
        "arithmetic trap",                      /*  6 T_ARITHTRAP */
        "system forced exception",              /*  7 T_ASTFLT */
        "",                                     /*  8 unused */
        "general protection fault",             /*  9 T_PROTFLT */
        "trace trap",                           /* 10 T_TRCTRAP */
        "",                                     /* 11 unused */
        "page fault",                           /* 12 T_PAGEFLT */
        "",                                     /* 13 unused */
        "alignment fault",                      /* 14 T_ALIGNFLT */
        "",                                     /* 15 unused */
        "",                                     /* 16 unused */
        "",                                     /* 17 unused */
        "integer divide fault",                 /* 18 T_DIVIDE */
        "non-maskable interrupt trap",          /* 19 T_NMI */
        "overflow trap",                        /* 20 T_OFLOW */
        "FPU bounds check fault",               /* 21 T_BOUND */
        "FPU device not available",             /* 22 T_DNA */
        "double fault",                         /* 23 T_DOUBLEFLT */
        "FPU operand fetch fault",              /* 24 T_FPOPFLT */
        "invalid TSS fault",                    /* 25 T_TSSFLT */
        "segment not present fault",            /* 26 T_SEGNPFLT */
        "stack fault",                          /* 27 T_STKFLT */
        "machine check trap",                   /* 28 T_MCHK */
};

#ifdef DDB
static int ddb_on_nmi = 1;
SYSCTL_INT(_machdep, OID_AUTO, ddb_on_nmi, CTLFLAG_RW,
        &ddb_on_nmi, 0, "Go to DDB on NMI");
#endif
static int panic_on_nmi = 1;
SYSCTL_INT(_machdep, OID_AUTO, panic_on_nmi, CTLFLAG_RW,
        &panic_on_nmi, 0, "Panic on NMI");
static int fast_release;
SYSCTL_INT(_machdep, OID_AUTO, fast_release, CTLFLAG_RW,
        &fast_release, 0, "Passive Release was optimal");
static int slow_release;
SYSCTL_INT(_machdep, OID_AUTO, slow_release, CTLFLAG_RW,
        &slow_release, 0, "Passive Release was nonoptimal");
#ifdef SMP
static int syscall_mpsafe = 1;
SYSCTL_INT(_kern, OID_AUTO, syscall_mpsafe, CTLFLAG_RW,
        &syscall_mpsafe, 0, "Allow MPSAFE marked syscalls to run without BGL");
TUNABLE_INT("kern.syscall_mpsafe", &syscall_mpsafe);
static int trap_mpsafe = 1;
SYSCTL_INT(_kern, OID_AUTO, trap_mpsafe, CTLFLAG_RW,
        &trap_mpsafe, 0, "Allow traps to mostly run without the BGL");
TUNABLE_INT("kern.trap_mpsafe", &trap_mpsafe);
#endif

MALLOC_DEFINE(M_SYSMSG, "sysmsg", "sysmsg structure");
extern int max_sysmsg;

/*
 * Passively intercepts the thread switch function to increase
 * the thread priority from a user priority to a kernel priority, reducing
 * syscall and trap overhead for the case where no switch occurs.
 *
 * Synchronizes td_ucred with p_ucred.  This is used by system calls,
 * signal handling, faults, AST traps, and anything else that enters the
 * kernel from userland and provides the kernel with a stable read-only
 * copy of the process ucred.
 */
static __inline void
userenter(struct thread *curtd, struct proc *curp)
{
        struct ucred *ocred;
        struct ucred *ncred;

        curtd->td_release = lwkt_passive_release;

        if (curtd->td_ucred != curp->p_ucred) {
                ncred = crhold(curp->p_ucred);
                ocred = curtd->td_ucred;
                curtd->td_ucred = ncred;
                if (ocred)
                        crfree(ocred);
        }
}

/*
 * Handle signals, upcalls, profiling, and other AST's and/or tasks that
 * must be completed before we can return to or try to return to userland.
 *
 * Note that td_sticks is a 64 bit quantity, but there's no point doing 64
 * arithmatic on the delta calculation so the absolute tick values are
 * truncated to an integer.
 */
static void
userret(struct lwp *lp, struct trapframe *frame, int sticks)
{
        struct proc *p = lp->lwp_proc;
        int sig;

        /*
         * Charge system time if profiling.  Note: times are in microseconds.
         * This may do a copyout and block, so do it first even though it
         * means some system time will be charged as user time.
         */
        if (p->p_flag & P_PROFIL) {
                addupc_task(p, frame->tf_eip, 
                        (u_int)((int)lp->lwp_thread->td_sticks - sticks));
        }

recheck:
        /*
         * If the jungle wants us dead, so be it.
         */
        if (lp->lwp_flag & LWP_WEXIT) {
                get_mplock();
                lwp_exit(0);
                rel_mplock();   /* NOT REACHED */
        }

        /*
         * Block here if we are in a stopped state.
         */
        if (p->p_stat == SSTOP) {
                get_mplock();
                tstop();
                rel_mplock();
                goto recheck;
        }

        /*
         * Post any pending upcalls
         */
        if (p->p_flag & P_UPCALLPEND) {
                get_mplock();
                p->p_flag &= ~P_UPCALLPEND;
                postupcall(lp);
                rel_mplock();
                goto recheck;
        }

        /*
         * Post any pending signals
         */
        if ((sig = CURSIG_TRACE(lp)) != 0) {
                get_mplock();
                postsig(sig);
                rel_mplock();
                goto recheck;
        }

        /*
         * block here if we are swapped out, but still process signals
         * (such as SIGKILL).  proc0 (the swapin scheduler) is already
         * aware of our situation, we do not have to wake it up.
         */
        if (p->p_flag & P_SWAPPEDOUT) {
                get_mplock();
                p->p_flag |= P_SWAPWAIT;
                swapin_request();
                if (p->p_flag & P_SWAPWAIT)
                        tsleep(p, PCATCH, "SWOUT", 0);
                p->p_flag &= ~P_SWAPWAIT;
                rel_mplock();
                goto recheck;
        }

        /*
         * Make sure postsig() handled request to restore old signal mask after
         * running signal handler.
         */
        KKASSERT((lp->lwp_flag & LWP_OLDMASK) == 0);
}

/*
 * Cleanup from userenter and any passive release that might have occured.
 * We must reclaim the current-process designation before we can return
 * to usermode.  We also handle both LWKT and USER reschedule requests.
 */
static __inline void
userexit(struct lwp *lp)
{
        struct thread *td = lp->lwp_thread;
        /* globaldata_t gd = td->td_gd; */

        /*
         * Handle stop requests at kernel priority.  Any requests queued
         * after this loop will generate another AST.
         */
        while (lp->lwp_proc->p_stat == SSTOP) {
                get_mplock();
                tstop();
                rel_mplock();
        }

        /*
         * Reduce our priority in preparation for a return to userland.  If
         * our passive release function was still in place, our priority was
         * never raised and does not need to be reduced.
         */
        lwkt_passive_recover(td);

        /*
         * Become the current user scheduled process if we aren't already,
         * and deal with reschedule requests and other factors.
         */
        lp->lwp_proc->p_usched->acquire_curproc(lp);
        /* WARNING: we may have migrated cpu's */
        /* gd = td->td_gd; */
}

#if !defined(KTR_KERNENTRY)
#define KTR_KERNENTRY   KTR_ALL
#endif
KTR_INFO_MASTER(kernentry);
KTR_INFO(KTR_KERNENTRY, kernentry, trap, 0, "pid=%d, tid=%d, trapno=%d, eva=%p",
         sizeof(int) + sizeof(int) + sizeof(int) + sizeof(vm_offset_t));
KTR_INFO(KTR_KERNENTRY, kernentry, trap_ret, 0, "pid=%d, tid=%d",
         sizeof(int) + sizeof(int));
KTR_INFO(KTR_KERNENTRY, kernentry, syscall, 0, "pid=%d, tid=%d, call=%d",
         sizeof(int) + sizeof(int) + sizeof(int));
KTR_INFO(KTR_KERNENTRY, kernentry, syscall_ret, 0, "pid=%d, tid=%d, err=%d",
         sizeof(int) + sizeof(int) + sizeof(int));
KTR_INFO(KTR_KERNENTRY, kernentry, fork_ret, 0, "pid=%d, tid=%d",
         sizeof(int) + sizeof(int));

/*
 * Exception, fault, and trap interface to the kernel.
 * This common code is called from assembly language IDT gate entry
 * routines that prepare a suitable stack frame, and restore this
 * frame after the exception has been processed.
 *
 * This function is also called from doreti in an interlock to handle ASTs.
 * For example:  hardwareint->INTROUTINE->(set ast)->doreti->trap
 *
 * NOTE!  We have to retrieve the fault address prior to obtaining the
 * MP lock because get_mplock() may switch out.  YYY cr2 really ought
 * to be retrieved by the assembly code, not here.
 *
 * XXX gd_trap_nesting_level currently prevents lwkt_switch() from panicing
 * if an attempt is made to switch from a fast interrupt or IPI.  This is
 * necessary to properly take fatal kernel traps on SMP machines if 
 * get_mplock() has to block.
 */

void
user_trap(struct trapframe *frame)
{
        struct globaldata *gd = mycpu;
        struct thread *td = gd->gd_curthread;
        struct lwp *lp = td->td_lwp;
        struct proc *p;
        int sticks = 0;
        int i = 0, ucode = 0, type, code;
#ifdef SMP
        int have_mplock = 0;
#endif
#ifdef INVARIANTS
        int crit_count = td->td_pri & ~TDPRI_MASK;
#endif
        vm_offset_t eva;

        p = td->td_proc;

        /*
         * This is a bad kludge to avoid changing the various trapframe
         * structures.  Because we are enabled as a virtual kernel,
         * the original tf_err field will be passed to us shifted 16
         * over in the tf_trapno field for T_PAGEFLT.
         */
        if (frame->tf_trapno == T_PAGEFLT)
                eva = frame->tf_err;
        else
                eva = 0;
#if 0
        kprintf("USER_TRAP AT %08x xflags %d trapno %d eva %08x\n", 
                frame->tf_eip, frame->tf_xflags, frame->tf_trapno, eva);
#endif

        /*
         * Everything coming from user mode runs through user_trap,
         * including system calls.
         */
        if (frame->tf_trapno == T_SYSCALL80) {
                syscall2(frame);
                return;
        }

        KTR_LOG(kernentry_trap, lp->lwp_proc->p_pid, lp->lwp_tid,
                frame->tf_trapno, eva);

#ifdef DDB
        if (db_active) {
                eva = (frame->tf_trapno == T_PAGEFLT ? rcr2() : 0);
                ++gd->gd_trap_nesting_level;
                MAKEMPSAFE(have_mplock);
                trap_fatal(frame, TRUE, eva);
                --gd->gd_trap_nesting_level;
                goto out2;
        }
#endif

        ++gd->gd_trap_nesting_level;
#ifdef SMP
        if (trap_mpsafe == 0)
                MAKEMPSAFE(have_mplock);
#endif

        --gd->gd_trap_nesting_level;

#if defined(I586_CPU) && !defined(NO_F00F_HACK)
restart:
#endif
        type = frame->tf_trapno;
        code = frame->tf_err;

        userenter(td, p);

        sticks = (int)td->td_sticks;
        lp->lwp_md.md_regs = frame;

        switch (type) {
        case T_PRIVINFLT:       /* privileged instruction fault */
                ucode = ILL_PRVOPC;
                i = SIGILL;
                break;

        case T_BPTFLT:          /* bpt instruction fault */
        case T_TRCTRAP:         /* trace trap */
                frame->tf_eflags &= ~PSL_T;
                ucode = TRAP_TRACE;
                i = SIGTRAP;
                break;

        case T_ARITHTRAP:       /* arithmetic trap */
                ucode = code;
                i = SIGFPE;
                break;

        case T_ASTFLT:          /* Allow process switch */
                mycpu->gd_cnt.v_soft++;
                if (mycpu->gd_reqflags & RQF_AST_OWEUPC) {
                        atomic_clear_int_nonlocked(&mycpu->gd_reqflags,
                                    RQF_AST_OWEUPC);
                        addupc_task(p, p->p_prof.pr_addr,
                                    p->p_prof.pr_ticks);
                }
                goto out;

                /*
                 * The following two traps can happen in
                 * vm86 mode, and, if so, we want to handle
                 * them specially.
                 */
        case T_PROTFLT:         /* general protection fault */
        case T_STKFLT:          /* stack fault */
#if 0
                if (frame->tf_eflags & PSL_VM) {
                        i = vm86_emulate((struct vm86frame *)frame);
                        if (i == 0)
                                goto out;
                        break;
                }
#endif
                i = SIGBUS;
                ucode = (type == T_PROTFLT) ? BUS_OBJERR : BUS_ADRERR;
                break;
        case T_SEGNPFLT:        /* segment not present fault */
                i = SIGBUS;
                ucode = BUS_ADRERR;
                break;
        case T_TSSFLT:          /* invalid TSS fault */
        case T_DOUBLEFLT:       /* double fault */
                i = SIGBUS;
                ucode = BUS_OBJERR;
        default:
#if 0
                ucode = code + BUS_SEGM_FAULT ; /* XXX: ???*/
#endif
                ucode = BUS_OBJERR;
                i = SIGBUS;
                break;

        case T_PAGEFLT:         /* page fault */
                MAKEMPSAFE(have_mplock);
                i = trap_pfault(frame, TRUE, eva);
                if (i == -1)
                        goto out;
#if defined(I586_CPU) && !defined(NO_F00F_HACK)
                if (i == -2)
                        goto restart;
#endif
                if (i == 0)
                        goto out;

#if 0
                ucode = T_PAGEFLT;
#endif
                if (i == SIGSEGV)
                        ucode = SEGV_MAPERR;
                else
                        ucode = BUS_ADRERR;
                break;

        case T_DIVIDE:          /* integer divide fault */
                ucode = FPE_INTDIV;
                i = SIGFPE;
                break;

#if NISA > 0
        case T_NMI:
                MAKEMPSAFE(have_mplock);
                /* machine/parity/power fail/"kitchen sink" faults */
                if (isa_nmi(code) == 0) {
#ifdef DDB
                        /*
                         * NMI can be hooked up to a pushbutton
                         * for debugging.
                         */
                        if (ddb_on_nmi) {
                                kprintf ("NMI ... going to debugger\n");
                                kdb_trap (type, 0, frame);
                        }
#endif /* DDB */
                        goto out2;
                } else if (panic_on_nmi)
                        panic("NMI indicates hardware failure");
                break;
#endif /* NISA > 0 */

        case T_OFLOW:           /* integer overflow fault */
                ucode = FPE_INTOVF;
                i = SIGFPE;
                break;

        case T_BOUND:           /* bounds check fault */
                ucode = FPE_FLTSUB;
                i = SIGFPE;
                break;

        case T_DNA:
                /*
                 * Virtual kernel intercept - pass the DNA exception
                 * to the (emulated) virtual kernel if it asked to handle 
                 * it.  This occurs when the virtual kernel is holding
                 * onto the FP context for a different emulated
                 * process then the one currently running.
                 *
                 * We must still call npxdna() since we may have
                 * saved FP state that the (emulated) virtual kernel
                 * needs to hand over to a different emulated process.
                 */
                if (lp->lwp_vkernel && lp->lwp_vkernel->ve &&
                    (td->td_pcb->pcb_flags & FP_VIRTFP)
                ) {
                        npxdna(frame);
                        break;
                }
#if NNPX > 0
                /* 
                 * The kernel may have switched out the FP unit's
                 * state, causing the user process to take a fault
                 * when it tries to use the FP unit.  Restore the
                 * state here
                 */
                if (npxdna(frame))
                        goto out;
#endif
                if (!pmath_emulate) {
                        i = SIGFPE;
                        ucode = FPE_FPU_NP_TRAP;
                        break;
                }
                i = (*pmath_emulate)(frame);
                if (i == 0) {
                        if (!(frame->tf_eflags & PSL_T))
                                goto out2;
                        frame->tf_eflags &= ~PSL_T;
                        i = SIGTRAP;
                }
                /* else ucode = emulator_only_knows() XXX */
                break;

        case T_FPOPFLT:         /* FPU operand fetch fault */
                ucode = ILL_COPROC;
                i = SIGILL;
                break;

        case T_XMMFLT:          /* SIMD floating-point exception */
                ucode = 0; /* XXX */
                i = SIGFPE;
                break;
        }

        /*
         * Virtual kernel intercept - if the fault is directly related to a
         * VM context managed by a virtual kernel then let the virtual kernel
         * handle it.
         */
        if (lp->lwp_vkernel && lp->lwp_vkernel->ve) {
                vkernel_trap(lp, frame);
                goto out;
        }

        /*
         * Translate fault for emulators (e.g. Linux) 
         */
        if (*p->p_sysent->sv_transtrap)
                i = (*p->p_sysent->sv_transtrap)(i, type);

        MAKEMPSAFE(have_mplock);
        trapsignal(lp, i, ucode);

#ifdef DEBUG
        if (type <= MAX_TRAP_MSG) {
                uprintf("fatal process exception: %s",
                        trap_msg[type]);
                if ((type == T_PAGEFLT) || (type == T_PROTFLT))
                        uprintf(", fault VA = 0x%lx", (u_long)eva);
                uprintf("\n");
        }
#endif

out:
#ifdef SMP
        KASSERT(td->td_mpcount == have_mplock, ("badmpcount trap/end from %p", (void *)frame->tf_eip));
#endif
        userret(lp, frame, sticks);
        userexit(lp);
out2:   ;
#ifdef SMP
        if (have_mplock)
                rel_mplock();
#endif
        KTR_LOG(kernentry_trap_ret, lp->lwp_proc->p_pid, lp->lwp_tid);
#ifdef INVARIANTS
        KASSERT(crit_count == (td->td_pri & ~TDPRI_MASK),
                ("syscall: critical section count mismatch! %d/%d",
                crit_count / TDPRI_CRIT, td->td_pri / TDPRI_CRIT));
#endif
}

void
kern_trap(struct trapframe *frame)
{
        struct globaldata *gd = mycpu;
        struct thread *td = gd->gd_curthread;
        struct lwp *lp;
        struct proc *p;
        int i = 0, ucode = 0, type, code;
#ifdef SMP
        int have_mplock = 0;
#endif
#ifdef INVARIANTS
        int crit_count = td->td_pri & ~TDPRI_MASK;
#endif
        vm_offset_t eva;

        lp = td->td_lwp;
        p = td->td_proc;

        if (frame->tf_trapno == T_PAGEFLT) 
                eva = frame->tf_err;
        else
                eva = 0;

#ifdef DDB
        if (db_active) {
                ++gd->gd_trap_nesting_level;
                MAKEMPSAFE(have_mplock);
                trap_fatal(frame, FALSE, eva);
                --gd->gd_trap_nesting_level;
                goto out2;
        }
#endif

        ++gd->gd_trap_nesting_level;

#ifdef SMP
        if (trap_mpsafe == 0)
                MAKEMPSAFE(have_mplock);
#endif

        --gd->gd_trap_nesting_level;

        type = frame->tf_trapno;
        code = frame->tf_err;

#if 0
kernel_trap:
#endif
        /* kernel trap */

        switch (type) {
        case T_PAGEFLT:                 /* page fault */
                MAKEMPSAFE(have_mplock);
                trap_pfault(frame, FALSE, eva);
                goto out2;

        case T_DNA:
#if NNPX > 0
                /*
                 * The kernel may be using npx for copying or other
                 * purposes.
                 */
                panic("kernel NPX should not happen");
                if (npxdna(frame))
                        goto out2;
#endif
                break;

        case T_PROTFLT:         /* general protection fault */
        case T_SEGNPFLT:        /* segment not present fault */
                /*
                 * Invalid segment selectors and out of bounds
                 * %eip's and %esp's can be set up in user mode.
                 * This causes a fault in kernel mode when the
                 * kernel tries to return to user mode.  We want
                 * to get this fault so that we can fix the
                 * problem here and not have to check all the
                 * selectors and pointers when the user changes
                 * them.
                 */
                if (mycpu->gd_intr_nesting_level == 0) {
                        if (td->td_pcb->pcb_onfault) {
                                frame->tf_eip = 
                                    (register_t)td->td_pcb->pcb_onfault;
                                goto out2;
                        }
                }
                break;

        case T_TSSFLT:
                /*
                 * PSL_NT can be set in user mode and isn't cleared
                 * automatically when the kernel is entered.  This
                 * causes a TSS fault when the kernel attempts to
                 * `iret' because the TSS link is uninitialized.  We
                 * want to get this fault so that we can fix the
                 * problem here and not every time the kernel is
                 * entered.
                 */
                if (frame->tf_eflags & PSL_NT) {
                        frame->tf_eflags &= ~PSL_NT;
                        goto out2;
                }
                break;

        case T_TRCTRAP:  /* trace trap */
#if 0
                if (frame->tf_eip == (int)IDTVEC(syscall)) {
                        /*
                         * We've just entered system mode via the
                         * syscall lcall.  Continue single stepping
                         * silently until the syscall handler has
                         * saved the flags.
                         */
                        goto out2;
                }
                if (frame->tf_eip == (int)IDTVEC(syscall) + 1) {
                        /*
                         * The syscall handler has now saved the
                         * flags.  Stop single stepping it.
                         */
                        frame->tf_eflags &= ~PSL_T;
                        goto out2;
                }
#endif
#if 0
                /*
                 * Ignore debug register trace traps due to
                 * accesses in the user's address space, which
                 * can happen under several conditions such as
                 * if a user sets a watchpoint on a buffer and
                 * then passes that buffer to a system call.
                 * We still want to get TRCTRAPS for addresses
                 * in kernel space because that is useful when
                 * debugging the kernel.
                 */
                if (user_dbreg_trap()) {
                        /*
                         * Reset breakpoint bits because the
                         * processor doesn't
                         */
                        load_dr6(rdr6() & 0xfffffff0);
                        goto out2;
                }
#endif
                /*
                 * Fall through (TRCTRAP kernel mode, kernel address)
                 */
        case T_BPTFLT:
                /*
                 * If DDB is enabled, let it handle the debugger trap.
                 * Otherwise, debugger traps "can't happen".
                 */
#ifdef DDB
                MAKEMPSAFE(have_mplock);
                if (kdb_trap (type, 0, frame))
                        goto out2;
#endif
                break;
        case T_DIVIDE:
                MAKEMPSAFE(have_mplock);
                trap_fatal(frame, FALSE, eva);
                goto out2;
        case T_NMI:
                MAKEMPSAFE(have_mplock);
                trap_fatal(frame, FALSE, eva);
                goto out2;
        case T_SYSCALL80:
                /*
                 * Ignore this trap generated from a spurious SIGTRAP.
                 *
                 * single stepping in / syscalls leads to spurious / SIGTRAP
                 * so ignore
                 *
                 * Haiku (c) 2007 Simon 'corecode' Schubert
                 */
                goto out2;
        }

        /*
         * Translate fault for emulators (e.g. Linux) 
         */
        if (*p->p_sysent->sv_transtrap)
                i = (*p->p_sysent->sv_transtrap)(i, type);

        MAKEMPSAFE(have_mplock);
        trapsignal(lp, i, ucode);

#ifdef DEBUG
        if (type <= MAX_TRAP_MSG) {
                uprintf("fatal process exception: %s",
                        trap_msg[type]);
                if ((type == T_PAGEFLT) || (type == T_PROTFLT))
                        uprintf(", fault VA = 0x%lx", (u_long)eva);
                uprintf("\n");
        }
#endif

out2:   
        ;
#ifdef SMP
        if (have_mplock)
                rel_mplock();
#endif
#ifdef INVARIANTS
        KASSERT(crit_count == (td->td_pri & ~TDPRI_MASK),
                ("syscall: critical section count mismatch! %d/%d",
                crit_count / TDPRI_CRIT, td->td_pri / TDPRI_CRIT));
#endif
}

int
trap_pfault(struct trapframe *frame, int usermode, vm_offset_t eva)
{
        vm_offset_t va;
        struct vmspace *vm = NULL;
        vm_map_t map = 0;
        int rv = 0;
        int fault_flags;
        vm_prot_t ftype;
        thread_t td = curthread;
        struct lwp *lp = td->td_lwp;

        va = trunc_page(eva);
        if (usermode == FALSE) {
                /*
                 * This is a fault on kernel virtual memory.
                 */
                map = &kernel_map;
        } else {
                /*
                 * This is a fault on non-kernel virtual memory.
                 * vm is initialized above to NULL. If curproc is NULL
                 * or curproc->p_vmspace is NULL the fault is fatal.
                 */
                if (lp != NULL)
                        vm = lp->lwp_vmspace;

                if (vm == NULL)
                        goto nogo;

                map = &vm->vm_map;
        }

        if (frame->tf_xflags & PGEX_W)
                ftype = VM_PROT_READ | VM_PROT_WRITE;
        else
                ftype = VM_PROT_READ;

        if (map != &kernel_map) {
                /*
                 * Keep swapout from messing with us during this
                 *      critical time.
                 */
                PHOLD(lp->lwp_proc);

                /*
                 * Issue fault
                 */
                fault_flags = 0;
                if (usermode)
                        fault_flags |= VM_FAULT_BURST;
                if (ftype & VM_PROT_WRITE)
                        fault_flags |= VM_FAULT_DIRTY;
                else
                        fault_flags |= VM_FAULT_NORMAL;
                rv = vm_fault(map, va, ftype, fault_flags);

                PRELE(lp->lwp_proc);
        } else {
                /*
                 * Don't have to worry about process locking or stacks in the kernel.
                 */
                rv = vm_fault(map, va, ftype, VM_FAULT_NORMAL);
        }

        if (rv == KERN_SUCCESS)
                return (0);
nogo:
        if (!usermode) {
                if (td->td_gd->gd_intr_nesting_level == 0 &&
                    td->td_pcb->pcb_onfault) {
                        frame->tf_eip = (register_t)td->td_pcb->pcb_onfault;
                        return (0);
                }
                trap_fatal(frame, usermode, eva);
                return (-1);
        }
        return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV);
}

static void
trap_fatal(struct trapframe *frame, int usermode, vm_offset_t eva)
{
        int code, type, ss, esp;

        code = frame->tf_xflags;
        type = frame->tf_trapno;

        if (type <= MAX_TRAP_MSG) {
                kprintf("\n\nFatal trap %d: %s while in %s mode\n",
                        type, trap_msg[type],
                        (usermode ? "user" : "kernel"));
        }
#ifdef SMP
        /* two separate prints in case of a trap on an unmapped page */
        kprintf("mp_lock = %08x; ", mp_lock);
        kprintf("cpuid = %d\n", mycpu->gd_cpuid);
#endif
        if (type == T_PAGEFLT) {
                kprintf("fault virtual address  = %p\n", (void *)eva);
                kprintf("fault code             = %s %s, %s\n",
                        usermode ? "user" : "supervisor",
                        code & PGEX_W ? "write" : "read",
                        code & PGEX_P ? "protection violation" : "page not present");
        }
        kprintf("instruction pointer    = 0x%x:0x%x\n",
               frame->tf_cs & 0xffff, frame->tf_eip);
        if (usermode) {
                ss = frame->tf_ss & 0xffff;
                esp = frame->tf_esp;
        } else {
                ss = GSEL(GDATA_SEL, SEL_KPL);
                esp = (int)&frame->tf_esp;
        }
        kprintf("stack pointer          = 0x%x:0x%x\n", ss, esp);
        kprintf("frame pointer          = 0x%x:0x%x\n", ss, frame->tf_ebp);
        kprintf("processor eflags       = ");
        if (frame->tf_eflags & PSL_T)
                kprintf("trace trap, ");
        if (frame->tf_eflags & PSL_I)
                kprintf("interrupt enabled, ");
        if (frame->tf_eflags & PSL_NT)
                kprintf("nested task, ");
        if (frame->tf_eflags & PSL_RF)
                kprintf("resume, ");
#if 0
        if (frame->tf_eflags & PSL_VM)
                kprintf("vm86, ");
#endif
        kprintf("IOPL = %d\n", (frame->tf_eflags & PSL_IOPL) >> 12);
        kprintf("current process                = ");
        if (curproc) {
                kprintf("%lu (%s)\n",
                    (u_long)curproc->p_pid, curproc->p_comm ?
                    curproc->p_comm : "");
        } else {
                kprintf("Idle\n");
        }
        kprintf("current thread          = pri %d ", curthread->td_pri);
        if (curthread->td_pri >= TDPRI_CRIT)
                kprintf("(CRIT)");
        kprintf("\n");
#ifdef SMP
/**
 *  XXX FIXME:
 *      we probably SHOULD have stopped the other CPUs before now!
 *      another CPU COULD have been touching cpl at this moment...
 */
        kprintf(" <- SMP: XXX");
#endif
        kprintf("\n");

#ifdef KDB
        if (kdb_trap(&psl))
                return;
#endif
#ifdef DDB
        if ((debugger_on_panic || db_active) && kdb_trap(type, code, frame))
                return;
#endif
        kprintf("trap number            = %d\n", type);
        if (type <= MAX_TRAP_MSG)
                panic("%s", trap_msg[type]);
        else
                panic("unknown/reserved trap");
}

/*
 * Double fault handler. Called when a fault occurs while writing
 * a frame for a trap/exception onto the stack. This usually occurs
 * when the stack overflows (such is the case with infinite recursion,
 * for example).
 *
 * XXX Note that the current PTD gets replaced by IdlePTD when the
 * task switch occurs. This means that the stack that was active at
 * the time of the double fault is not available at <kstack> unless
 * the machine was idle when the double fault occurred. The downside
 * of this is that "trace <ebp>" in ddb won't work.
 */
void
dblfault_handler(void)
{
        struct mdglobaldata *gd = mdcpu;

        kprintf("\nFatal double fault:\n");
        kprintf("eip = 0x%x\n", gd->gd_common_tss.tss_eip);
        kprintf("esp = 0x%x\n", gd->gd_common_tss.tss_esp);
        kprintf("ebp = 0x%x\n", gd->gd_common_tss.tss_ebp);
#ifdef SMP
        /* two separate prints in case of a trap on an unmapped page */
        kprintf("mp_lock = %08x; ", mp_lock);
        kprintf("cpuid = %d\n", mycpu->gd_cpuid);
#endif
        panic("double fault");
}

/*
 * syscall2 -   MP aware system call request C handler
 *
 * A system call is essentially treated as a trap except that the
 * MP lock is not held on entry or return.  We are responsible for
 * obtaining the MP lock if necessary and for handling ASTs
 * (e.g. a task switch) prior to return.
 *
 * MPSAFE
 */
void
syscall2(struct trapframe *frame)
{
        struct thread *td = curthread;
        struct proc *p = td->td_proc;
        struct lwp *lp = td->td_lwp;
        caddr_t params;
        struct sysent *callp;
        register_t orig_tf_eflags;
        int sticks;
        int error;
        int narg;
#ifdef INVARIANTS
        int crit_count = td->td_pri & ~TDPRI_MASK;
#endif
#ifdef SMP
        int have_mplock = 0;
#endif
        u_int code;
        union sysunion args;

        KTR_LOG(kernentry_syscall, lp->lwp_proc->p_pid, lp->lwp_tid,
                frame->tf_eax);

#ifdef SMP
        KASSERT(td->td_mpcount == 0, ("badmpcount syscall2 from %p", (void *)frame->tf_eip));
        if (syscall_mpsafe == 0)
                MAKEMPSAFE(have_mplock);
#endif
        userenter(td, p);       /* lazy raise our priority */

        /*
         * Misc
         */
        sticks = (int)td->td_sticks;
        orig_tf_eflags = frame->tf_eflags;

        /*
         * Virtual kernel intercept - if a VM context managed by a virtual
         * kernel issues a system call the virtual kernel handles it, not us.
         * Restore the virtual kernel context and return from its system
         * call.  The current frame is copied out to the virtual kernel.
         */
        if (lp->lwp_vkernel && lp->lwp_vkernel->ve) {
                vkernel_trap(lp, frame);
                error = EJUSTRETURN;
                goto out;
        }

        /*
         * Get the system call parameters and account for time
         */
        lp->lwp_md.md_regs = frame;
        params = (caddr_t)frame->tf_esp + sizeof(int);
        code = frame->tf_eax;

        if (p->p_sysent->sv_prepsyscall) {
                (*p->p_sysent->sv_prepsyscall)(
                        frame, (int *)(&args.nosys.sysmsg + 1),
                        &code, &params);
        } else {
                /*
                 * Need to check if this is a 32 bit or 64 bit syscall.
                 * fuword is MP aware.
                 */
                if (code == SYS_syscall) {
                        /*
                         * Code is first argument, followed by actual args.
                         */
                        code = fuword(params);
                        params += sizeof(int);
                } else if (code == SYS___syscall) {
                        /*
                         * Like syscall, but code is a quad, so as to maintain
                         * quad alignment for the rest of the arguments.
                         */
                        code = fuword(params);
                        params += sizeof(quad_t);
                }
        }

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

        narg = callp->sy_narg & SYF_ARGMASK;

        /*
         * copyin is MP aware, but the tracing code is not
         */
        if (narg && params) {
                error = copyin(params, (caddr_t)(&args.nosys.sysmsg + 1),
                                narg * sizeof(register_t));
                if (error) {
#ifdef KTRACE
                        if (KTRPOINT(td, KTR_SYSCALL)) {
                                MAKEMPSAFE(have_mplock);
                                
                                ktrsyscall(lp, code, narg,
                                        (void *)(&args.nosys.sysmsg + 1));
                        }
#endif
                        goto bad;
                }
        }

#ifdef KTRACE
        if (KTRPOINT(td, KTR_SYSCALL)) {
                MAKEMPSAFE(have_mplock);
                ktrsyscall(lp, code, narg, (void *)(&args.nosys.sysmsg + 1));
        }
#endif

        /*
         * For traditional syscall code edx is left untouched when 32 bit
         * results are returned.  Since edx is loaded from fds[1] when the 
         * system call returns we pre-set it here.
         */
        args.sysmsg_fds[0] = 0;
        args.sysmsg_fds[1] = frame->tf_edx;

        /*
         * The syscall might manipulate the trap frame. If it does it
         * will probably return EJUSTRETURN.
         */
        args.sysmsg_frame = frame;

        STOPEVENT(p, S_SCE, narg);      /* MP aware */

        /*
         * NOTE: All system calls run MPSAFE now.  The system call itself
         *       is responsible for getting the MP lock.
         */
        error = (*callp->sy_call)(&args);

#if 0
        kprintf("system call %d returned %d\n", code, error);
#endif

out:
        /*
         * MP SAFE (we may or may not have the MP lock at this point)
         */
        switch (error) {
        case 0:
                /*
                 * Reinitialize proc pointer `p' as it may be different
                 * if this is a child returning from fork syscall.
                 */
                p = curproc;
                lp = curthread->td_lwp;
                frame->tf_eax = args.sysmsg_fds[0];
                frame->tf_edx = args.sysmsg_fds[1];
                frame->tf_eflags &= ~PSL_C;
                break;
        case ERESTART:
                /*
                 * Reconstruct pc, assuming lcall $X,y is 7 bytes,
                 * int 0x80 is 2 bytes. We saved this in tf_err.
                 */
                frame->tf_eip -= frame->tf_err;
                break;
        case EJUSTRETURN:
                break;
        case EASYNC:
                panic("Unexpected EASYNC return value (for now)");
        default:
bad:
                if (p->p_sysent->sv_errsize) {
                        if (error >= p->p_sysent->sv_errsize)
                                error = -1;     /* XXX */
                        else
                                error = p->p_sysent->sv_errtbl[error];
                }
                frame->tf_eax = error;
                frame->tf_eflags |= PSL_C;
                break;
        }

        /*
         * Traced syscall.  trapsignal() is not MP aware.
         */
        if ((orig_tf_eflags & PSL_T) /*&& !(orig_tf_eflags & PSL_VM)*/) {
                MAKEMPSAFE(have_mplock);
                frame->tf_eflags &= ~PSL_T;
                trapsignal(lp, SIGTRAP, TRAP_TRACE);
        }

        /*
         * Handle reschedule and other end-of-syscall issues
         */
        userret(lp, frame, sticks);

#ifdef KTRACE
        if (KTRPOINT(td, KTR_SYSRET)) {
                MAKEMPSAFE(have_mplock);
                ktrsysret(lp, code, error, args.sysmsg_result);
        }
#endif

        /*
         * This works because errno is findable through the
         * register set.  If we ever support an emulation where this
         * is not the case, this code will need to be revisited.
         */
        STOPEVENT(p, S_SCX, code);

        userexit(lp);
#ifdef SMP
        /*
         * Release the MP lock if we had to get it
         */
        KASSERT(td->td_mpcount == have_mplock, 
                ("badmpcount syscall2/end from %p", (void *)frame->tf_eip));
        if (have_mplock)
                rel_mplock();
#endif
        KTR_LOG(kernentry_syscall_ret, lp->lwp_proc->p_pid, lp->lwp_tid, error);
#ifdef INVARIANTS
        KASSERT(crit_count == (td->td_pri & ~TDPRI_MASK), 
                ("syscall: critical section count mismatch! %d/%d",
                crit_count / TDPRI_CRIT, td->td_pri / TDPRI_CRIT));
#endif
}

void
fork_return(struct lwp *lp, struct trapframe *frame)
{
        frame->tf_eax = 0;              /* Child returns zero */
        frame->tf_eflags &= ~PSL_C;     /* success */
        frame->tf_edx = 1;

        generic_lwp_return(lp, frame);
        KTR_LOG(kernentry_fork_ret, lp->lwp_proc->p_pid, lp->lwp_tid);
}

/*
 * Simplified back end of syscall(), used when returning from fork()
 * or lwp_create() directly into user mode.  MP lock is held on entry and
 * should be released on return.  This code will return back into the fork
 * trampoline code which then runs doreti.
 */
void
generic_lwp_return(struct lwp *lp, struct trapframe *frame)
{
        struct proc *p = lp->lwp_proc;

        /*
         * Newly forked processes are given a kernel priority.  We have to
         * adjust the priority to a normal user priority and fake entry
         * into the kernel (call userenter()) to install a passive release
         * function just in case userret() decides to stop the process.  This
         * can occur when ^Z races a fork.  If we do not install the passive
         * release function the current process designation will not be
         * released when the thread goes to sleep.
         */
        lwkt_setpri_self(TDPRI_USER_NORM);
        userenter(lp->lwp_thread, p);
        userret(lp, frame, 0);
#ifdef KTRACE
        if (KTRPOINT(lp->lwp_thread, KTR_SYSRET))
                ktrsysret(lp, SYS_fork, 0, 0);
#endif
        p->p_flag |= P_PASSIVE_ACQ;
        userexit(lp);
        p->p_flag &= ~P_PASSIVE_ACQ;
#ifdef SMP
        KKASSERT(lp->lwp_thread->td_mpcount == 1);
        rel_mplock();
#endif
}

/*
 * doreti has turned into this.  The frame is directly on the stack.  We
 * pull everything else we need (fpu and tls context) from the current
 * thread.
 *
 * Note on fpu interactions: In a virtual kernel, the fpu context for
 * an emulated user mode process is not shared with the virtual kernel's
 * fpu context, so we only have to 'stack' fpu contexts within the virtual
 * kernel itself, and not even then since the signal() contexts that we care
 * about save and restore the FPU state (I think anyhow).
 *
 * vmspace_ctl() returns an error only if it had problems instaling the
 * context we supplied or problems copying data to/from our VM space.
 */
void
go_user(struct intrframe *frame)
{
        struct trapframe *tf = (void *)&frame->if_gs;
        int r;

        /*
         * Interrupts may be disabled on entry, make sure all signals
         * can be received before beginning our loop.
         */
        sigsetmask(0);

        /*
         * Switch to the current simulated user process, then call
         * user_trap() when we break out of it (usually due to a signal).
         */
        for (;;) {
                /*
                 * Tell the real kernel whether it is ok to use the FP
                 * unit or not.
                 *
                 * The critical section is required to prevent an interrupt
                 * from causing a preemptive task switch and changing
                 * the FP state.
                 */
                crit_enter();
                if (mdcpu->gd_npxthread == curthread) {
                        tf->tf_xflags &= ~PGEX_FPFAULT;
                } else {
                        tf->tf_xflags |= PGEX_FPFAULT;
                }

                /*
                 * Run emulated user process context.  This call interlocks
                 * with new mailbox signals.
                 *
                 * Set PGEX_U unconditionally, indicating a user frame (the
                 * bit is normally set only by T_PAGEFLT).
                 */
                r = vmspace_ctl(&curproc->p_vmspace->vm_pmap, VMSPACE_CTL_RUN,
                                tf, &curthread->td_savevext);
                crit_exit();
                frame->if_xflags |= PGEX_U;
#if 0
                kprintf("GO USER %d trap %d EVA %08x EIP %08x ESP %08x XFLAGS %02x/%02x\n", 
                        r, tf->tf_trapno, tf->tf_err, tf->tf_eip, tf->tf_esp,
                        tf->tf_xflags, frame->if_xflags);
#endif
                if (r < 0) {
                        if (errno != EINTR)
                                panic("vmspace_ctl failed error %d", errno);
                } else {
                        if (tf->tf_trapno) {
                                user_trap(tf);
                        }
                }
                if (mycpu->gd_reqflags & RQF_AST_MASK) {
                        tf->tf_trapno = T_ASTFLT;
                        user_trap(tf);
                }
                tf->tf_trapno = 0;
        }
}

/*
 * If PGEX_FPFAULT is set then set FP_VIRTFP in the PCB to force a T_DNA
 * fault (which is then passed back to the virtual kernel) if an attempt is
 * made to use the FP unit.
 * 
 * XXX this is a fairly big hack.
 */
void
set_vkernel_fp(struct trapframe *frame)
{
        struct thread *td = curthread;

        if (frame->tf_xflags & PGEX_FPFAULT) {
                td->td_pcb->pcb_flags |= FP_VIRTFP;
                if (mdcpu->gd_npxthread == td)
                        npxexit();
        } else {
                td->td_pcb->pcb_flags &= ~FP_VIRTFP;
        }
}

/*
 * Called from vkernel_trap() to fixup the vkernel's syscall
 * frame for vmspace_ctl() return.
 */
void
cpu_vkernel_trap(struct trapframe *frame, int error)
{
        frame->tf_eax = error;
        if (error)
                frame->tf_eflags |= PSL_C;
        else
                frame->tf_eflags &= ~PSL_C;
}