modules/unix/unix.c

/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */
/*
 * Copyright (c) 1999, 2010, Oracle and/or its affiliates. All rights reserved.
 * Copyright 2018 OmniOS Community Edition (OmniOSce) Association.
 * Copyright 2019 Joyent, Inc.
 */

#include <mdb/mdb_modapi.h>
#include <mdb/mdb_ctf.h>
#include <mdb/mdb_x86util.h>
#include <sys/cpuvar.h>
#include <sys/systm.h>
#include <sys/traptrace.h>
#include <sys/x_call.h>
#include <sys/xc_levels.h>
#include <sys/avintr.h>
#include <sys/systm.h>
#include <sys/trap.h>
#include <sys/mutex.h>
#include <sys/mutex_impl.h>
#include "i86mmu.h"
#include "unix_sup.h"
#include <sys/apix.h>
#include <sys/x86_archext.h>
#include <sys/bitmap.h>
#include <sys/controlregs.h>

#define	TT_HDLR_WIDTH	17


/* apix only */
static apix_impl_t *d_apixs[NCPU];
static int use_apix = 0;

static int
ttrace_ttr_size_check(void)
{
	mdb_ctf_id_t ttrtid;
	ssize_t ttr_size;

	if (mdb_ctf_lookup_by_name("trap_trace_rec_t", &ttrtid) != 0 ||
	    mdb_ctf_type_resolve(ttrtid, &ttrtid) != 0) {
		mdb_warn("failed to determine size of trap_trace_rec_t; "
		    "non-TRAPTRACE kernel?\n");
		return (0);
	}

	if ((ttr_size = mdb_ctf_type_size(ttrtid)) !=
	    sizeof (trap_trace_rec_t)) {
		/*
		 * On Intel machines, this will happen when TTR_STACK_DEPTH
		 * is changed.  This code could be smarter, and could
		 * dynamically adapt to different depths, but not until a
		 * need for such adaptation is demonstrated.
		 */
		mdb_warn("size of trap_trace_rec_t (%d bytes) doesn't "
		    "match expected %d\n", ttr_size, sizeof (trap_trace_rec_t));
		return (0);
	}

	return (1);
}

int
ttrace_walk_init(mdb_walk_state_t *wsp)
{
	trap_trace_ctl_t *ttcp;
	size_t ttc_size = sizeof (trap_trace_ctl_t) * NCPU;
	int i;

	if (!ttrace_ttr_size_check())
		return (WALK_ERR);

	ttcp = mdb_zalloc(ttc_size, UM_SLEEP);

	if (wsp->walk_addr != 0) {
		mdb_warn("ttrace only supports global walks\n");
		return (WALK_ERR);
	}

	if (mdb_readsym(ttcp, ttc_size, "trap_trace_ctl") == -1) {
		mdb_warn("symbol 'trap_trace_ctl' not found; "
		    "non-TRAPTRACE kernel?\n");
		mdb_free(ttcp, ttc_size);
		return (WALK_ERR);
	}

	/*
	 * We'll poach the ttc_current pointer (which isn't used for
	 * anything) to store a pointer to our current TRAPTRACE record.
	 * This allows us to only keep the array of trap_trace_ctl structures
	 * as our walker state (ttc_current may be the only kernel data
	 * structure member added exclusively to make writing the mdb walker
	 * a little easier).
	 */
	for (i = 0; i < NCPU; i++) {
		trap_trace_ctl_t *ttc = &ttcp[i];

		if (ttc->ttc_first == 0)
			continue;

		/*
		 * Assign ttc_current to be the last completed record.
		 * Note that the error checking (i.e. in the ttc_next ==
		 * ttc_first case) is performed in the step function.
		 */
		ttc->ttc_current = ttc->ttc_next - sizeof (trap_trace_rec_t);
	}

	wsp->walk_data = ttcp;
	return (WALK_NEXT);
}

int
ttrace_walk_step(mdb_walk_state_t *wsp)
{
	trap_trace_ctl_t *ttcp = wsp->walk_data, *ttc, *latest_ttc;
	trap_trace_rec_t rec;
	int rval, i, recsize = sizeof (trap_trace_rec_t);
	hrtime_t latest = 0;

	/*
	 * Loop through the CPUs, looking for the latest trap trace record
	 * (we want to walk through the trap trace records in reverse
	 * chronological order).
	 */
	for (i = 0; i < NCPU; i++) {
		ttc = &ttcp[i];

		if (ttc->ttc_current == 0)
			continue;

		if (ttc->ttc_current < ttc->ttc_first)
			ttc->ttc_current = ttc->ttc_limit - recsize;

		if (mdb_vread(&rec, sizeof (rec), ttc->ttc_current) == -1) {
			mdb_warn("couldn't read rec at %p", ttc->ttc_current);
			return (WALK_ERR);
		}

		if (rec.ttr_stamp > latest) {
			latest = rec.ttr_stamp;
			latest_ttc = ttc;
		}
	}

	if (latest == 0)
		return (WALK_DONE);

	ttc = latest_ttc;

	if (mdb_vread(&rec, sizeof (rec), ttc->ttc_current) == -1) {
		mdb_warn("couldn't read rec at %p", ttc->ttc_current);
		return (WALK_ERR);
	}

	rval = wsp->walk_callback(ttc->ttc_current, &rec, wsp->walk_cbdata);

	if (ttc->ttc_current == ttc->ttc_next)
		ttc->ttc_current = 0;
	else
		ttc->ttc_current -= sizeof (trap_trace_rec_t);

	return (rval);
}

void
ttrace_walk_fini(mdb_walk_state_t *wsp)
{
	mdb_free(wsp->walk_data, sizeof (trap_trace_ctl_t) * NCPU);
}

static int
ttrace_syscall(trap_trace_rec_t *rec)
{
	GElf_Sym sym;
	int sysnum = rec->ttr_sysnum;
	uintptr_t addr;
	struct sysent sys;

	mdb_printf("%-3x", sysnum);

	if (rec->ttr_sysnum > NSYSCALL) {
		mdb_printf(" %-*d", TT_HDLR_WIDTH, rec->ttr_sysnum);
		return (0);
	}

	if (mdb_lookup_by_name("sysent", &sym) == -1) {
		mdb_warn("\ncouldn't find 'sysent'");
		return (-1);
	}

	addr = (uintptr_t)sym.st_value + sysnum * sizeof (struct sysent);

	if (addr >= (uintptr_t)sym.st_value + sym.st_size) {
		mdb_warn("\nsysnum %d out-of-range\n", sysnum);
		return (-1);
	}

	if (mdb_vread(&sys, sizeof (sys), addr) == -1) {
		mdb_warn("\nfailed to read sysent at %p", addr);
		return (-1);
	}

	mdb_printf(" %-*a", TT_HDLR_WIDTH, sys.sy_callc);

	return (0);
}

static int
ttrace_interrupt(trap_trace_rec_t *rec)
{
	GElf_Sym sym;
	uintptr_t addr;
	struct av_head hd;
	struct autovec av;

	switch (rec->ttr_regs.r_trapno) {
	case T_SOFTINT:
		mdb_printf("%-3s %-*s", "-", TT_HDLR_WIDTH, "(fakesoftint)");
		return (0);
	default:
		break;
	}

	mdb_printf("%-3x ", rec->ttr_vector);

	if (mdb_lookup_by_name("autovect", &sym) == -1) {
		mdb_warn("\ncouldn't find 'autovect'");
		return (-1);
	}

	addr = (uintptr_t)sym.st_value +
	    rec->ttr_vector * sizeof (struct av_head);

	if (addr >= (uintptr_t)sym.st_value + sym.st_size) {
		mdb_warn("\nav_head for vec %x is corrupt\n", rec->ttr_vector);
		return (-1);
	}

	if (mdb_vread(&hd, sizeof (hd), addr) == -1) {
		mdb_warn("\ncouldn't read av_head for vec %x", rec->ttr_vector);
		return (-1);
	}

	if (hd.avh_link == NULL) {
		if (rec->ttr_ipl == XC_CPUPOKE_PIL)
			mdb_printf("%-*s", TT_HDLR_WIDTH, "(cpupoke)");
		else
			mdb_printf("%-*s", TT_HDLR_WIDTH, "(spurious)");
	} else {
		if (mdb_vread(&av, sizeof (av), (uintptr_t)hd.avh_link) == -1) {
			mdb_warn("couldn't read autovec at %p",
			    (uintptr_t)hd.avh_link);
		}

		mdb_printf("%-*a", TT_HDLR_WIDTH, av.av_vector);
	}

	return (0);
}

static int
ttrace_apix_interrupt(trap_trace_rec_t *rec)
{
	struct autovec av;
	apix_impl_t apix;
	apix_vector_t apix_vector;

	switch (rec->ttr_regs.r_trapno) {
	case T_SOFTINT:
		mdb_printf("%-3s %-*s", "-", TT_HDLR_WIDTH, "(fakesoftint)");
		return (0);
	default:
		break;
	}

	mdb_printf("%-3x ", rec->ttr_vector);

	/* Read the per CPU apix entry */
	if (mdb_vread(&apix, sizeof (apix_impl_t),
	    (uintptr_t)d_apixs[rec->ttr_cpuid]) == -1) {
		mdb_warn("\ncouldn't read apix[%d]", rec->ttr_cpuid);
		return (-1);
	}
	if (mdb_vread(&apix_vector, sizeof (apix_vector_t),
	    (uintptr_t)apix.x_vectbl[rec->ttr_vector]) == -1) {
		mdb_warn("\ncouldn't read apix_vector_t[%d]", rec->ttr_vector);
		return (-1);
	}
	if (apix_vector.v_share == 0) {
		if (rec->ttr_ipl == XC_CPUPOKE_PIL)
			mdb_printf("%-*s", TT_HDLR_WIDTH, "(cpupoke)");
		else
			mdb_printf("%-*s", TT_HDLR_WIDTH, "(spurious)");
	} else {
		if (mdb_vread(&av, sizeof (struct autovec),
		    (uintptr_t)(apix_vector.v_autovect)) == -1) {
			mdb_warn("couldn't read autovec at %p",
			    (uintptr_t)apix_vector.v_autovect);
		}

		mdb_printf("%-*a", TT_HDLR_WIDTH, av.av_vector);
	}

	return (0);
}


static struct {
	int tt_trapno;
	char *tt_name;
} ttrace_traps[] = {
	{ T_ZERODIV,	"divide-error" },
	{ T_SGLSTP,	"debug-exception" },
	{ T_NMIFLT,	"nmi-interrupt" },
	{ T_BPTFLT,	"breakpoint" },
	{ T_OVFLW,	"into-overflow" },
	{ T_BOUNDFLT,	"bound-exceeded" },
	{ T_ILLINST,	"invalid-opcode" },
	{ T_NOEXTFLT,	"device-not-avail" },
	{ T_DBLFLT,	"double-fault" },
	{ T_EXTOVRFLT,	"segment-overrun" },
	{ T_TSSFLT,	"invalid-tss" },
	{ T_SEGFLT,	"segment-not-pres" },
	{ T_STKFLT,	"stack-fault" },
	{ T_GPFLT,	"general-protectn" },
	{ T_PGFLT,	"page-fault" },
	{ T_EXTERRFLT,	"error-fault" },
	{ T_ALIGNMENT,	"alignment-check" },
	{ T_MCE,	"machine-check" },
	{ T_SIMDFPE,	"sse-exception" },

	{ T_DBGENTR,	"debug-enter" },
	{ T_FASTTRAP,	"fasttrap-0xd2" },
	{ T_SYSCALLINT,	"syscall-0x91" },
	{ T_DTRACE_RET,	"dtrace-ret" },
	{ T_SOFTINT,	"softint" },
	{ T_INTERRUPT,	"interrupt" },
	{ T_FAULT,	"fault" },
	{ T_AST,	"ast" },
	{ T_SYSCALL,	"syscall" },

	{ 0,		NULL }
};

static int
ttrace_trap(trap_trace_rec_t *rec)
{
	int i;

	if (rec->ttr_regs.r_trapno == T_AST)
		mdb_printf("%-3s ", "-");
	else
		mdb_printf("%-3x ", rec->ttr_regs.r_trapno);

	for (i = 0; ttrace_traps[i].tt_name != NULL; i++) {
		if (rec->ttr_regs.r_trapno == ttrace_traps[i].tt_trapno)
			break;
	}

	if (ttrace_traps[i].tt_name == NULL)
		mdb_printf("%-*s", TT_HDLR_WIDTH, "(unknown)");
	else
		mdb_printf("%-*s", TT_HDLR_WIDTH, ttrace_traps[i].tt_name);

	return (0);
}

static void
ttrace_intr_detail(trap_trace_rec_t *rec)
{
	mdb_printf("\tirq %x ipl %d oldpri %d basepri %d\n", rec->ttr_vector,
	    rec->ttr_ipl, rec->ttr_pri, rec->ttr_spl);
}

static struct {
	uchar_t t_marker;
	char *t_name;
	int (*t_hdlr)(trap_trace_rec_t *);
} ttrace_hdlr[] = {
	{ TT_SYSCALL, "sysc", ttrace_syscall },
	{ TT_SYSENTER, "syse", ttrace_syscall },
	{ TT_SYSC, "asys", ttrace_syscall },
	{ TT_SYSC64, "sc64", ttrace_syscall },
	{ TT_INTERRUPT, "intr", ttrace_interrupt },
	{ TT_TRAP, "trap", ttrace_trap },
	{ TT_EVENT, "evnt", ttrace_trap },
	{ 0, NULL, NULL }
};

typedef struct ttrace_dcmd {
	processorid_t ttd_cpu;
	uint_t ttd_extended;
	uintptr_t ttd_kthread;
	trap_trace_ctl_t ttd_ttc[NCPU];
} ttrace_dcmd_t;

#if defined(__amd64)

#define	DUMP(reg) #reg, regs->r_##reg
#define	THREEREGS	"         %3s: %16lx %3s: %16lx %3s: %16lx\n"

static void
ttrace_dumpregs(trap_trace_rec_t *rec)
{
	struct regs *regs = &rec->ttr_regs;

	mdb_printf(THREEREGS, DUMP(rdi), DUMP(rsi), DUMP(rdx));
	mdb_printf(THREEREGS, DUMP(rcx), DUMP(r8), DUMP(r9));
	mdb_printf(THREEREGS, DUMP(rax), DUMP(rbx), DUMP(rbp));
	mdb_printf(THREEREGS, DUMP(r10), DUMP(r11), DUMP(r12));
	mdb_printf(THREEREGS, DUMP(r13), DUMP(r14), DUMP(r15));
	mdb_printf(THREEREGS, DUMP(ds), DUMP(es), DUMP(fs));
	mdb_printf(THREEREGS, DUMP(gs), "trp", regs->r_trapno, DUMP(err));
	mdb_printf(THREEREGS, DUMP(rip), DUMP(cs), DUMP(rfl));
	mdb_printf(THREEREGS, DUMP(rsp), DUMP(ss), "cr2", rec->ttr_cr2);
	mdb_printf("         %3s: %16lx %3s: %16lx\n",
	    "fsb", regs->__r_fsbase,
	    "gsb", regs->__r_gsbase);
	mdb_printf("\n");
}

#else

#define	DUMP(reg) #reg, regs->r_##reg
#define	FOURREGS	"         %3s: %08x %3s: %08x %3s: %08x %3s: %08x\n"

static void
ttrace_dumpregs(trap_trace_rec_t *rec)
{
	struct regs *regs = &rec->ttr_regs;

	mdb_printf(FOURREGS, DUMP(gs), DUMP(fs), DUMP(es), DUMP(ds));
	mdb_printf(FOURREGS, DUMP(edi), DUMP(esi), DUMP(ebp), DUMP(esp));
	mdb_printf(FOURREGS, DUMP(ebx), DUMP(edx), DUMP(ecx), DUMP(eax));
	mdb_printf(FOURREGS, "trp", regs->r_trapno, DUMP(err),
	    DUMP(pc), DUMP(cs));
	mdb_printf(FOURREGS, DUMP(efl), "usp", regs->r_uesp, DUMP(ss),
	    "cr2", rec->ttr_cr2);
	mdb_printf("\n");
}

#endif	/* __amd64 */

int
ttrace_walk(uintptr_t addr, trap_trace_rec_t *rec, ttrace_dcmd_t *dcmd)
{
	struct regs *regs = &rec->ttr_regs;
	processorid_t cpu = -1, i;

	for (i = 0; i < NCPU; i++) {
		if (addr >= dcmd->ttd_ttc[i].ttc_first &&
		    addr < dcmd->ttd_ttc[i].ttc_limit) {
			cpu = i;
			break;
		}
	}

	if (cpu == -1) {
		mdb_warn("couldn't find %p in any trap trace ctl\n", addr);
		return (WALK_ERR);
	}

	if (dcmd->ttd_cpu != -1 && cpu != dcmd->ttd_cpu)
		return (WALK_NEXT);

	if (dcmd->ttd_kthread != 0 &&
	    dcmd->ttd_kthread != rec->ttr_curthread)
		return (WALK_NEXT);

	mdb_printf("%3d %15llx ", cpu, rec->ttr_stamp);

	for (i = 0; ttrace_hdlr[i].t_hdlr != NULL; i++) {
		if (rec->ttr_marker != ttrace_hdlr[i].t_marker)
			continue;
		mdb_printf("%4s ", ttrace_hdlr[i].t_name);
		if (ttrace_hdlr[i].t_hdlr(rec) == -1)
			return (WALK_ERR);
	}

	mdb_printf(" %a\n", regs->r_pc);

	if (dcmd->ttd_extended == FALSE)
		return (WALK_NEXT);

	if (rec->ttr_marker == TT_INTERRUPT)
		ttrace_intr_detail(rec);
	else
		ttrace_dumpregs(rec);

	if (rec->ttr_sdepth > 0) {
		for (i = 0; i < rec->ttr_sdepth; i++) {
			if (i >= TTR_STACK_DEPTH) {
				mdb_printf("%17s*** invalid ttr_sdepth (is %d, "
				    "should be <= %d)\n", " ", rec->ttr_sdepth,
				    TTR_STACK_DEPTH);
				break;
			}

			mdb_printf("%17s %a()\n", " ", rec->ttr_stack[i]);
		}
		mdb_printf("\n");
	}

	return (WALK_NEXT);
}

int
ttrace(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	ttrace_dcmd_t dcmd;
	trap_trace_ctl_t *ttc = dcmd.ttd_ttc;
	trap_trace_rec_t rec;
	size_t ttc_size = sizeof (trap_trace_ctl_t) * NCPU;

	if (!ttrace_ttr_size_check())
		return (WALK_ERR);

	bzero(&dcmd, sizeof (dcmd));
	dcmd.ttd_cpu = -1;
	dcmd.ttd_extended = FALSE;

	if (mdb_readsym(ttc, ttc_size, "trap_trace_ctl") == -1) {
		mdb_warn("symbol 'trap_trace_ctl' not found; "
		    "non-TRAPTRACE kernel?\n");
		return (DCMD_ERR);
	}

	if (mdb_getopts(argc, argv,
	    'x', MDB_OPT_SETBITS, TRUE, &dcmd.ttd_extended,
	    't', MDB_OPT_UINTPTR, &dcmd.ttd_kthread, NULL) != argc)
		return (DCMD_USAGE);

	if (DCMD_HDRSPEC(flags)) {
		mdb_printf("%3s %15s %4s %2s %-*s%s\n", "CPU",
		    "TIMESTAMP", "TYPE", "Vec", TT_HDLR_WIDTH, "HANDLER",
		    " EIP");
	}

	if (flags & DCMD_ADDRSPEC) {
		if (addr >= NCPU) {
			if (mdb_vread(&rec, sizeof (rec), addr) == -1) {
				mdb_warn("couldn't read trap trace record "
				    "at %p", addr);
				return (DCMD_ERR);
			}

			if (ttrace_walk(addr, &rec, &dcmd) == WALK_ERR)
				return (DCMD_ERR);

			return (DCMD_OK);
		}
		dcmd.ttd_cpu = addr;
	}

	if (mdb_readvar(&use_apix, "apix_enable") == -1) {
		mdb_warn("failed to read apix_enable");
		use_apix = 0;
	}

	if (use_apix) {
		if (mdb_readvar(&d_apixs, "apixs") == -1) {
			mdb_warn("\nfailed to read apixs.");
			return (DCMD_ERR);
		}
		/* change to apix ttrace interrupt handler */
		ttrace_hdlr[4].t_hdlr = ttrace_apix_interrupt;
	}

	if (mdb_walk("ttrace", (mdb_walk_cb_t)ttrace_walk, &dcmd) == -1) {
		mdb_warn("couldn't walk 'ttrace'");
		return (DCMD_ERR);
	}

	return (DCMD_OK);
}

/*ARGSUSED*/
int
mutex_owner_init(mdb_walk_state_t *wsp)
{
	return (WALK_NEXT);
}

int
mutex_owner_step(mdb_walk_state_t *wsp)
{
	uintptr_t addr = wsp->walk_addr;
	mutex_impl_t mtx;
	uintptr_t owner;
	kthread_t thr;

	if (mdb_vread(&mtx, sizeof (mtx), addr) == -1)
		return (WALK_ERR);

	if (!MUTEX_TYPE_ADAPTIVE(&mtx))
		return (WALK_DONE);

	if ((owner = (uintptr_t)MUTEX_OWNER(&mtx)) == 0)
		return (WALK_DONE);

	if (mdb_vread(&thr, sizeof (thr), owner) != -1)
		(void) wsp->walk_callback(owner, &thr, wsp->walk_cbdata);

	return (WALK_DONE);
}

static void
gate_desc_dump(gate_desc_t *gate, const char *label, int header)
{
	const char *lastnm;
	uint_t lastval;
	char type[4];

	switch (gate->sgd_type) {
	case SDT_SYSIGT:
		strcpy(type, "int");
		break;
	case SDT_SYSTGT:
		strcpy(type, "trp");
		break;
	case SDT_SYSTASKGT:
		strcpy(type, "tsk");
		break;
	default:
		(void) mdb_snprintf(type, sizeof (type), "%3x", gate->sgd_type);
	}

#if defined(__amd64)
	lastnm = "IST";
	lastval = gate->sgd_ist;
#else
	lastnm = "STK";
	lastval = gate->sgd_stkcpy;
#endif

	if (header) {
		mdb_printf("%*s%<u>%-30s%</u> %<u>%-4s%</u> %<u>%3s%</u> "
		    "%<u>%1s%</u> %<u>%3s%</u> %<u>%3s%</u>\n", strlen(label),
		    "", "HANDLER", "SEL", "DPL", "P", "TYP", lastnm);
	}

	mdb_printf("%s", label);

	if (gate->sgd_type == SDT_SYSTASKGT)
		mdb_printf("%-30s ", "-");
	else
		mdb_printf("%-30a ", GATESEG_GETOFFSET(gate));

	mdb_printf("%4x  %d  %c %3s %2x\n", gate->sgd_selector,
	    gate->sgd_dpl, (gate->sgd_p ? '+' : ' '), type, lastval);
}

/*ARGSUSED*/
static int
gate_desc(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	gate_desc_t gate;

	if (argc != 0 || !(flags & DCMD_ADDRSPEC))
		return (DCMD_USAGE);

	if (mdb_vread(&gate, sizeof (gate_desc_t), addr) !=
	    sizeof (gate_desc_t)) {
		mdb_warn("failed to read gate descriptor at %p\n", addr);
		return (DCMD_ERR);
	}

	gate_desc_dump(&gate, "", DCMD_HDRSPEC(flags));

	return (DCMD_OK);
}

/*ARGSUSED*/
static int
idt(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	int i;

	if (!(flags & DCMD_ADDRSPEC)) {
		GElf_Sym idt0_va;
		gate_desc_t *idt0;

		if (mdb_lookup_by_name("idt0", &idt0_va) < 0) {
			mdb_warn("failed to find VA of idt0");
			return (DCMD_ERR);
		}

		addr = idt0_va.st_value;
		if (mdb_vread(&idt0, sizeof (idt0), addr) != sizeof (idt0)) {
			mdb_warn("failed to read idt0 at %p\n", addr);
			return (DCMD_ERR);
		}

		addr = (uintptr_t)idt0;
	}

	for (i = 0; i < NIDT; i++, addr += sizeof (gate_desc_t)) {
		gate_desc_t gate;
		char label[6];

		if (mdb_vread(&gate, sizeof (gate_desc_t), addr) !=
		    sizeof (gate_desc_t)) {
			mdb_warn("failed to read gate descriptor at %p\n",
			    addr);
			return (DCMD_ERR);
		}

		(void) mdb_snprintf(label, sizeof (label), "%3d: ", i);
		gate_desc_dump(&gate, label, i == 0);
	}

	return (DCMD_OK);
}

static void
htables_help(void)
{
	mdb_printf(
	    "Given a (hat_t *), generates the list of all (htable_t *)s\n"
	    "that correspond to that address space\n");
}

static void
report_maps_help(void)
{
	mdb_printf(
	    "Given a PFN, report HAT structures that map the page, or use\n"
	    "the page as a pagetable.\n"
	    "\n"
	    "-m Interpret the PFN as an MFN (machine frame number)\n");
}

static void
ptable_help(void)
{
	mdb_printf(
	    "Given a PFN holding a page table, print its contents, and\n"
	    "the address of the corresponding htable structure.\n"
	    "\n"
	    "-m Interpret the PFN as an MFN (machine frame number)\n"
	    "-l force page table level (3 is top)\n");
}

static void
ptmap_help(void)
{
	mdb_printf(
	    "Report all mappings represented by the page table hierarchy\n"
	    "rooted at the given cr3 value / physical address.\n"
	    "\n"
	    "-w run ::whatis on mapping start addresses\n");
}

static const char *const scalehrtime_desc =
	"Scales a timestamp from ticks to nanoseconds. Unscaled timestamps\n"
	"are used as both a quick way of accumulating relative time (as for\n"
	"usage) and as a quick way of getting the absolute current time.\n"
	"These uses require slightly different scaling algorithms. By\n"
	"default, if a specified time is greater than half of the unscaled\n"
	"time at the last tick (that is, if the unscaled time represents\n"
	"more than half the time since boot), the timestamp is assumed to\n"
	"be absolute, and the scaling algorithm used mimics that which the\n"
	"kernel uses in gethrtime(). Otherwise, the timestamp is assumed to\n"
	"be relative, and the algorithm mimics scalehrtime(). This behavior\n"
	"can be overridden by forcing the unscaled time to be interpreted\n"
	"as relative (via -r) or absolute (via -a).\n";

static void
scalehrtime_help(void)
{
	mdb_printf("%s", scalehrtime_desc);
}

/*
 * NSEC_SHIFT is replicated here (it is not defined in a header file),
 * but for amusement, the reader is directed to the comment that explains
 * the rationale for this particular value on x86.  Spoiler:  the value is
 * selected to accommodate 60 MHz Pentiums!  (And a confession:  if the voice
 * in that comment sounds too familiar, it's because your author also wrote
 * that code -- some fifteen years prior to this writing in 2011...)
 */
#define	NSEC_SHIFT 5

/*ARGSUSED*/
static int
scalehrtime_dcmd(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	uint32_t nsec_scale;
	hrtime_t tsc = addr, hrt, tsc_last, base, mult = 1;
	unsigned int *tscp = (unsigned int *)&tsc;
	uintptr_t scalehrtimef;
	uint64_t scale;
	GElf_Sym sym;
	int expected = !(flags & DCMD_ADDRSPEC);
	uint_t absolute = FALSE, relative = FALSE;

	if (mdb_getopts(argc, argv,
	    'a', MDB_OPT_SETBITS, TRUE, &absolute,
	    'r', MDB_OPT_SETBITS, TRUE, &relative, NULL) != argc - expected)
		return (DCMD_USAGE);

	if (absolute && relative) {
		mdb_warn("can't specify both -a and -r\n");
		return (DCMD_USAGE);
	}

	if (expected == 1) {
		switch (argv[argc - 1].a_type) {
		case MDB_TYPE_STRING:
			tsc = mdb_strtoull(argv[argc - 1].a_un.a_str);
			break;
		case MDB_TYPE_IMMEDIATE:
			tsc = argv[argc - 1].a_un.a_val;
			break;
		default:
			return (DCMD_USAGE);
		}
	}

	if (mdb_readsym(&scalehrtimef,
	    sizeof (scalehrtimef), "scalehrtimef") == -1) {
		mdb_warn("couldn't read 'scalehrtimef'");
		return (DCMD_ERR);
	}

	if (mdb_lookup_by_name("tsc_scalehrtime", &sym) == -1) {
		mdb_warn("couldn't find 'tsc_scalehrtime'");
		return (DCMD_ERR);
	}

	if (sym.st_value != scalehrtimef) {
		mdb_warn("::scalehrtime requires that scalehrtimef "
		    "be set to tsc_scalehrtime\n");
		return (DCMD_ERR);
	}

	if (mdb_readsym(&nsec_scale, sizeof (nsec_scale), "nsec_scale") == -1) {
		mdb_warn("couldn't read 'nsec_scale'");
		return (DCMD_ERR);
	}

	if (mdb_readsym(&tsc_last, sizeof (tsc_last), "tsc_last") == -1) {
		mdb_warn("couldn't read 'tsc_last'");
		return (DCMD_ERR);
	}

	if (mdb_readsym(&base, sizeof (base), "tsc_hrtime_base") == -1) {
		mdb_warn("couldn't read 'tsc_hrtime_base'");
		return (DCMD_ERR);
	}

	/*
	 * If our time is greater than half of tsc_last, we will take our
	 * delta against tsc_last, convert it, and add that to (or subtract it
	 * from) tsc_hrtime_base.  This mimics what the kernel actually does
	 * in gethrtime() (modulo the tsc_sync_tick_delta) and gets us a much
	 * higher precision result than trying to convert a large tsc value.
	 */
	if (absolute || (tsc > (tsc_last >> 1) && !relative)) {
		if (tsc > tsc_last) {
			tsc = tsc - tsc_last;
		} else {
			tsc = tsc_last - tsc;
			mult = -1;
		}
	} else {
		base = 0;
	}

	scale = (uint64_t)nsec_scale;

	hrt = ((uint64_t)tscp[1] * scale) << NSEC_SHIFT;
	hrt += ((uint64_t)tscp[0] * scale) >> (32 - NSEC_SHIFT);

	mdb_printf("0x%llx\n", base + (hrt * mult));

	return (DCMD_OK);
}

/*
 * The x86 feature set is implemented as a bitmap array. That bitmap array is
 * stored across a number of uchars based on the BT_SIZEOFMAP(NUM_X86_FEATURES)
 * macro. We have the names for each of these features in unix's text segment
 * so we do not have to duplicate them and instead just look them up.
 */
/*ARGSUSED*/
static int
x86_featureset_dcmd(uintptr_t addr, uint_t flags, int argc,
    const mdb_arg_t *argv)
{
	void *fset;
	GElf_Sym sym;
	uintptr_t nptr;
	char name[128];
	int ii;

	size_t sz = sizeof (uchar_t) * BT_SIZEOFMAP(NUM_X86_FEATURES);

	if (argc != 0)
		return (DCMD_USAGE);

	if (mdb_lookup_by_name("x86_feature_names", &sym) == -1) {
		mdb_warn("couldn't find x86_feature_names");
		return (DCMD_ERR);
	}

	fset = mdb_zalloc(sz, UM_NOSLEEP);
	if (fset == NULL) {
		mdb_warn("failed to allocate memory for x86_featureset");
		return (DCMD_ERR);
	}

	if (mdb_readvar(fset, "x86_featureset") != sz) {
		mdb_warn("failed to read x86_featureset");
		mdb_free(fset, sz);
		return (DCMD_ERR);
	}

	for (ii = 0; ii < NUM_X86_FEATURES; ii++) {
		if (!BT_TEST((ulong_t *)fset, ii))
			continue;

		if (mdb_vread(&nptr, sizeof (char *), sym.st_value +
		    sizeof (void *) * ii) != sizeof (char *)) {
			mdb_warn("failed to read feature array %d", ii);
			mdb_free(fset, sz);
			return (DCMD_ERR);
		}

		if (mdb_readstr(name, sizeof (name), nptr) == -1) {
			mdb_warn("failed to read feature %d", ii);
			mdb_free(fset, sz);
			return (DCMD_ERR);
		}
		mdb_printf("%s\n", name);
	}

	mdb_free(fset, sz);
	return (DCMD_OK);
}

#ifdef _KMDB
/* ARGSUSED */
static int
sysregs_dcmd(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	struct sysregs sregs = { 0 };
	desctbr_t gdtr;
	boolean_t longmode = B_FALSE;

#ifdef __amd64
	longmode = B_TRUE;
#endif

	sregs.sr_cr0 = kmdb_unix_getcr0();
	sregs.sr_cr2 = kmdb_unix_getcr2();
	sregs.sr_cr3 = kmdb_unix_getcr3();
	sregs.sr_cr4 = kmdb_unix_getcr4();

	kmdb_unix_getgdtr(&gdtr);
	sregs.sr_gdtr.d_base = gdtr.dtr_base;
	sregs.sr_gdtr.d_lim = gdtr.dtr_limit;

	mdb_x86_print_sysregs(&sregs, longmode);

	return (DCMD_OK);
}
#endif

extern void xcall_help(void);
extern int xcall_dcmd(uintptr_t, uint_t, int, const mdb_arg_t *);

static const mdb_dcmd_t dcmds[] = {
	{ "gate_desc", ":", "dump a gate descriptor", gate_desc },
	{ "idt", ":[-v]", "dump an IDT", idt },
	{ "ttrace", "[-x] [-t kthread]", "dump trap trace buffers", ttrace },
	{ "vatopfn", ":[-a as]", "translate address to physical page",
	    va2pfn_dcmd },
	{ "report_maps", ":[-m]",
	    "Given PFN, report mappings / page table usage",
	    report_maps_dcmd, report_maps_help },
	{ "htables", "", "Given hat_t *, lists all its htable_t * values",
	    htables_dcmd, htables_help },
	{ "ptable", ":[-lm]", "Given PFN, dump contents of a page table",
	    ptable_dcmd, ptable_help },
	{ "ptmap", ":", "Given a cr3 value, dump all mappings",
	    ptmap_dcmd, ptmap_help },
	{ "pte", ":[-l N]", "print human readable page table entry",
	    pte_dcmd },
	{ "pfntomfn", ":", "convert physical page to hypervisor machine page",
	    pfntomfn_dcmd },
	{ "mfntopfn", ":", "convert hypervisor machine page to physical page",
	    mfntopfn_dcmd },
	{ "memseg_list", ":", "show memseg list", memseg_list },
	{ "scalehrtime", ":[-a|-r]", "scale an unscaled high-res time",
	    scalehrtime_dcmd, scalehrtime_help },
	{ "x86_featureset", NULL, "dump the x86_featureset vector",
		x86_featureset_dcmd },
	{ "xcall", ":", "print CPU cross-call state", xcall_dcmd, xcall_help },
#ifdef _KMDB
	{ "sysregs", NULL, "dump system registers", sysregs_dcmd },
#endif
	{ NULL }
};

static const mdb_walker_t walkers[] = {
	{ "ttrace", "walks trap trace buffers in reverse chronological order",
		ttrace_walk_init, ttrace_walk_step, ttrace_walk_fini },
	{ "mutex_owner", "walks the owner of a mutex",
		mutex_owner_init, mutex_owner_step },
	{ "memseg", "walk the memseg structures",
		memseg_walk_init, memseg_walk_step, memseg_walk_fini },
	{ NULL }
};

static const mdb_modinfo_t modinfo = { MDB_API_VERSION, dcmds, walkers };

const mdb_modinfo_t *
_mdb_init(void)
{
	return (&modinfo);
}

void
_mdb_fini(void)
{
	free_mmu();
}