usr/src/cmd/mdb/common/modules/genunix/genunix.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 2011 Nexenta Systems, Inc.  All rights reserved.
 * Copyright (c) 1999, 2010, Oracle and/or its affiliates. All rights reserved.
 * Copyright (c) 2018, Joyent, Inc.
 * Copyright (c) 2013 by Delphix. All rights reserved.
 */

#include <mdb/mdb_param.h>
#include <mdb/mdb_modapi.h>
#include <mdb/mdb_ks.h>
#include <mdb/mdb_ctf.h>

#include <sys/types.h>
#include <sys/thread.h>
#include <sys/session.h>
#include <sys/user.h>
#include <sys/proc.h>
#include <sys/var.h>
#include <sys/t_lock.h>
#include <sys/callo.h>
#include <sys/priocntl.h>
#include <sys/class.h>
#include <sys/regset.h>
#include <sys/stack.h>
#include <sys/cpuvar.h>
#include <sys/vnode.h>
#include <sys/vfs.h>
#include <sys/flock_impl.h>
#include <sys/kmem_impl.h>
#include <sys/vmem_impl.h>
#include <sys/kstat.h>
#include <sys/dditypes.h>
#include <sys/ddi_impldefs.h>
#include <sys/sysmacros.h>
#include <sys/sysconf.h>
#include <sys/task.h>
#include <sys/project.h>
#include <sys/errorq_impl.h>
#include <sys/cred_impl.h>
#include <sys/zone.h>
#include <sys/panic.h>
#include <regex.h>
#include <sys/port_impl.h>

#include "avl.h"
#include "bio.h"
#include "bitset.h"
#include "combined.h"
#include "contract.h"
#include "cpupart_mdb.h"
#include "cred.h"
#include "ctxop.h"
#include "cyclic.h"
#include "damap.h"
#include "ddi_periodic.h"
#include "devinfo.h"
#include "dnlc.h"
#include "findstack.h"
#include "fm.h"
#include "gcore.h"
#include "group.h"
#include "irm.h"
#include "kgrep.h"
#include "kmem.h"
#include "ldi.h"
#include "leaky.h"
#include "lgrp.h"
#include "list.h"
#include "log.h"
#include "mdi.h"
#include "memory.h"
#include "mmd.h"
#include "modhash.h"
#include "ndievents.h"
#include "net.h"
#include "netstack.h"
#include "nvpair.h"
#include "pg.h"
#include "rctl.h"
#include "sobj.h"
#include "streams.h"
#include "sysevent.h"
#include "taskq.h"
#include "thread.h"
#include "tsd.h"
#include "tsol.h"
#include "typegraph.h"
#include "vfs.h"
#include "zone.h"
#include "hotplug.h"

/*
 * Surely this is defined somewhere...
 */
#define	NINTR		16

#define	KILOS		10
#define	MEGS		20
#define	GIGS		30

#ifndef STACK_BIAS
#define	STACK_BIAS	0
#endif

static char
pstat2ch(uchar_t state)
{
	switch (state) {
		case SSLEEP: return ('S');
		case SRUN: return ('R');
		case SZOMB: return ('Z');
		case SIDL: return ('I');
		case SONPROC: return ('O');
		case SSTOP: return ('T');
		case SWAIT: return ('W');
		default: return ('?');
	}
}

#define	PS_PRTTHREADS	0x1
#define	PS_PRTLWPS	0x2
#define	PS_PSARGS	0x4
#define	PS_TASKS	0x8
#define	PS_PROJECTS	0x10
#define	PS_ZONES	0x20

static int
ps_threadprint(uintptr_t addr, const void *data, void *private)
{
	const kthread_t *t = (const kthread_t *)data;
	uint_t prt_flags = *((uint_t *)private);

	static const mdb_bitmask_t t_state_bits[] = {
		{ "TS_FREE",	UINT_MAX,	TS_FREE		},
		{ "TS_SLEEP",	TS_SLEEP,	TS_SLEEP	},
		{ "TS_RUN",	TS_RUN,		TS_RUN		},
		{ "TS_ONPROC",	TS_ONPROC,	TS_ONPROC	},
		{ "TS_ZOMB",	TS_ZOMB,	TS_ZOMB		},
		{ "TS_STOPPED",	TS_STOPPED,	TS_STOPPED	},
		{ "TS_WAIT",	TS_WAIT,	TS_WAIT		},
		{ NULL,		0,		0		}
	};

	if (prt_flags & PS_PRTTHREADS)
		mdb_printf("\tT  %?a <%b>\n", addr, t->t_state, t_state_bits);

	if (prt_flags & PS_PRTLWPS) {
		char desc[128] = "";

		(void) thread_getdesc(addr, B_FALSE, desc, sizeof (desc));

		mdb_printf("\tL  %?a ID: %s\n", t->t_lwp, desc);
	}

	return (WALK_NEXT);
}

typedef struct mdb_pflags_proc {
	struct pid	*p_pidp;
	ushort_t	p_pidflag;
	uint_t		p_proc_flag;
	uint_t		p_flag;
} mdb_pflags_proc_t;

static int
pflags(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	mdb_pflags_proc_t pr;
	struct pid pid;

	static const mdb_bitmask_t p_flag_bits[] = {
		{ "SSYS",		SSYS,		SSYS		},
		{ "SEXITING",		SEXITING,	SEXITING	},
		{ "SITBUSY",		SITBUSY,	SITBUSY		},
		{ "SFORKING",		SFORKING,	SFORKING	},
		{ "SWATCHOK",		SWATCHOK,	SWATCHOK	},
		{ "SKILLED",		SKILLED,	SKILLED		},
		{ "SSCONT",		SSCONT,		SSCONT		},
		{ "SZONETOP",		SZONETOP,	SZONETOP	},
		{ "SEXTKILLED",		SEXTKILLED,	SEXTKILLED	},
		{ "SUGID",		SUGID,		SUGID		},
		{ "SEXECED",		SEXECED,	SEXECED		},
		{ "SJCTL",		SJCTL,		SJCTL		},
		{ "SNOWAIT",		SNOWAIT,	SNOWAIT		},
		{ "SVFORK",		SVFORK,		SVFORK		},
		{ "SVFWAIT",		SVFWAIT,	SVFWAIT		},
		{ "SEXITLWPS",		SEXITLWPS,	SEXITLWPS	},
		{ "SHOLDFORK",		SHOLDFORK,	SHOLDFORK	},
		{ "SHOLDFORK1",		SHOLDFORK1,	SHOLDFORK1	},
		{ "SCOREDUMP",		SCOREDUMP,	SCOREDUMP	},
		{ "SMSACCT",		SMSACCT,	SMSACCT		},
		{ "SLWPWRAP",		SLWPWRAP,	SLWPWRAP	},
		{ "SAUTOLPG",		SAUTOLPG,	SAUTOLPG	},
		{ "SNOCD",		SNOCD,		SNOCD		},
		{ "SHOLDWATCH",		SHOLDWATCH,	SHOLDWATCH	},
		{ "SMSFORK",		SMSFORK,	SMSFORK		},
		{ "SDOCORE",		SDOCORE,	SDOCORE		},
		{ NULL,			0,		0		}
	};

	static const mdb_bitmask_t p_pidflag_bits[] = {
		{ "CLDPEND",		CLDPEND,	CLDPEND		},
		{ "CLDCONT",		CLDCONT,	CLDCONT		},
		{ "CLDNOSIGCHLD",	CLDNOSIGCHLD,	CLDNOSIGCHLD	},
		{ "CLDWAITPID",		CLDWAITPID,	CLDWAITPID	},
		{ NULL,			0,		0		}
	};

	static const mdb_bitmask_t p_proc_flag_bits[] = {
		{ "P_PR_TRACE",		P_PR_TRACE,	P_PR_TRACE	},
		{ "P_PR_PTRACE",	P_PR_PTRACE,	P_PR_PTRACE	},
		{ "P_PR_FORK",		P_PR_FORK,	P_PR_FORK	},
		{ "P_PR_LOCK",		P_PR_LOCK,	P_PR_LOCK	},
		{ "P_PR_ASYNC",		P_PR_ASYNC,	P_PR_ASYNC	},
		{ "P_PR_EXEC",		P_PR_EXEC,	P_PR_EXEC	},
		{ "P_PR_BPTADJ",	P_PR_BPTADJ,	P_PR_BPTADJ	},
		{ "P_PR_RUNLCL",	P_PR_RUNLCL,	P_PR_RUNLCL	},
		{ "P_PR_KILLCL",	P_PR_KILLCL,	P_PR_KILLCL	},
		{ NULL,			0,		0		}
	};

	if (!(flags & DCMD_ADDRSPEC)) {
		if (mdb_walk_dcmd("proc", "pflags", argc, argv) == -1) {
			mdb_warn("can't walk 'proc'");
			return (DCMD_ERR);
		}
		return (DCMD_OK);
	}

	if (mdb_ctf_vread(&pr, "proc_t", "mdb_pflags_proc_t", addr, 0) == -1 ||
	    mdb_vread(&pid, sizeof (pid), (uintptr_t)pr.p_pidp) == -1) {
		mdb_warn("cannot read proc_t or pid");
		return (DCMD_ERR);
	}

	mdb_printf("%p [pid %d]:\n", addr, pid.pid_id);
	mdb_printf("\tp_flag:      %08x <%b>\n", pr.p_flag, pr.p_flag,
	    p_flag_bits);
	mdb_printf("\tp_pidflag:   %08x <%b>\n", pr.p_pidflag, pr.p_pidflag,
	    p_pidflag_bits);
	mdb_printf("\tp_proc_flag: %08x <%b>\n", pr.p_proc_flag, pr.p_proc_flag,
	    p_proc_flag_bits);

	return (DCMD_OK);
}

typedef struct mdb_ps_proc {
	char		p_stat;
	struct pid	*p_pidp;
	struct pid	*p_pgidp;
	struct cred	*p_cred;
	struct sess	*p_sessp;
	struct task	*p_task;
	struct zone	*p_zone;
	pid_t		p_ppid;
	uint_t		p_flag;
	struct {
		char		u_comm[MAXCOMLEN + 1];
		char		u_psargs[PSARGSZ];
	} p_user;
} mdb_ps_proc_t;

int
ps(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	uint_t prt_flags = 0;
	mdb_ps_proc_t pr;
	struct pid pid, pgid, sid;
	sess_t session;
	cred_t cred;
	task_t tk;
	kproject_t pj;
	zone_t zn;

	if (!(flags & DCMD_ADDRSPEC)) {
		if (mdb_walk_dcmd("proc", "ps", argc, argv) == -1) {
			mdb_warn("can't walk 'proc'");
			return (DCMD_ERR);
		}
		return (DCMD_OK);
	}

	if (mdb_getopts(argc, argv,
	    'f', MDB_OPT_SETBITS, PS_PSARGS, &prt_flags,
	    'l', MDB_OPT_SETBITS, PS_PRTLWPS, &prt_flags,
	    'T', MDB_OPT_SETBITS, PS_TASKS, &prt_flags,
	    'P', MDB_OPT_SETBITS, PS_PROJECTS, &prt_flags,
	    'z', MDB_OPT_SETBITS, PS_ZONES, &prt_flags,
	    't', MDB_OPT_SETBITS, PS_PRTTHREADS, &prt_flags, NULL) != argc)
		return (DCMD_USAGE);

	if (DCMD_HDRSPEC(flags)) {
		mdb_printf("%<u>%1s %6s %6s %6s %6s ",
		    "S", "PID", "PPID", "PGID", "SID");
		if (prt_flags & PS_TASKS)
			mdb_printf("%5s ", "TASK");
		if (prt_flags & PS_PROJECTS)
			mdb_printf("%5s ", "PROJ");
		if (prt_flags & PS_ZONES)
			mdb_printf("%5s ", "ZONE");
		mdb_printf("%6s %10s %?s %s%</u>\n",
		    "UID", "FLAGS", "ADDR", "NAME");
	}

	if (mdb_ctf_vread(&pr, "proc_t", "mdb_ps_proc_t", addr, 0) == -1)
		return (DCMD_ERR);

	mdb_vread(&pid, sizeof (pid), (uintptr_t)pr.p_pidp);
	mdb_vread(&pgid, sizeof (pgid), (uintptr_t)pr.p_pgidp);
	mdb_vread(&cred, sizeof (cred), (uintptr_t)pr.p_cred);
	mdb_vread(&session, sizeof (session), (uintptr_t)pr.p_sessp);
	mdb_vread(&sid, sizeof (sid), (uintptr_t)session.s_sidp);
	if (prt_flags & (PS_TASKS | PS_PROJECTS))
		mdb_vread(&tk, sizeof (tk), (uintptr_t)pr.p_task);
	if (prt_flags & PS_PROJECTS)
		mdb_vread(&pj, sizeof (pj), (uintptr_t)tk.tk_proj);
	if (prt_flags & PS_ZONES)
		mdb_vread(&zn, sizeof (zn), (uintptr_t)pr.p_zone);

	mdb_printf("%c %6d %6d %6d %6d ",
	    pstat2ch(pr.p_stat), pid.pid_id, pr.p_ppid, pgid.pid_id,
	    sid.pid_id);
	if (prt_flags & PS_TASKS)
		mdb_printf("%5d ", tk.tk_tkid);
	if (prt_flags & PS_PROJECTS)
		mdb_printf("%5d ", pj.kpj_id);
	if (prt_flags & PS_ZONES)
		mdb_printf("%5d ", zn.zone_id);
	mdb_printf("%6d 0x%08x %0?p %s\n",
	    cred.cr_uid, pr.p_flag, addr,
	    (prt_flags & PS_PSARGS) ? pr.p_user.u_psargs : pr.p_user.u_comm);

	if (prt_flags & ~PS_PSARGS)
		(void) mdb_pwalk("thread", ps_threadprint, &prt_flags, addr);

	return (DCMD_OK);
}

#define	PG_NEWEST	0x0001
#define	PG_OLDEST	0x0002
#define	PG_PIPE_OUT	0x0004
#define	PG_EXACT_MATCH	0x0008

typedef struct pgrep_data {
	uint_t pg_flags;
	uint_t pg_psflags;
	uintptr_t pg_xaddr;
	hrtime_t pg_xstart;
	const char *pg_pat;
#ifndef _KMDB
	regex_t pg_reg;
#endif
} pgrep_data_t;

typedef struct mdb_pgrep_proc {
	struct {
		timestruc_t	u_start;
		char		u_comm[MAXCOMLEN + 1];
	} p_user;
} mdb_pgrep_proc_t;

/*ARGSUSED*/
static int
pgrep_cb(uintptr_t addr, const void *ignored, void *data)
{
	mdb_pgrep_proc_t p;
	pgrep_data_t *pgp = data;
#ifndef _KMDB
	regmatch_t pmatch;
#endif

	if (mdb_ctf_vread(&p, "proc_t", "mdb_pgrep_proc_t", addr, 0) == -1)
		return (WALK_ERR);

	/*
	 * kmdb doesn't have access to the reg* functions, so we fall back
	 * to strstr/strcmp.
	 */
#ifdef _KMDB
	if ((pgp->pg_flags & PG_EXACT_MATCH) ?
	    (strcmp(p.p_user.u_comm, pgp->pg_pat) != 0) :
	    (strstr(p.p_user.u_comm, pgp->pg_pat) == NULL))
		return (WALK_NEXT);
#else
	if (regexec(&pgp->pg_reg, p.p_user.u_comm, 1, &pmatch, 0) != 0)
		return (WALK_NEXT);

	if ((pgp->pg_flags & PG_EXACT_MATCH) &&
	    (pmatch.rm_so != 0 || p.p_user.u_comm[pmatch.rm_eo] != '\0'))
		return (WALK_NEXT);
#endif

	if (pgp->pg_flags & (PG_NEWEST | PG_OLDEST)) {
		hrtime_t start;

		start = (hrtime_t)p.p_user.u_start.tv_sec * NANOSEC +
		    p.p_user.u_start.tv_nsec;

		if (pgp->pg_flags & PG_NEWEST) {
			if (pgp->pg_xaddr == NULL || start > pgp->pg_xstart) {
				pgp->pg_xaddr = addr;
				pgp->pg_xstart = start;
			}
		} else {
			if (pgp->pg_xaddr == NULL || start < pgp->pg_xstart) {
				pgp->pg_xaddr = addr;
				pgp->pg_xstart = start;
			}
		}

	} else if (pgp->pg_flags & PG_PIPE_OUT) {
		mdb_printf("%p\n", addr);

	} else {
		if (mdb_call_dcmd("ps", addr, pgp->pg_psflags, 0, NULL) != 0) {
			mdb_warn("can't invoke 'ps'");
			return (WALK_DONE);
		}
		pgp->pg_psflags &= ~DCMD_LOOPFIRST;
	}

	return (WALK_NEXT);
}

/*ARGSUSED*/
int
pgrep(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	pgrep_data_t pg;
	int i;
#ifndef _KMDB
	int err;
#endif

	if (flags & DCMD_ADDRSPEC)
		return (DCMD_USAGE);

	pg.pg_flags = 0;
	pg.pg_xaddr = 0;

	i = mdb_getopts(argc, argv,
	    'n', MDB_OPT_SETBITS, PG_NEWEST, &pg.pg_flags,
	    'o', MDB_OPT_SETBITS, PG_OLDEST, &pg.pg_flags,
	    'x', MDB_OPT_SETBITS, PG_EXACT_MATCH, &pg.pg_flags,
	    NULL);

	argc -= i;
	argv += i;

	if (argc != 1)
		return (DCMD_USAGE);

	/*
	 * -n and -o are mutually exclusive.
	 */
	if ((pg.pg_flags & PG_NEWEST) && (pg.pg_flags & PG_OLDEST))
		return (DCMD_USAGE);

	if (argv->a_type != MDB_TYPE_STRING)
		return (DCMD_USAGE);

	if (flags & DCMD_PIPE_OUT)
		pg.pg_flags |= PG_PIPE_OUT;

	pg.pg_pat = argv->a_un.a_str;
	if (DCMD_HDRSPEC(flags))
		pg.pg_psflags = DCMD_ADDRSPEC | DCMD_LOOP | DCMD_LOOPFIRST;
	else
		pg.pg_psflags = DCMD_ADDRSPEC | DCMD_LOOP;

#ifndef _KMDB
	if ((err = regcomp(&pg.pg_reg, pg.pg_pat, REG_EXTENDED)) != 0) {
		size_t nbytes;
		char *buf;

		nbytes = regerror(err, &pg.pg_reg, NULL, 0);
		buf = mdb_alloc(nbytes + 1, UM_SLEEP | UM_GC);
		(void) regerror(err, &pg.pg_reg, buf, nbytes);
		mdb_warn("%s\n", buf);

		return (DCMD_ERR);
	}
#endif

	if (mdb_walk("proc", pgrep_cb, &pg) != 0) {
		mdb_warn("can't walk 'proc'");
		return (DCMD_ERR);
	}

	if (pg.pg_xaddr != 0 && (pg.pg_flags & (PG_NEWEST | PG_OLDEST))) {
		if (pg.pg_flags & PG_PIPE_OUT) {
			mdb_printf("%p\n", pg.pg_xaddr);
		} else {
			if (mdb_call_dcmd("ps", pg.pg_xaddr, pg.pg_psflags,
			    0, NULL) != 0) {
				mdb_warn("can't invoke 'ps'");
				return (DCMD_ERR);
			}
		}
	}

	return (DCMD_OK);
}

int
task(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	task_t tk;
	kproject_t pj;

	if (!(flags & DCMD_ADDRSPEC)) {
		if (mdb_walk_dcmd("task_cache", "task", argc, argv) == -1) {
			mdb_warn("can't walk task_cache");
			return (DCMD_ERR);
		}
		return (DCMD_OK);
	}
	if (DCMD_HDRSPEC(flags)) {
		mdb_printf("%<u>%?s %6s %6s %6s %6s %10s%</u>\n",
		    "ADDR", "TASKID", "PROJID", "ZONEID", "REFCNT", "FLAGS");
	}
	if (mdb_vread(&tk, sizeof (task_t), addr) == -1) {
		mdb_warn("can't read task_t structure at %p", addr);
		return (DCMD_ERR);
	}
	if (mdb_vread(&pj, sizeof (kproject_t), (uintptr_t)tk.tk_proj) == -1) {
		mdb_warn("can't read project_t structure at %p", addr);
		return (DCMD_ERR);
	}
	mdb_printf("%0?p %6d %6d %6d %6u 0x%08x\n",
	    addr, tk.tk_tkid, pj.kpj_id, pj.kpj_zoneid, tk.tk_hold_count,
	    tk.tk_flags);
	return (DCMD_OK);
}

int
project(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	kproject_t pj;

	if (!(flags & DCMD_ADDRSPEC)) {
		if (mdb_walk_dcmd("projects", "project", argc, argv) == -1) {
			mdb_warn("can't walk projects");
			return (DCMD_ERR);
		}
		return (DCMD_OK);
	}
	if (DCMD_HDRSPEC(flags)) {
		mdb_printf("%<u>%?s %6s %6s %6s%</u>\n",
		    "ADDR", "PROJID", "ZONEID", "REFCNT");
	}
	if (mdb_vread(&pj, sizeof (kproject_t), addr) == -1) {
		mdb_warn("can't read kproject_t structure at %p", addr);
		return (DCMD_ERR);
	}
	mdb_printf("%0?p %6d %6d %6u\n", addr, pj.kpj_id, pj.kpj_zoneid,
	    pj.kpj_count);
	return (DCMD_OK);
}

/* walk callouts themselves, either by list or id hash. */
int
callout_walk_init(mdb_walk_state_t *wsp)
{
	if (wsp->walk_addr == NULL) {
		mdb_warn("callout doesn't support global walk");
		return (WALK_ERR);
	}
	wsp->walk_data = mdb_alloc(sizeof (callout_t), UM_SLEEP);
	return (WALK_NEXT);
}

#define	CALLOUT_WALK_BYLIST	0
#define	CALLOUT_WALK_BYID	1

/* the walker arg switches between walking by list (0) and walking by id (1). */
int
callout_walk_step(mdb_walk_state_t *wsp)
{
	int retval;

	if (wsp->walk_addr == NULL) {
		return (WALK_DONE);
	}
	if (mdb_vread(wsp->walk_data, sizeof (callout_t),
	    wsp->walk_addr) == -1) {
		mdb_warn("failed to read callout at %p", wsp->walk_addr);
		return (WALK_DONE);
	}
	retval = wsp->walk_callback(wsp->walk_addr, wsp->walk_data,
	    wsp->walk_cbdata);

	if ((ulong_t)wsp->walk_arg == CALLOUT_WALK_BYID) {
		wsp->walk_addr =
		    (uintptr_t)(((callout_t *)wsp->walk_data)->c_idnext);
	} else {
		wsp->walk_addr =
		    (uintptr_t)(((callout_t *)wsp->walk_data)->c_clnext);
	}

	return (retval);
}

void
callout_walk_fini(mdb_walk_state_t *wsp)
{
	mdb_free(wsp->walk_data, sizeof (callout_t));
}

/*
 * walker for callout lists. This is different from hashes and callouts.
 * Thankfully, it's also simpler.
 */
int
callout_list_walk_init(mdb_walk_state_t *wsp)
{
	if (wsp->walk_addr == NULL) {
		mdb_warn("callout list doesn't support global walk");
		return (WALK_ERR);
	}
	wsp->walk_data = mdb_alloc(sizeof (callout_list_t), UM_SLEEP);
	return (WALK_NEXT);
}

int
callout_list_walk_step(mdb_walk_state_t *wsp)
{
	int retval;

	if (wsp->walk_addr == NULL) {
		return (WALK_DONE);
	}
	if (mdb_vread(wsp->walk_data, sizeof (callout_list_t),
	    wsp->walk_addr) != sizeof (callout_list_t)) {
		mdb_warn("failed to read callout_list at %p", wsp->walk_addr);
		return (WALK_ERR);
	}
	retval = wsp->walk_callback(wsp->walk_addr, wsp->walk_data,
	    wsp->walk_cbdata);

	wsp->walk_addr = (uintptr_t)
	    (((callout_list_t *)wsp->walk_data)->cl_next);

	return (retval);
}

void
callout_list_walk_fini(mdb_walk_state_t *wsp)
{
	mdb_free(wsp->walk_data, sizeof (callout_list_t));
}

/* routines/structs to walk callout table(s) */
typedef struct cot_data {
	callout_table_t *ct0;
	callout_table_t ct;
	callout_hash_t cot_idhash[CALLOUT_BUCKETS];
	callout_hash_t cot_clhash[CALLOUT_BUCKETS];
	kstat_named_t ct_kstat_data[CALLOUT_NUM_STATS];
	int cotndx;
	int cotsize;
} cot_data_t;

int
callout_table_walk_init(mdb_walk_state_t *wsp)
{
	int max_ncpus;
	cot_data_t *cot_walk_data;

	cot_walk_data = mdb_alloc(sizeof (cot_data_t), UM_SLEEP);

	if (wsp->walk_addr == NULL) {
		if (mdb_readvar(&cot_walk_data->ct0, "callout_table") == -1) {
			mdb_warn("failed to read 'callout_table'");
			return (WALK_ERR);
		}
		if (mdb_readvar(&max_ncpus, "max_ncpus") == -1) {
			mdb_warn("failed to get callout_table array size");
			return (WALK_ERR);
		}
		cot_walk_data->cotsize = CALLOUT_NTYPES * max_ncpus;
		wsp->walk_addr = (uintptr_t)cot_walk_data->ct0;
	} else {
		/* not a global walk */
		cot_walk_data->cotsize = 1;
	}

	cot_walk_data->cotndx = 0;
	wsp->walk_data = cot_walk_data;

	return (WALK_NEXT);
}

int
callout_table_walk_step(mdb_walk_state_t *wsp)
{
	int retval;
	cot_data_t *cotwd = (cot_data_t *)wsp->walk_data;
	size_t size;

	if (cotwd->cotndx >= cotwd->cotsize) {
		return (WALK_DONE);
	}
	if (mdb_vread(&(cotwd->ct), sizeof (callout_table_t),
	    wsp->walk_addr) != sizeof (callout_table_t)) {
		mdb_warn("failed to read callout_table at %p", wsp->walk_addr);
		return (WALK_ERR);
	}

	size = sizeof (callout_hash_t) * CALLOUT_BUCKETS;
	if (cotwd->ct.ct_idhash != NULL) {
		if (mdb_vread(cotwd->cot_idhash, size,
		    (uintptr_t)(cotwd->ct.ct_idhash)) != size) {
			mdb_warn("failed to read id_hash at %p",
			    cotwd->ct.ct_idhash);
			return (WALK_ERR);
		}
	}
	if (cotwd->ct.ct_clhash != NULL) {
		if (mdb_vread(&(cotwd->cot_clhash), size,
		    (uintptr_t)cotwd->ct.ct_clhash) == -1) {
			mdb_warn("failed to read cl_hash at %p",
			    cotwd->ct.ct_clhash);
			return (WALK_ERR);
		}
	}
	size = sizeof (kstat_named_t) * CALLOUT_NUM_STATS;
	if (cotwd->ct.ct_kstat_data != NULL) {
		if (mdb_vread(&(cotwd->ct_kstat_data), size,
		    (uintptr_t)cotwd->ct.ct_kstat_data) == -1) {
			mdb_warn("failed to read kstats at %p",
			    cotwd->ct.ct_kstat_data);
			return (WALK_ERR);
		}
	}
	retval = wsp->walk_callback(wsp->walk_addr, (void *)cotwd,
	    wsp->walk_cbdata);

	cotwd->cotndx++;
	if (cotwd->cotndx >= cotwd->cotsize) {
		return (WALK_DONE);
	}
	wsp->walk_addr = (uintptr_t)((char *)wsp->walk_addr +
	    sizeof (callout_table_t));

	return (retval);
}

void
callout_table_walk_fini(mdb_walk_state_t *wsp)
{
	mdb_free(wsp->walk_data, sizeof (cot_data_t));
}

static const char *co_typenames[] = { "R", "N" };

#define	CO_PLAIN_ID(xid)	((xid) & CALLOUT_ID_MASK)

#define	TABLE_TO_SEQID(x)	((x) >> CALLOUT_TYPE_BITS)

/* callout flags, in no particular order */
#define	COF_REAL	0x00000001
#define	COF_NORM	0x00000002
#define	COF_LONG	0x00000004
#define	COF_SHORT	0x00000008
#define	COF_EMPTY	0x00000010
#define	COF_TIME	0x00000020
#define	COF_BEFORE	0x00000040
#define	COF_AFTER	0x00000080
#define	COF_SEQID	0x00000100
#define	COF_FUNC	0x00000200
#define	COF_ADDR	0x00000400
#define	COF_EXEC	0x00000800
#define	COF_HIRES	0x00001000
#define	COF_ABS		0x00002000
#define	COF_TABLE	0x00004000
#define	COF_BYIDH	0x00008000
#define	COF_FREE	0x00010000
#define	COF_LIST	0x00020000
#define	COF_EXPREL	0x00040000
#define	COF_HDR		0x00080000
#define	COF_VERBOSE	0x00100000
#define	COF_LONGLIST	0x00200000
#define	COF_THDR	0x00400000
#define	COF_LHDR	0x00800000
#define	COF_CHDR	0x01000000
#define	COF_PARAM	0x02000000
#define	COF_DECODE	0x04000000
#define	COF_HEAP	0x08000000
#define	COF_QUEUE	0x10000000

/* show real and normal, short and long, expired and unexpired. */
#define	COF_DEFAULT	(COF_REAL | COF_NORM | COF_LONG | COF_SHORT)

#define	COF_LIST_FLAGS	\
	(CALLOUT_LIST_FLAG_HRESTIME | CALLOUT_LIST_FLAG_ABSOLUTE)

/* private callout data for callback functions */
typedef struct callout_data {
	uint_t flags;		/* COF_* */
	cpu_t *cpu;		/* cpu pointer if given */
	int seqid;		/* cpu seqid, or -1 */
	hrtime_t time;		/* expiration time value */
	hrtime_t atime;		/* expiration before value */
	hrtime_t btime;		/* expiration after value */
	uintptr_t funcaddr;	/* function address or NULL */
	uintptr_t param;	/* parameter to function or NULL */
	hrtime_t now;		/* current system time */
	int nsec_per_tick;	/* for conversions */
	ulong_t ctbits;		/* for decoding xid */
	callout_table_t *co_table;	/* top of callout table array */
	int ndx;		/* table index. */
	int bucket;		/* which list/id bucket are we in */
	hrtime_t exp;		/* expire time */
	int list_flags;		/* copy of cl_flags */
} callout_data_t;

/* this callback does the actual callback itself (finally). */
/*ARGSUSED*/
static int
callouts_cb(uintptr_t addr, const void *data, void *priv)
{
	callout_data_t *coargs = (callout_data_t *)priv;
	callout_t *co = (callout_t *)data;
	int tableid, list_flags;
	callout_id_t coid;

	if ((coargs == NULL) || (co == NULL)) {
		return (WALK_ERR);
	}

	if ((coargs->flags & COF_FREE) && !(co->c_xid & CALLOUT_ID_FREE)) {
		/*
		 * The callout must have been reallocated. No point in
		 * walking any more.
		 */
		return (WALK_DONE);
	}
	if (!(coargs->flags & COF_FREE) && (co->c_xid & CALLOUT_ID_FREE)) {
		/*
		 * The callout must have been freed. No point in
		 * walking any more.
		 */
		return (WALK_DONE);
	}
	if ((coargs->flags & COF_FUNC) &&
	    (coargs->funcaddr != (uintptr_t)co->c_func)) {
		return (WALK_NEXT);
	}
	if ((coargs->flags & COF_PARAM) &&
	    (coargs->param != (uintptr_t)co->c_arg)) {
		return (WALK_NEXT);
	}
	if (!(coargs->flags & COF_LONG) && (co->c_xid & CALLOUT_LONGTERM)) {
		return (WALK_NEXT);
	}
	if (!(coargs->flags & COF_SHORT) && !(co->c_xid & CALLOUT_LONGTERM)) {
		return (WALK_NEXT);
	}
	if ((coargs->flags & COF_EXEC) && !(co->c_xid & CALLOUT_EXECUTING)) {
		return (WALK_NEXT);
	}
	/* it is possible we don't have the exp time or flags */
	if (coargs->flags & COF_BYIDH) {
		if (!(coargs->flags & COF_FREE)) {
			/* we have to fetch the expire time ourselves. */
			if (mdb_vread(&coargs->exp, sizeof (hrtime_t),
			    (uintptr_t)co->c_list + offsetof(callout_list_t,
			    cl_expiration)) == -1) {
				mdb_warn("failed to read expiration "
				    "time from %p", co->c_list);
				coargs->exp = 0;
			}
			/* and flags. */
			if (mdb_vread(&coargs->list_flags, sizeof (int),
			    (uintptr_t)co->c_list + offsetof(callout_list_t,
			    cl_flags)) == -1) {
				mdb_warn("failed to read list flags"
				    "from %p", co->c_list);
				coargs->list_flags = 0;
			}
		} else {
			/* free callouts can't use list pointer. */
			coargs->exp = 0;
			coargs->list_flags = 0;
		}
		if (coargs->exp != 0) {
			if ((coargs->flags & COF_TIME) &&
			    (coargs->exp != coargs->time)) {
				return (WALK_NEXT);
			}
			if ((coargs->flags & COF_BEFORE) &&
			    (coargs->exp > coargs->btime)) {
				return (WALK_NEXT);
			}
			if ((coargs->flags & COF_AFTER) &&
			    (coargs->exp < coargs->atime)) {
				return (WALK_NEXT);
			}
		}
		/* tricky part, since both HIRES and ABS can be set */
		list_flags = coargs->list_flags;
		if ((coargs->flags & COF_HIRES) && (coargs->flags & COF_ABS)) {
			/* both flags are set, only skip "regular" ones */
			if (! (list_flags & COF_LIST_FLAGS)) {
				return (WALK_NEXT);
			}
		} else {
			/* individual flags, or no flags */
			if ((coargs->flags & COF_HIRES) &&
			    !(list_flags & CALLOUT_LIST_FLAG_HRESTIME)) {
				return (WALK_NEXT);
			}
			if ((coargs->flags & COF_ABS) &&
			    !(list_flags & CALLOUT_LIST_FLAG_ABSOLUTE)) {
				return (WALK_NEXT);
			}
		}
		/*
		 * We do the checks for COF_HEAP and COF_QUEUE here only if we
		 * are traversing BYIDH. If the traversal is by callout list,
		 * we do this check in callout_list_cb() to be more
		 * efficient.
		 */
		if ((coargs->flags & COF_HEAP) &&
		    !(list_flags & CALLOUT_LIST_FLAG_HEAPED)) {
			return (WALK_NEXT);
		}

		if ((coargs->flags & COF_QUEUE) &&
		    !(list_flags & CALLOUT_LIST_FLAG_QUEUED)) {
			return (WALK_NEXT);
		}
	}

#define	callout_table_mask	((1 << coargs->ctbits) - 1)
	tableid = CALLOUT_ID_TO_TABLE(co->c_xid);
#undef	callout_table_mask
	coid = CO_PLAIN_ID(co->c_xid);

	if ((coargs->flags & COF_CHDR) && !(coargs->flags & COF_ADDR)) {
		/*
		 * We need to print the headers. If walking by id, then
		 * the list header isn't printed, so we must include
		 * that info here.
		 */
		if (!(coargs->flags & COF_VERBOSE)) {
			mdb_printf("%<u>%3s %-1s %-14s %</u>",
			    "SEQ", "T", "EXP");
		} else if (coargs->flags & COF_BYIDH) {
			mdb_printf("%<u>%-14s %</u>", "EXP");
		}
		mdb_printf("%<u>%-4s %-?s %-20s%</u>",
		    "XHAL", "XID", "FUNC(ARG)");
		if (coargs->flags & COF_LONGLIST) {
			mdb_printf("%<u> %-?s %-?s %-?s %-?s%</u>",
			    "PREVID", "NEXTID", "PREVL", "NEXTL");
			mdb_printf("%<u> %-?s %-4s %-?s%</u>",
			    "DONE", "UTOS", "THREAD");
		}
		mdb_printf("\n");
		coargs->flags &= ~COF_CHDR;
		coargs->flags |= (COF_THDR | COF_LHDR);
	}

	if (!(coargs->flags & COF_ADDR)) {
		if (!(coargs->flags & COF_VERBOSE)) {
			mdb_printf("%-3d %1s %-14llx ",
			    TABLE_TO_SEQID(tableid),
			    co_typenames[tableid & CALLOUT_TYPE_MASK],
			    (coargs->flags & COF_EXPREL) ?
			    coargs->exp - coargs->now : coargs->exp);
		} else if (coargs->flags & COF_BYIDH) {
			mdb_printf("%-14x ",
			    (coargs->flags & COF_EXPREL) ?
			    coargs->exp - coargs->now : coargs->exp);
		}
		list_flags = coargs->list_flags;
		mdb_printf("%1s%1s%1s%1s %-?llx %a(%p)",
		    (co->c_xid & CALLOUT_EXECUTING) ? "X" : " ",
		    (list_flags & CALLOUT_LIST_FLAG_HRESTIME) ? "H" : " ",
		    (list_flags & CALLOUT_LIST_FLAG_ABSOLUTE) ? "A" : " ",
		    (co->c_xid & CALLOUT_LONGTERM) ? "L" : " ",
		    (long long)coid, co->c_func, co->c_arg);
		if (coargs->flags & COF_LONGLIST) {
			mdb_printf(" %-?p %-?p %-?p %-?p",
			    co->c_idprev, co->c_idnext, co->c_clprev,
			    co->c_clnext);
			mdb_printf(" %-?p %-4d %-0?p",
			    co->c_done, co->c_waiting, co->c_executor);
		}
	} else {
		/* address only */
		mdb_printf("%-0p", addr);
	}
	mdb_printf("\n");
	return (WALK_NEXT);
}

/* this callback is for callout list handling. idhash is done by callout_t_cb */
/*ARGSUSED*/
static int
callout_list_cb(uintptr_t addr, const void *data, void *priv)
{
	callout_data_t *coargs = (callout_data_t *)priv;
	callout_list_t *cl = (callout_list_t *)data;
	callout_t *coptr;
	int list_flags;

	if ((coargs == NULL) || (cl == NULL)) {
		return (WALK_ERR);
	}

	coargs->exp = cl->cl_expiration;
	coargs->list_flags = cl->cl_flags;
	if ((coargs->flags & COF_FREE) &&
	    !(cl->cl_flags & CALLOUT_LIST_FLAG_FREE)) {
		/*
		 * The callout list must have been reallocated. No point in
		 * walking any more.
		 */
		return (WALK_DONE);
	}
	if (!(coargs->flags & COF_FREE) &&
	    (cl->cl_flags & CALLOUT_LIST_FLAG_FREE)) {
		/*
		 * The callout list must have been freed. No point in
		 * walking any more.
		 */
		return (WALK_DONE);
	}
	if ((coargs->flags & COF_TIME) &&
	    (cl->cl_expiration != coargs->time)) {
		return (WALK_NEXT);
	}
	if ((coargs->flags & COF_BEFORE) &&
	    (cl->cl_expiration > coargs->btime)) {
		return (WALK_NEXT);
	}
	if ((coargs->flags & COF_AFTER) &&
	    (cl->cl_expiration < coargs->atime)) {
		return (WALK_NEXT);
	}
	if (!(coargs->flags & COF_EMPTY) &&
	    (cl->cl_callouts.ch_head == NULL)) {
		return (WALK_NEXT);
	}
	/* FOUR cases, each different, !A!B, !AB, A!B, AB */
	if ((coargs->flags & COF_HIRES) && (coargs->flags & COF_ABS)) {
		/* both flags are set, only skip "regular" ones */
		if (! (cl->cl_flags & COF_LIST_FLAGS)) {
			return (WALK_NEXT);
		}
	} else {
		if ((coargs->flags & COF_HIRES) &&
		    !(cl->cl_flags & CALLOUT_LIST_FLAG_HRESTIME)) {
			return (WALK_NEXT);
		}
		if ((coargs->flags & COF_ABS) &&
		    !(cl->cl_flags & CALLOUT_LIST_FLAG_ABSOLUTE)) {
			return (WALK_NEXT);
		}
	}

	if ((coargs->flags & COF_HEAP) &&
	    !(coargs->list_flags & CALLOUT_LIST_FLAG_HEAPED)) {
		return (WALK_NEXT);
	}

	if ((coargs->flags & COF_QUEUE) &&
	    !(coargs->list_flags & CALLOUT_LIST_FLAG_QUEUED)) {
		return (WALK_NEXT);
	}

	if ((coargs->flags & COF_LHDR) && !(coargs->flags & COF_ADDR) &&
	    (coargs->flags & (COF_LIST | COF_VERBOSE))) {
		if (!(coargs->flags & COF_VERBOSE)) {
			/* don't be redundant again */
			mdb_printf("%<u>SEQ T %</u>");
		}
		mdb_printf("%<u>EXP            HA BUCKET "
		    "CALLOUTS         %</u>");

		if (coargs->flags & COF_LONGLIST) {
			mdb_printf("%<u> %-?s %-?s%</u>",
			    "PREV", "NEXT");
		}
		mdb_printf("\n");
		coargs->flags &= ~COF_LHDR;
		coargs->flags |= (COF_THDR | COF_CHDR);
	}
	if (coargs->flags & (COF_LIST | COF_VERBOSE)) {
		if (!(coargs->flags & COF_ADDR)) {
			if (!(coargs->flags & COF_VERBOSE)) {
				mdb_printf("%3d %1s ",
				    TABLE_TO_SEQID(coargs->ndx),
				    co_typenames[coargs->ndx &
				    CALLOUT_TYPE_MASK]);
			}

			list_flags = coargs->list_flags;
			mdb_printf("%-14llx %1s%1s %-6d %-0?p ",
			    (coargs->flags & COF_EXPREL) ?
			    coargs->exp - coargs->now : coargs->exp,
			    (list_flags & CALLOUT_LIST_FLAG_HRESTIME) ?
			    "H" : " ",
			    (list_flags & CALLOUT_LIST_FLAG_ABSOLUTE) ?
			    "A" : " ",
			    coargs->bucket, cl->cl_callouts.ch_head);

			if (coargs->flags & COF_LONGLIST) {
				mdb_printf(" %-?p %-?p",
				    cl->cl_prev, cl->cl_next);
			}
		} else {
			/* address only */
			mdb_printf("%-0p", addr);
		}
		mdb_printf("\n");
		if (coargs->flags & COF_LIST) {
			return (WALK_NEXT);
		}
	}
	/* yet another layer as we walk the actual callouts via list. */
	if (cl->cl_callouts.ch_head == NULL) {
		return (WALK_NEXT);
	}
	/* free list structures do not have valid callouts off of them. */
	if (coargs->flags & COF_FREE) {
		return (WALK_NEXT);
	}
	coptr = (callout_t *)cl->cl_callouts.ch_head;

	if (coargs->flags & COF_VERBOSE) {
		mdb_inc_indent(4);
	}
	/*
	 * walk callouts using yet another callback routine.
	 * we use callouts_bytime because id hash is handled via
	 * the callout_t_cb callback.
	 */
	if (mdb_pwalk("callouts_bytime", callouts_cb, coargs,
	    (uintptr_t)coptr) == -1) {
		mdb_warn("cannot walk callouts at %p", coptr);
		return (WALK_ERR);
	}
	if (coargs->flags & COF_VERBOSE) {
		mdb_dec_indent(4);
	}

	return (WALK_NEXT);
}

/* this callback handles the details of callout table walking. */
static int
callout_t_cb(uintptr_t addr, const void *data, void *priv)
{
	callout_data_t *coargs = (callout_data_t *)priv;
	cot_data_t *cotwd = (cot_data_t *)data;
	callout_table_t *ct = &(cotwd->ct);
	int index, seqid, cotype;
	int i;
	callout_list_t *clptr;
	callout_t *coptr;

	if ((coargs == NULL) || (ct == NULL) || (coargs->co_table == NULL)) {
		return (WALK_ERR);
	}

	index =  ((char *)addr - (char *)coargs->co_table) /
	    sizeof (callout_table_t);
	cotype = index & CALLOUT_TYPE_MASK;
	seqid = TABLE_TO_SEQID(index);

	if ((coargs->flags & COF_SEQID) && (coargs->seqid != seqid)) {
		return (WALK_NEXT);
	}

	if (!(coargs->flags & COF_REAL) && (cotype == CALLOUT_REALTIME)) {
		return (WALK_NEXT);
	}

	if (!(coargs->flags & COF_NORM) && (cotype == CALLOUT_NORMAL)) {
		return (WALK_NEXT);
	}

	if (!(coargs->flags & COF_EMPTY) && (
	    (ct->ct_heap == NULL) || (ct->ct_cyclic == NULL))) {
		return (WALK_NEXT);
	}

	if ((coargs->flags & COF_THDR) && !(coargs->flags & COF_ADDR) &&
	    (coargs->flags & (COF_TABLE | COF_VERBOSE))) {
		/* print table hdr */
		mdb_printf("%<u>%-3s %-1s %-?s %-?s %-?s %-?s%</u>",
		    "SEQ", "T", "FREE", "LFREE", "CYCLIC", "HEAP");
		coargs->flags &= ~COF_THDR;
		coargs->flags |= (COF_LHDR | COF_CHDR);
		if (coargs->flags & COF_LONGLIST) {
			/* more info! */
			mdb_printf("%<u> %-T%-7s %-7s %-?s %-?s %-?s"
			    " %-?s %-?s %-?s%</u>",
			    "HEAPNUM", "HEAPMAX", "TASKQ", "EXPQ", "QUE",
			    "PEND", "FREE", "LOCK");
		}
		mdb_printf("\n");
	}
	if (coargs->flags & (COF_TABLE | COF_VERBOSE)) {
		if (!(coargs->flags & COF_ADDR)) {
			mdb_printf("%-3d %-1s %-0?p %-0?p %-0?p %-?p",
			    seqid, co_typenames[cotype],
			    ct->ct_free, ct->ct_lfree, ct->ct_cyclic,
			    ct->ct_heap);
			if (coargs->flags & COF_LONGLIST)  {
				/* more info! */
				mdb_printf(" %-7d %-7d %-?p %-?p %-?p"
				    " %-?lld %-?lld %-?p",
				    ct->ct_heap_num,  ct->ct_heap_max,
				    ct->ct_taskq, ct->ct_expired.ch_head,
				    ct->ct_queue.ch_head,
				    cotwd->ct_timeouts_pending,
				    cotwd->ct_allocations -
				    cotwd->ct_timeouts_pending,
				    ct->ct_mutex);
			}
		} else {
			/* address only */
			mdb_printf("%-0?p", addr);
		}
		mdb_printf("\n");
		if (coargs->flags & COF_TABLE) {
			return (WALK_NEXT);
		}
	}

	coargs->ndx = index;
	if (coargs->flags & COF_VERBOSE) {
		mdb_inc_indent(4);
	}
	/* keep digging. */
	if (!(coargs->flags & COF_BYIDH)) {
		/* walk the list hash table */
		if (coargs->flags & COF_FREE) {
			clptr = ct->ct_lfree;
			coargs->bucket = 0;
			if (clptr == NULL) {
				return (WALK_NEXT);
			}
			if (mdb_pwalk("callout_list", callout_list_cb, coargs,
			    (uintptr_t)clptr) == -1) {
				mdb_warn("cannot walk callout free list at %p",
				    clptr);
				return (WALK_ERR);
			}
		} else {
			/* first print the expired list. */
			clptr = (callout_list_t *)ct->ct_expired.ch_head;
			if (clptr != NULL) {
				coargs->bucket = -1;
				if (mdb_pwalk("callout_list", callout_list_cb,
				    coargs, (uintptr_t)clptr) == -1) {
					mdb_warn("cannot walk callout_list"
					    " at %p", clptr);
					return (WALK_ERR);
				}
			}
			/* then, print the callout queue */
			clptr = (callout_list_t *)ct->ct_queue.ch_head;
			if (clptr != NULL) {
				coargs->bucket = -1;
				if (mdb_pwalk("callout_list", callout_list_cb,
				    coargs, (uintptr_t)clptr) == -1) {
					mdb_warn("cannot walk callout_list"
					    " at %p", clptr);
					return (WALK_ERR);
				}
			}
			for (i = 0; i < CALLOUT_BUCKETS; i++) {
				if (ct->ct_clhash == NULL) {
					/* nothing to do */
					break;
				}
				if (cotwd->cot_clhash[i].ch_head == NULL) {
					continue;
				}
				clptr = (callout_list_t *)
				    cotwd->cot_clhash[i].ch_head;
				coargs->bucket = i;
				/* walk list with callback routine. */
				if (mdb_pwalk("callout_list", callout_list_cb,
				    coargs, (uintptr_t)clptr) == -1) {
					mdb_warn("cannot walk callout_list"
					    " at %p", clptr);
					return (WALK_ERR);
				}
			}
		}
	} else {
		/* walk the id hash table. */
		if (coargs->flags & COF_FREE) {
			coptr = ct->ct_free;
			coargs->bucket = 0;
			if (coptr == NULL) {
				return (WALK_NEXT);
			}
			if (mdb_pwalk("callouts_byid", callouts_cb, coargs,
			    (uintptr_t)coptr) == -1) {
				mdb_warn("cannot walk callout id free list"
				    " at %p", coptr);
				return (WALK_ERR);
			}
		} else {
			for (i = 0; i < CALLOUT_BUCKETS; i++) {
				if (ct->ct_idhash == NULL) {
					break;
				}
				coptr = (callout_t *)
				    cotwd->cot_idhash[i].ch_head;
				if (coptr == NULL) {
					continue;
				}
				coargs->bucket = i;

				/*
				 * walk callouts directly by id. For id
				 * chain, the callout list is just a header,
				 * so there's no need to walk it.
				 */
				if (mdb_pwalk("callouts_byid", callouts_cb,
				    coargs, (uintptr_t)coptr) == -1) {
					mdb_warn("cannot walk callouts at %p",
					    coptr);
					return (WALK_ERR);
				}
			}
		}
	}
	if (coargs->flags & COF_VERBOSE) {
		mdb_dec_indent(4);
	}
	return (WALK_NEXT);
}

/*
 * initialize some common info for both callout dcmds.
 */
int
callout_common_init(callout_data_t *coargs)
{
	/* we need a couple of things */
	if (mdb_readvar(&(coargs->co_table), "callout_table") == -1) {
		mdb_warn("failed to read 'callout_table'");
		return (DCMD_ERR);
	}
	/* need to get now in nsecs. Approximate with hrtime vars */
	if (mdb_readsym(&(coargs->now), sizeof (hrtime_t), "hrtime_last") !=
	    sizeof (hrtime_t)) {
		if (mdb_readsym(&(coargs->now), sizeof (hrtime_t),
		    "hrtime_base") != sizeof (hrtime_t)) {
			mdb_warn("Could not determine current system time");
			return (DCMD_ERR);
		}
	}

	if (mdb_readvar(&(coargs->ctbits), "callout_table_bits") == -1) {
		mdb_warn("failed to read 'callout_table_bits'");
		return (DCMD_ERR);
	}
	if (mdb_readvar(&(coargs->nsec_per_tick), "nsec_per_tick") == -1) {
		mdb_warn("failed to read 'nsec_per_tick'");
		return (DCMD_ERR);
	}
	return (DCMD_OK);
}

/*
 * dcmd to print callouts.  Optional addr limits to specific table.
 * Parses lots of options that get passed to callbacks for walkers.
 * Has it's own help function.
 */
/*ARGSUSED*/
int
callout(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	callout_data_t coargs;
	/* getopts doesn't help much with stuff like this */
	boolean_t Sflag, Cflag, tflag, aflag, bflag, dflag, kflag;
	char *funcname = NULL;
	char *paramstr = NULL;
	uintptr_t Stmp, Ctmp;	/* for getopt. */
	int retval;

	coargs.flags = COF_DEFAULT;
	Sflag = Cflag = tflag = bflag = aflag = dflag = kflag = FALSE;
	coargs.seqid = -1;

	if (mdb_getopts(argc, argv,
	    'r', MDB_OPT_CLRBITS, COF_NORM, &coargs.flags,
	    'n', MDB_OPT_CLRBITS, COF_REAL, &coargs.flags,
	    'l', MDB_OPT_CLRBITS, COF_SHORT, &coargs.flags,
	    's', MDB_OPT_CLRBITS, COF_LONG, &coargs.flags,
	    'x', MDB_OPT_SETBITS, COF_EXEC, &coargs.flags,
	    'h', MDB_OPT_SETBITS, COF_HIRES, &coargs.flags,
	    'B', MDB_OPT_SETBITS, COF_ABS, &coargs.flags,
	    'E', MDB_OPT_SETBITS, COF_EMPTY, &coargs.flags,
	    'd', MDB_OPT_SETBITS, 1, &dflag,
	    'C', MDB_OPT_UINTPTR_SET, &Cflag, &Ctmp,
	    'S', MDB_OPT_UINTPTR_SET, &Sflag, &Stmp,
	    't', MDB_OPT_UINTPTR_SET, &tflag, (uintptr_t *)&coargs.time,
	    'a', MDB_OPT_UINTPTR_SET, &aflag, (uintptr_t *)&coargs.atime,
	    'b', MDB_OPT_UINTPTR_SET, &bflag, (uintptr_t *)&coargs.btime,
	    'k', MDB_OPT_SETBITS, 1, &kflag,
	    'f', MDB_OPT_STR, &funcname,
	    'p', MDB_OPT_STR, &paramstr,
	    'T', MDB_OPT_SETBITS, COF_TABLE, &coargs.flags,
	    'D', MDB_OPT_SETBITS, COF_EXPREL, &coargs.flags,
	    'L', MDB_OPT_SETBITS, COF_LIST, &coargs.flags,
	    'V', MDB_OPT_SETBITS, COF_VERBOSE, &coargs.flags,
	    'v', MDB_OPT_SETBITS, COF_LONGLIST, &coargs.flags,
	    'i', MDB_OPT_SETBITS, COF_BYIDH, &coargs.flags,
	    'F', MDB_OPT_SETBITS, COF_FREE, &coargs.flags,
	    'H', MDB_OPT_SETBITS, COF_HEAP, &coargs.flags,
	    'Q', MDB_OPT_SETBITS, COF_QUEUE, &coargs.flags,
	    'A', MDB_OPT_SETBITS, COF_ADDR, &coargs.flags,
	    NULL) != argc) {
		return (DCMD_USAGE);
	}

	/* initialize from kernel variables */
	if ((retval = callout_common_init(&coargs)) != DCMD_OK) {
		return (retval);
	}

	/* do some option post-processing */
	if (kflag) {
		coargs.time *= coargs.nsec_per_tick;
		coargs.atime *= coargs.nsec_per_tick;
		coargs.btime *= coargs.nsec_per_tick;
	}

	if (dflag) {
		coargs.time += coargs.now;
		coargs.atime += coargs.now;
		coargs.btime += coargs.now;
	}
	if (Sflag) {
		if (flags & DCMD_ADDRSPEC) {
			mdb_printf("-S option conflicts with explicit"
			    " address\n");
			return (DCMD_USAGE);
		}
		coargs.flags |= COF_SEQID;
		coargs.seqid = (int)Stmp;
	}
	if (Cflag) {
		if (flags & DCMD_ADDRSPEC) {
			mdb_printf("-C option conflicts with explicit"
			    " address\n");
			return (DCMD_USAGE);
		}
		if (coargs.flags & COF_SEQID) {
			mdb_printf("-C and -S are mutually exclusive\n");
			return (DCMD_USAGE);
		}
		coargs.cpu = (cpu_t *)Ctmp;
		if (mdb_vread(&coargs.seqid, sizeof (processorid_t),
		    (uintptr_t)&(coargs.cpu->cpu_seqid)) == -1) {
			mdb_warn("failed to read cpu_t at %p", Ctmp);
			return (DCMD_ERR);
		}
		coargs.flags |= COF_SEQID;
	}
	/* avoid null outputs. */
	if (!(coargs.flags & (COF_REAL | COF_NORM))) {
		coargs.flags |= COF_REAL | COF_NORM;
	}
	if (!(coargs.flags & (COF_LONG | COF_SHORT))) {
		coargs.flags |= COF_LONG | COF_SHORT;
	}
	if (tflag) {
		if (aflag || bflag) {
			mdb_printf("-t and -a|b are mutually exclusive\n");
			return (DCMD_USAGE);
		}
		coargs.flags |= COF_TIME;
	}
	if (aflag) {
		coargs.flags |= COF_AFTER;
	}
	if (bflag) {
		coargs.flags |= COF_BEFORE;
	}
	if ((aflag && bflag) && (coargs.btime <= coargs.atime)) {
		mdb_printf("value for -a must be earlier than the value"
		    " for -b.\n");
		return (DCMD_USAGE);
	}

	if ((coargs.flags & COF_HEAP) && (coargs.flags & COF_QUEUE)) {
		mdb_printf("-H and -Q are mutually exclusive\n");
		return (DCMD_USAGE);
	}

	if (funcname != NULL) {
		GElf_Sym sym;

		if (mdb_lookup_by_name(funcname, &sym) != 0) {
			coargs.funcaddr = mdb_strtoull(funcname);
		} else {
			coargs.funcaddr = sym.st_value;
		}
		coargs.flags |= COF_FUNC;
	}

	if (paramstr != NULL) {
		GElf_Sym sym;

		if (mdb_lookup_by_name(paramstr, &sym) != 0) {
			coargs.param = mdb_strtoull(paramstr);
		} else {
			coargs.param = sym.st_value;
		}
		coargs.flags |= COF_PARAM;
	}

	if (!(flags & DCMD_ADDRSPEC)) {
		/* don't pass "dot" if no addr. */
		addr = NULL;
	}
	if (addr != NULL) {
		/*
		 * a callout table was specified. Ignore -r|n option
		 * to avoid null output.
		 */
		coargs.flags |= (COF_REAL | COF_NORM);
	}

	if (DCMD_HDRSPEC(flags) || (coargs.flags & COF_VERBOSE)) {
		coargs.flags |= COF_THDR | COF_LHDR | COF_CHDR;
	}
	if (coargs.flags & COF_FREE) {
		coargs.flags |= COF_EMPTY;
		/* -F = free callouts, -FL = free lists */
		if (!(coargs.flags & COF_LIST)) {
			coargs.flags |= COF_BYIDH;
		}
	}

	/* walk table, using specialized callback routine. */
	if (mdb_pwalk("callout_table", callout_t_cb, &coargs, addr) == -1) {
		mdb_warn("cannot walk callout_table");
		return (DCMD_ERR);
	}
	return (DCMD_OK);
}


/*
 * Given an extended callout id, dump its information.
 */
/*ARGSUSED*/
int
calloutid(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	callout_data_t coargs;
	callout_table_t *ctptr;
	callout_table_t ct;
	callout_id_t coid;
	callout_t *coptr;
	int tableid;
	callout_id_t xid;
	ulong_t idhash;
	int i, retval;
	const mdb_arg_t *arg;
	size_t size;
	callout_hash_t cot_idhash[CALLOUT_BUCKETS];

	coargs.flags = COF_DEFAULT | COF_BYIDH;
	i = mdb_getopts(argc, argv,
	    'd', MDB_OPT_SETBITS, COF_DECODE, &coargs.flags,
	    'v', MDB_OPT_SETBITS, COF_LONGLIST, &coargs.flags,
	    NULL);
	argc -= i;
	argv += i;

	if (argc != 1) {
		return (DCMD_USAGE);
	}
	arg = &argv[0];

	if (arg->a_type == MDB_TYPE_IMMEDIATE) {
		xid = arg->a_un.a_val;
	} else {
		xid = (callout_id_t)mdb_strtoull(arg->a_un.a_str);
	}

	if (DCMD_HDRSPEC(flags)) {
		coargs.flags |= COF_CHDR;
	}


	/* initialize from kernel variables */
	if ((retval = callout_common_init(&coargs)) != DCMD_OK) {
		return (retval);
	}

	/* we must massage the environment so that the macros will play nice */
#define	callout_table_mask	((1 << coargs.ctbits) - 1)
#define	callout_table_bits	coargs.ctbits
#define	nsec_per_tick		coargs.nsec_per_tick
	tableid = CALLOUT_ID_TO_TABLE(xid);
	idhash = CALLOUT_IDHASH(xid);
#undef	callouts_table_bits
#undef	callout_table_mask
#undef	nsec_per_tick
	coid = CO_PLAIN_ID(xid);

	if (flags & DCMD_ADDRSPEC) {
		mdb_printf("calloutid does not accept explicit address.\n");
		return (DCMD_USAGE);
	}

	if (coargs.flags & COF_DECODE) {
		if (DCMD_HDRSPEC(flags)) {
			mdb_printf("%<u>%3s %1s %2s %-?s %-6s %</u>\n",
			    "SEQ", "T", "XL", "XID", "IDHASH");
		}
		mdb_printf("%-3d %1s %1s%1s %-?llx %-6d\n",
		    TABLE_TO_SEQID(tableid),
		    co_typenames[tableid & CALLOUT_TYPE_MASK],
		    (xid & CALLOUT_EXECUTING) ? "X" : " ",
		    (xid & CALLOUT_LONGTERM) ? "L" : " ",
		    (long long)coid, idhash);
		return (DCMD_OK);
	}

	/* get our table. Note this relies on the types being correct */
	ctptr = coargs.co_table + tableid;
	if (mdb_vread(&ct, sizeof (callout_table_t), (uintptr_t)ctptr) == -1) {
		mdb_warn("failed to read callout_table at %p", ctptr);
		return (DCMD_ERR);
	}
	size = sizeof (callout_hash_t) * CALLOUT_BUCKETS;
	if (ct.ct_idhash != NULL) {
		if (mdb_vread(&(cot_idhash), size,
		    (uintptr_t)ct.ct_idhash) == -1) {
			mdb_warn("failed to read id_hash at %p",
			    ct.ct_idhash);
			return (WALK_ERR);
		}
	}

	/* callout at beginning of hash chain */
	if (ct.ct_idhash == NULL) {
		mdb_printf("id hash chain for this xid is empty\n");
		return (DCMD_ERR);
	}
	coptr = (callout_t *)cot_idhash[idhash].ch_head;
	if (coptr == NULL) {
		mdb_printf("id hash chain for this xid is empty\n");
		return (DCMD_ERR);
	}

	coargs.ndx = tableid;
	coargs.bucket = idhash;

	/* use the walker, luke */
	if (mdb_pwalk("callouts_byid", callouts_cb, &coargs,
	    (uintptr_t)coptr) == -1) {
		mdb_warn("cannot walk callouts at %p", coptr);
		return (WALK_ERR);
	}

	return (DCMD_OK);
}

void
callout_help(void)
{
	mdb_printf("callout: display callouts.\n"
	    "Given a callout table address, display callouts from table.\n"
	    "Without an address, display callouts from all tables.\n"
	    "options:\n"
	    " -r|n : limit display to (r)ealtime or (n)ormal type callouts\n"
	    " -s|l : limit display to (s)hort-term ids or (l)ong-term ids\n"
	    " -x : limit display to callouts which are executing\n"
	    " -h : limit display to callouts based on hrestime\n"
	    " -B : limit display to callouts based on absolute time\n"
	    " -t|a|b nsec: limit display to callouts that expire a(t) time,"
	    " (a)fter time,\n     or (b)efore time. Use -a and -b together "
	    " to specify a range.\n     For \"now\", use -d[t|a|b] 0.\n"
	    " -d : interpret time option to -t|a|b as delta from current time\n"
	    " -k : use ticks instead of nanoseconds as arguments to"
	    " -t|a|b. Note that\n     ticks are less accurate and may not"
	    " match other tick times (ie: lbolt).\n"
	    " -D : display exiration time as delta from current time\n"
	    " -S seqid : limit display to callouts for this cpu sequence id\n"
	    " -C addr :  limit display to callouts for this cpu pointer\n"
	    " -f name|addr : limit display to callouts with this function\n"
	    " -p name|addr : limit display to callouts functions with this"
	    " parameter\n"
	    " -T : display the callout table itself, instead of callouts\n"
	    " -L : display callout lists instead of callouts\n"
	    " -E : with -T or L, display empty data structures.\n"
	    " -i : traverse callouts by id hash instead of list hash\n"
	    " -F : walk free callout list (free list with -i) instead\n"
	    " -v : display more info for each item\n"
	    " -V : show details of each level of info as it is traversed\n"
	    " -H : limit display to callouts in the callout heap\n"
	    " -Q : limit display to callouts in the callout queue\n"
	    " -A : show only addresses. Useful for pipelines.\n");
}

void
calloutid_help(void)
{
	mdb_printf("calloutid: display callout by id.\n"
	    "Given an extended callout id, display the callout infomation.\n"
	    "options:\n"
	    " -d : do not dereference callout, just decode the id.\n"
	    " -v : verbose display more info about the callout\n");
}

/*ARGSUSED*/
int
class(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	long num_classes, i;
	sclass_t *class_tbl;
	GElf_Sym g_sclass;
	char class_name[PC_CLNMSZ];
	size_t tbl_size;

	if (mdb_lookup_by_name("sclass", &g_sclass) == -1) {
		mdb_warn("failed to find symbol sclass\n");
		return (DCMD_ERR);
	}

	tbl_size = (size_t)g_sclass.st_size;
	num_classes = tbl_size / (sizeof (sclass_t));
	class_tbl = mdb_alloc(tbl_size, UM_SLEEP | UM_GC);

	if (mdb_readsym(class_tbl, tbl_size, "sclass") == -1) {
		mdb_warn("failed to read sclass");
		return (DCMD_ERR);
	}

	mdb_printf("%<u>%4s %-10s %-24s %-24s%</u>\n", "SLOT", "NAME",
	    "INIT FCN", "CLASS FCN");

	for (i = 0; i < num_classes; i++) {
		if (mdb_vread(class_name, sizeof (class_name),
		    (uintptr_t)class_tbl[i].cl_name) == -1)
			(void) strcpy(class_name, "???");

		mdb_printf("%4ld %-10s %-24a %-24a\n", i, class_name,
		    class_tbl[i].cl_init, class_tbl[i].cl_funcs);
	}

	return (DCMD_OK);
}

#define	FSNAMELEN	32	/* Max len of FS name we read from vnodeops */

int
vnode2path(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	uintptr_t rootdir;
	vnode_t vn;
	char buf[MAXPATHLEN];

	uint_t opt_F = FALSE;

	if (mdb_getopts(argc, argv,
	    'F', MDB_OPT_SETBITS, TRUE, &opt_F, NULL) != argc)
		return (DCMD_USAGE);

	if (!(flags & DCMD_ADDRSPEC)) {
		mdb_warn("expected explicit vnode_t address before ::\n");
		return (DCMD_USAGE);
	}

	if (mdb_readvar(&rootdir, "rootdir") == -1) {
		mdb_warn("failed to read rootdir");
		return (DCMD_ERR);
	}

	if (mdb_vnode2path(addr, buf, sizeof (buf)) == -1)
		return (DCMD_ERR);

	if (*buf == '\0') {
		mdb_printf("??\n");
		return (DCMD_OK);
	}

	mdb_printf("%s", buf);
	if (opt_F && buf[strlen(buf)-1] != '/' &&
	    mdb_vread(&vn, sizeof (vn), addr) == sizeof (vn))
		mdb_printf("%c", mdb_vtype2chr(vn.v_type, 0));
	mdb_printf("\n");

	return (DCMD_OK);
}

int
ld_walk_init(mdb_walk_state_t *wsp)
{
	wsp->walk_data = (void *)wsp->walk_addr;
	return (WALK_NEXT);
}

int
ld_walk_step(mdb_walk_state_t *wsp)
{
	int status;
	lock_descriptor_t ld;

	if (mdb_vread(&ld, sizeof (lock_descriptor_t), wsp->walk_addr) == -1) {
		mdb_warn("couldn't read lock_descriptor_t at %p\n",
		    wsp->walk_addr);
		return (WALK_ERR);
	}

	status = wsp->walk_callback(wsp->walk_addr, &ld, wsp->walk_cbdata);
	if (status == WALK_ERR)
		return (WALK_ERR);

	wsp->walk_addr = (uintptr_t)ld.l_next;
	if (wsp->walk_addr == (uintptr_t)wsp->walk_data)
		return (WALK_DONE);

	return (status);
}

int
lg_walk_init(mdb_walk_state_t *wsp)
{
	GElf_Sym sym;

	if (mdb_lookup_by_name("lock_graph", &sym) == -1) {
		mdb_warn("failed to find symbol 'lock_graph'\n");
		return (WALK_ERR);
	}

	wsp->walk_addr = (uintptr_t)sym.st_value;
	wsp->walk_data = (void *)(uintptr_t)(sym.st_value + sym.st_size);

	return (WALK_NEXT);
}

typedef struct lg_walk_data {
	uintptr_t startaddr;
	mdb_walk_cb_t callback;
	void *data;
} lg_walk_data_t;

/*
 * We can't use ::walk lock_descriptor directly, because the head of each graph
 * is really a dummy lock.  Rather than trying to dynamically determine if this
 * is a dummy node or not, we just filter out the initial element of the
 * list.
 */
static int
lg_walk_cb(uintptr_t addr, const void *data, void *priv)
{
	lg_walk_data_t *lw = priv;

	if (addr != lw->startaddr)
		return (lw->callback(addr, data, lw->data));

	return (WALK_NEXT);
}

int
lg_walk_step(mdb_walk_state_t *wsp)
{
	graph_t *graph;
	lg_walk_data_t lw;

	if (wsp->walk_addr >= (uintptr_t)wsp->walk_data)
		return (WALK_DONE);

	if (mdb_vread(&graph, sizeof (graph), wsp->walk_addr) == -1) {
		mdb_warn("failed to read graph_t at %p", wsp->walk_addr);
		return (WALK_ERR);
	}

	wsp->walk_addr += sizeof (graph);

	if (graph == NULL)
		return (WALK_NEXT);

	lw.callback = wsp->walk_callback;
	lw.data = wsp->walk_cbdata;

	lw.startaddr = (uintptr_t)&(graph->active_locks);
	if (mdb_pwalk("lock_descriptor", lg_walk_cb, &lw, lw.startaddr)) {
		mdb_warn("couldn't walk lock_descriptor at %p\n", lw.startaddr);
		return (WALK_ERR);
	}

	lw.startaddr = (uintptr_t)&(graph->sleeping_locks);
	if (mdb_pwalk("lock_descriptor", lg_walk_cb, &lw, lw.startaddr)) {
		mdb_warn("couldn't walk lock_descriptor at %p\n", lw.startaddr);
		return (WALK_ERR);
	}

	return (WALK_NEXT);
}

/*
 * The space available for the path corresponding to the locked vnode depends
 * on whether we are printing 32- or 64-bit addresses.
 */
#ifdef _LP64
#define	LM_VNPATHLEN	20
#else
#define	LM_VNPATHLEN	30
#endif

typedef struct mdb_lminfo_proc {
	struct {
		char		u_comm[MAXCOMLEN + 1];
	} p_user;
} mdb_lminfo_proc_t;

/*ARGSUSED*/
static int
lminfo_cb(uintptr_t addr, const void *data, void *priv)
{
	const lock_descriptor_t *ld = data;
	char buf[LM_VNPATHLEN];
	mdb_lminfo_proc_t p;
	uintptr_t paddr = 0;

	if (ld->l_flock.l_pid != 0)
		paddr = mdb_pid2proc(ld->l_flock.l_pid, NULL);

	if (paddr != 0)
		mdb_ctf_vread(&p, "proc_t", "mdb_lminfo_proc_t", paddr, 0);

	mdb_printf("%-?p %2s %04x %6d %-16s %-?p ",
	    addr, ld->l_type == F_RDLCK ? "RD" :
	    ld->l_type == F_WRLCK ? "WR" : "??",
	    ld->l_state, ld->l_flock.l_pid,
	    ld->l_flock.l_pid == 0 ? "<kernel>" :
	    paddr == 0 ? "<defunct>" : p.p_user.u_comm, ld->l_vnode);

	mdb_vnode2path((uintptr_t)ld->l_vnode, buf,
	    sizeof (buf));
	mdb_printf("%s\n", buf);

	return (WALK_NEXT);
}

/*ARGSUSED*/
int
lminfo(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	if (DCMD_HDRSPEC(flags))
		mdb_printf("%<u>%-?s %2s %4s %6s %-16s %-?s %s%</u>\n",
		    "ADDR", "TP", "FLAG", "PID", "COMM", "VNODE", "PATH");

	return (mdb_pwalk("lock_graph", lminfo_cb, NULL, NULL));
}

/*ARGSUSED*/
int
whereopen_fwalk(uintptr_t addr, struct file *f, uintptr_t *target)
{
	if ((uintptr_t)f->f_vnode == *target) {
		mdb_printf("file %p\n", addr);
		*target = NULL;
	}

	return (WALK_NEXT);
}

/*ARGSUSED*/
int
whereopen_pwalk(uintptr_t addr, void *ignored, uintptr_t *target)
{
	uintptr_t t = *target;

	if (mdb_pwalk("file", (mdb_walk_cb_t)whereopen_fwalk, &t, addr) == -1) {
		mdb_warn("couldn't file walk proc %p", addr);
		return (WALK_ERR);
	}

	if (t == NULL)
		mdb_printf("%p\n", addr);

	return (WALK_NEXT);
}

/*ARGSUSED*/
int
whereopen(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	uintptr_t target = addr;

	if (!(flags & DCMD_ADDRSPEC) || addr == NULL)
		return (DCMD_USAGE);

	if (mdb_walk("proc", (mdb_walk_cb_t)whereopen_pwalk, &target) == -1) {
		mdb_warn("can't proc walk");
		return (DCMD_ERR);
	}

	return (DCMD_OK);
}

typedef struct datafmt {
	char	*hdr1;
	char	*hdr2;
	char	*dashes;
	char	*fmt;
} datafmt_t;

static datafmt_t kmemfmt[] = {
	{ "cache                    ", "name                     ",
	"-------------------------", "%-25s "				},
	{ "   buf",	"  size",	"------",	"%6u "		},
	{ "   buf",	"in use",	"------",	"%6u "		},
	{ "   buf",	" total",	"------",	"%6u "		},
	{ "   memory",	"   in use",	"----------",	"%10lu%c "	},
	{ "    alloc",	"  succeed",	"---------",	"%9u "		},
	{ "alloc",	" fail",	"-----",	"%5u "		},
	{ NULL,		NULL,		NULL,		NULL		}
};

static datafmt_t vmemfmt[] = {
	{ "vmem                     ", "name                     ",
	"-------------------------", "%-*s "				},
	{ "   memory",	"   in use",	"----------",	"%9llu%c "	},
	{ "    memory",	"     total",	"-----------",	"%10llu%c "	},
	{ "   memory",	"   import",	"----------",	"%9llu%c "	},
	{ "    alloc",	"  succeed",	"---------",	"%9llu "	},
	{ "alloc",	" fail",	"-----",	"%5llu "	},
	{ NULL,		NULL,		NULL,		NULL		}
};

/*ARGSUSED*/
static int
kmastat_cpu_avail(uintptr_t addr, const kmem_cpu_cache_t *ccp, int *avail)
{
	short rounds, prounds;

	if (KMEM_DUMPCC(ccp)) {
		rounds = ccp->cc_dump_rounds;
		prounds = ccp->cc_dump_prounds;
	} else {
		rounds = ccp->cc_rounds;
		prounds = ccp->cc_prounds;
	}
	if (rounds > 0)
		*avail += rounds;
	if (prounds > 0)
		*avail += prounds;

	return (WALK_NEXT);
}

/*ARGSUSED*/
static int
kmastat_cpu_alloc(uintptr_t addr, const kmem_cpu_cache_t *ccp, int *alloc)
{
	*alloc += ccp->cc_alloc;

	return (WALK_NEXT);
}

/*ARGSUSED*/
static int
kmastat_slab_avail(uintptr_t addr, const kmem_slab_t *sp, int *avail)
{
	*avail += sp->slab_chunks - sp->slab_refcnt;

	return (WALK_NEXT);
}

typedef struct kmastat_vmem {
	uintptr_t kv_addr;
	struct kmastat_vmem *kv_next;
	size_t kv_meminuse;
	int kv_alloc;
	int kv_fail;
} kmastat_vmem_t;

typedef struct kmastat_args {
	kmastat_vmem_t **ka_kvpp;
	uint_t ka_shift;
} kmastat_args_t;

static int
kmastat_cache(uintptr_t addr, const kmem_cache_t *cp, kmastat_args_t *kap)
{
	kmastat_vmem_t **kvpp = kap->ka_kvpp;
	kmastat_vmem_t *kv;
	datafmt_t *dfp = kmemfmt;
	int magsize;

	int avail, alloc, total;
	size_t meminuse = (cp->cache_slab_create - cp->cache_slab_destroy) *
	    cp->cache_slabsize;

	mdb_walk_cb_t cpu_avail = (mdb_walk_cb_t)kmastat_cpu_avail;
	mdb_walk_cb_t cpu_alloc = (mdb_walk_cb_t)kmastat_cpu_alloc;
	mdb_walk_cb_t slab_avail = (mdb_walk_cb_t)kmastat_slab_avail;

	magsize = kmem_get_magsize(cp);

	alloc = cp->cache_slab_alloc + cp->cache_full.ml_alloc;
	avail = cp->cache_full.ml_total * magsize;
	total = cp->cache_buftotal;

	(void) mdb_pwalk("kmem_cpu_cache", cpu_alloc, &alloc, addr);
	(void) mdb_pwalk("kmem_cpu_cache", cpu_avail, &avail, addr);
	(void) mdb_pwalk("kmem_slab_partial", slab_avail, &avail, addr);

	for (kv = *kvpp; kv != NULL; kv = kv->kv_next) {
		if (kv->kv_addr == (uintptr_t)cp->cache_arena)
			goto out;
	}

	kv = mdb_zalloc(sizeof (kmastat_vmem_t), UM_SLEEP | UM_GC);
	kv->kv_next = *kvpp;
	kv->kv_addr = (uintptr_t)cp->cache_arena;
	*kvpp = kv;
out:
	kv->kv_meminuse += meminuse;
	kv->kv_alloc += alloc;
	kv->kv_fail += cp->cache_alloc_fail;

	mdb_printf((dfp++)->fmt, cp->cache_name);
	mdb_printf((dfp++)->fmt, cp->cache_bufsize);
	mdb_printf((dfp++)->fmt, total - avail);
	mdb_printf((dfp++)->fmt, total);
	mdb_printf((dfp++)->fmt, meminuse >> kap->ka_shift,
	    kap->ka_shift == GIGS ? 'G' : kap->ka_shift == MEGS ? 'M' :
	    kap->ka_shift == KILOS ? 'K' : 'B');
	mdb_printf((dfp++)->fmt, alloc);
	mdb_printf((dfp++)->fmt, cp->cache_alloc_fail);
	mdb_printf("\n");

	return (WALK_NEXT);
}

static int
kmastat_vmem_totals(uintptr_t addr, const vmem_t *v, kmastat_args_t *kap)
{
	kmastat_vmem_t *kv = *kap->ka_kvpp;
	size_t len;

	while (kv != NULL && kv->kv_addr != addr)
		kv = kv->kv_next;

	if (kv == NULL || kv->kv_alloc == 0)
		return (WALK_NEXT);

	len = MIN(17, strlen(v->vm_name));

	mdb_printf("Total [%s]%*s %6s %6s %6s %10lu%c %9u %5u\n", v->vm_name,
	    17 - len, "", "", "", "",
	    kv->kv_meminuse >> kap->ka_shift,
	    kap->ka_shift == GIGS ? 'G' : kap->ka_shift == MEGS ? 'M' :
	    kap->ka_shift == KILOS ? 'K' : 'B', kv->kv_alloc, kv->kv_fail);

	return (WALK_NEXT);
}

/*ARGSUSED*/
static int
kmastat_vmem(uintptr_t addr, const vmem_t *v, const uint_t *shiftp)
{
	datafmt_t *dfp = vmemfmt;
	const vmem_kstat_t *vkp = &v->vm_kstat;
	uintptr_t paddr;
	vmem_t parent;
	int ident = 0;

	for (paddr = (uintptr_t)v->vm_source; paddr != NULL; ident += 4) {
		if (mdb_vread(&parent, sizeof (parent), paddr) == -1) {
			mdb_warn("couldn't trace %p's ancestry", addr);
			ident = 0;
			break;
		}
		paddr = (uintptr_t)parent.vm_source;
	}

	mdb_printf("%*s", ident, "");
	mdb_printf((dfp++)->fmt, 25 - ident, v->vm_name);
	mdb_printf((dfp++)->fmt, vkp->vk_mem_inuse.value.ui64 >> *shiftp,
	    *shiftp == GIGS ? 'G' : *shiftp == MEGS ? 'M' :
	    *shiftp == KILOS ? 'K' : 'B');
	mdb_printf((dfp++)->fmt, vkp->vk_mem_total.value.ui64 >> *shiftp,
	    *shiftp == GIGS ? 'G' : *shiftp == MEGS ? 'M' :
	    *shiftp == KILOS ? 'K' : 'B');
	mdb_printf((dfp++)->fmt, vkp->vk_mem_import.value.ui64 >> *shiftp,
	    *shiftp == GIGS ? 'G' : *shiftp == MEGS ? 'M' :
	    *shiftp == KILOS ? 'K' : 'B');
	mdb_printf((dfp++)->fmt, vkp->vk_alloc.value.ui64);
	mdb_printf((dfp++)->fmt, vkp->vk_fail.value.ui64);

	mdb_printf("\n");

	return (WALK_NEXT);
}

/*ARGSUSED*/
int
kmastat(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
{
	kmastat_vmem_t *kv = NULL;
	datafmt_t *dfp;
	kmastat_args_t ka;

	ka.ka_shift = 0;
	if (mdb_getopts(argc, argv,
	    'k', MDB_OPT_SETBITS, KILOS, &ka.ka_shift,
	    'm', MDB_OPT_SETBITS, MEGS, &ka.ka_shift,
	    'g', MDB_OPT_SETBITS, GIGS, &ka.ka_shift, NULL) != argc)
		return (DCMD_USAGE);

	for (dfp = kmemfmt; dfp->hdr1 != NULL; dfp++)
		mdb_printf("%s ", dfp->hdr1);
	mdb_printf("\n");

	for (dfp = kmemfmt; dfp->hdr1 != NULL; dfp++)
		mdb_printf("%s ", dfp->hdr2);
	mdb_printf("\n");

	for (dfp = kmemfmt; dfp->hdr1 != NULL; dfp++)
		mdb_printf("%s ", dfp->dashes);
	mdb_printf("\n");

	ka.ka_kvpp = &kv;
	if (mdb_walk("kmem_cache", (mdb_walk_cb_t)kmastat_cache, &ka) == -1) {
		mdb_warn("can't walk 'kmem_cache'");
		return (DCMD_ERR);
	}

	for (dfp = kmemfmt; dfp->hdr1 != NULL; dfp++)
		mdb_printf("%s ", dfp->dashes);
	mdb_printf("\n");

	if (mdb_walk("vmem", (mdb_walk_cb_t)kmastat_vmem_totals, &ka) == -1) {
		mdb_warn("can't walk 'vmem'");
		return (DCMD_ERR);
	}

	for (dfp = kmemfmt; dfp->hdr1 != NULL; dfp++)
		mdb_printf("%s ", dfp->dashes);
	mdb_printf("\n");

	mdb_printf("\n");

	for (dfp = vmemfmt; dfp->hdr1 != NULL; dfp++)
		mdb_printf("%s ", dfp->hdr1);
	mdb_printf("\n");

	for (dfp = vmemfmt; dfp->hdr1 != NULL; dfp++)
		mdb_printf("%s ", dfp->hdr2);
	mdb_printf("\n");

	for (dfp = vmemfmt; dfp->hdr1 != NULL; dfp++)
		mdb_printf("%s ", dfp->dashes);
	mdb_printf("\n");

	if (mdb_walk("vmem", (mdb_walk_cb_t)kmastat_vmem, &ka.ka_shift) == -1) {
		mdb_warn("can't walk 'vmem'");
		return (DCMD_ERR);
	}

	for (dfp = vmemfmt; dfp->hdr1 != NULL; dfp++)
		mdb_printf("%s ", dfp->dashes);
	mdb_printf("\n");
	return (DCMD_OK);
}

/*
 * Our ::kgrep callback scans the entire kernel VA space (kas).  kas is made
 * up of a set of 'struct seg's.  We could just scan each seg en masse, but
 * unfortunately, a few of the segs are both large and sparse, so we could
 * spend quite a bit of time scanning VAs which have no backing pages.
 *
 * So for the few very sparse segs, we skip the segment itself, and scan
 * the allocated vmem_segs in the vmem arena which manages that part of kas.
 * Currently, we do this for:
 *
 *	SEG		VMEM ARENA
 *	kvseg		heap_arena
 *	kvseg32		heap32_arena
 *	kvseg_core	heap_core_arena
 *
 * In addition, we skip the segkpm segment in its entirety, since it is very
 * sparse, and contains no new kernel data.
 */
typedef struct kgrep_walk_data {
	kgrep_cb_func *kg_cb;
	void *kg_cbdata;
	uintptr_t kg_kvseg;
	uintptr_t kg_kvseg32;
	uintptr_t kg_kvseg_core;
	uintptr_t kg_segkpm;
	uintptr_t kg_heap_lp_base;
	uintptr_t kg_heap_lp_end;
} kgrep_walk_data_t;

static int
kgrep_walk_seg(uintptr_t addr, const struct seg *seg, kgrep_walk_data_t *kg)
{
	uintptr_t base = (uintptr_t)seg->s_base;

	if (addr == kg->kg_kvseg || addr == kg->kg_kvseg32 ||
	    addr == kg->kg_kvseg_core)
		return (WALK_NEXT);

	if ((uintptr_t)seg->s_ops == kg->kg_segkpm)
		return (WALK_NEXT);

	return (kg->kg_cb(base, base + seg->s_size, kg->kg_cbdata));
}

/*ARGSUSED*/
static int
kgrep_walk_vseg(uintptr_t addr, const vmem_seg_t *seg, kgrep_walk_data_t *kg)
{
	/*
	 * skip large page heap address range - it is scanned by walking
	 * allocated vmem_segs in the heap_lp_arena
	 */
	if (seg->vs_start == kg->kg_heap_lp_base &&
	    seg->vs_end == kg->kg_heap_lp_end)
		return (WALK_NEXT);

	return (kg->kg_cb(seg->vs_start, seg->vs_end, kg->kg_cbdata));
}

/*ARGSUSED*/
static int
kgrep_xwalk_vseg(uintptr_t addr, const vmem_seg_t *seg, kgrep_walk_data_t *kg)
{
	return (kg->kg_cb(seg->vs_start, seg->vs_end, kg->kg_cbdata));
}

static int
kgrep_walk_vmem(uintptr_t addr, const vmem_t *vmem, kgrep_walk_data_t *kg)
{
	mdb_walk_cb_t walk_vseg = (mdb_walk_cb_t)kgrep_walk_vseg;

	if (strcmp(vmem->vm_name, "heap") != 0 &&
	    strcmp(vmem->vm_name, "heap32") != 0 &&
	    strcmp(vmem->vm_name, "heap_core") != 0 &&
	    strcmp(vmem->vm_name, "heap_lp") != 0)
		return (WALK_NEXT);

	if (strcmp(vmem->vm_name, "heap_lp") == 0)
		walk_vseg = (mdb_walk_cb_t)kgrep_xwalk_vseg;

	if (mdb_pwalk("vmem_alloc", walk_vseg, kg, addr) == -1) {
		mdb_warn("couldn't walk vmem_alloc for vmem %p", addr);
		return (WALK_ERR);
	}

	return (WALK_NEXT);
}

int
kgrep_subr(kgrep_cb_func *cb, void *cbdata)
{
	GElf_Sym kas, kvseg, kvseg32, kvseg_core, segkpm;
	kgrep_walk_data_t kg;

	if (mdb_get_state() == MDB_STATE_RUNNING) {
		mdb_warn("kgrep can only be run on a system "
		    "dump or under kmdb; see dumpadm(1M)\n");
		return (DCMD_ERR);
	}

	if (mdb_lookup_by_name("kas", &kas) == -1) {
		mdb_warn("failed to locate 'kas' symbol\n");
		return (DCMD_ERR);
	}

	if (mdb_lookup_by_name("kvseg", &kvseg) == -1) {
		mdb_warn("failed to locate 'kvseg' symbol\n");
		return (DCMD_ERR);
	}

	if (mdb_lookup_by_name("kvseg32", &kvseg32) == -1) {
		mdb_warn("failed to locate 'kvseg32' symbol\n");
		return (DCMD_ERR);
	}

	if (mdb_lookup_by_name("kvseg_core", &kvseg_core) == -1) {
		mdb_warn("failed to locate 'kvseg_core' symbol\n");
		return (DCMD_ERR);
	}

	if (mdb_lookup_by_name("segkpm_ops", &segkpm) == -1) {
		mdb_warn("failed to locate 'segkpm_ops' symbol\n");
		return (DCMD_ERR);
	}

	if (mdb_readvar(&kg.kg_heap_lp_base, "heap_lp_base") == -1) {
		mdb_warn("failed to read 'heap_lp_base'\n");
		return (DCMD_ERR);
	}

	if (mdb_readvar(&kg.kg_heap_lp_end, "heap_lp_end") == -1) {
		mdb_warn("failed to read 'heap_lp_end'\n");
		return (DCMD_ERR);
	}

	kg.kg_cb = cb;
	kg.kg_cbdata = cbdata;
	kg.kg_kvseg = (uintptr_t)kvseg.st_value;
	kg.kg_kvseg32 = (uintptr_t)kvseg32.st_value;
	kg.kg_kvseg_core = (uintptr_t)kvseg_core.st_value;
	kg.kg_segkpm = (uintptr_t)segkpm.st_value;

	if (mdb_pwalk("seg", (mdb_walk_cb_t)kgrep_walk_seg,
	    &kg, kas.st_value) == -1) {
		mdb_warn("failed to walk kas segments");
		return (DCMD_ERR);
	}

	if (mdb_walk("vmem", (mdb_walk_cb_t)kgrep_walk_vmem, &kg) == -1) {
		mdb_warn("failed to walk heap/heap32 vmem arenas");
		return (DCMD_ERR);
	}

	return (DCMD_OK);
}

size_t
kgrep_subr_pagesize(void)
{
	return (PAGESIZE);
}

typedef struct file_walk_data {
	struct uf_entry *fw_flist;
	int fw_flistsz;
	int fw_ndx;
	int fw_nofiles;
} file_walk_data_t;

typedef struct mdb_file_proc {
	struct {
		struct {
			int			fi_nfiles;
			uf_entry_t *volatile	fi_list;
		} u_finfo;
	} p_user;
} mdb_file_proc_t;

int
file_walk_init(mdb_walk_state_t *wsp)
{
	file_walk_data_t *fw;
	mdb_file_proc_t p;

	if (wsp->walk_addr == NULL) {
		mdb_warn("file walk doesn't support global walks\n");
		return (WALK_ERR);
	}

	fw = mdb_alloc(sizeof (file_walk_data_t), UM_SLEEP);

	if (mdb_ctf_vread(&p, "proc_t", "mdb_file_proc_t",
	    wsp->walk_addr, 0) == -1) {
		mdb_free(fw, sizeof (file_walk_data_t));
		mdb_warn("failed to read proc structure at %p", wsp->walk_addr);
		return (WALK_ERR);
	}

	if (p.p_user.u_finfo.fi_nfiles == 0) {
		mdb_free(fw, sizeof (file_walk_data_t));
		return (WALK_DONE);
	}

	fw->fw_nofiles = p.p_user.u_finfo.fi_nfiles;
	fw->fw_flistsz = sizeof (struct uf_entry) * fw->fw_nofiles;
	fw->fw_flist = mdb_alloc(fw->fw_flistsz, UM_SLEEP);

	if (mdb_vread(fw->fw_flist, fw->fw_flistsz,
	    (uintptr_t)p.p_user.u_finfo.fi_list) == -1) {
		mdb_warn("failed to read file array at %p",
		    p.p_user.u_finfo.fi_list);
		mdb_free(fw->fw_flist, fw->fw_flistsz);
		mdb_free(fw, sizeof (file_walk_data_t));
		return (WALK_ERR);
	}

	fw->fw_ndx = 0;
	wsp->walk_data = fw;

	return (WALK_NEXT);
}

int
file_walk_step(mdb_walk_state_t *wsp)
{
	file_walk_data_t *fw = (file_walk_data_t *)wsp->walk_data;
	struct file file;
	uintptr_t fp;

again:
	if (fw->fw_ndx == fw->fw_nofiles)
		return (WALK_DONE);

	if ((fp = (uintptr_t)fw->fw_flist[fw->fw_ndx++].uf_file) == NULL)
		goto again;

	(void) mdb_vread(&file, sizeof (file), (uintptr_t)fp);
	return (wsp->walk_callback(fp, &file, wsp->walk_cbdata));
}

int
allfile_walk_step(mdb_walk_state_t *wsp)
{
	file_walk_data_t *fw = (file_walk_data_t *)wsp->walk_data;
	struct file file;
	uintptr_t fp;

	if (fw->fw_ndx == fw->fw_nofiles)
		return (WALK_DONE);

	if ((fp = (uintptr_t)fw->fw_flist[fw->fw_ndx++].uf_file) != NULL)
		(void) mdb_vread(&file, sizeof (file), (uintptr_t)fp);
	else
		bzero(&file, sizeof (file));

	return (wsp->walk_callback(fp, &file, wsp->walk_cbdata));
}

void
file_walk_fini(mdb_walk_state_t *wsp)
{
	file_walk_data_t *fw = (file_walk_data_t *)wsp->walk_data;

	mdb_free(fw->fw_flist, fw->fw_flistsz);
	mdb_free(fw, sizeof (file_walk_data_t));
}

int
port_walk_init(mdb_walk_state_t *wsp)
{
	if (wsp->walk_addr == NULL) {
		mdb_warn("port walk doesn't support global walks\n");
		return (WALK_ERR);