xref: /freebsd/sys/kern/subr_witness.c (revision 6b3455a7665208c366849f0b2b3bc916fb97516e)

/*-
 * Copyright (c) 1998 Berkeley Software Design, Inc. All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. Berkeley Software Design Inc's name may not be used to endorse or
 *    promote products derived from this software without specific prior
 *    written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY BERKELEY SOFTWARE DESIGN INC ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL BERKELEY SOFTWARE DESIGN INC BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 *	from BSDI $Id: mutex_witness.c,v 1.1.2.20 2000/04/27 03:10:27 cp Exp $
 *	and BSDI $Id: synch_machdep.c,v 2.3.2.39 2000/04/27 03:10:25 cp Exp $
 */

/*
 * Implementation of the `witness' lock verifier.  Originally implemented for
 * mutexes in BSD/OS.  Extended to handle generic lock objects and lock
 * classes in FreeBSD.
 */

/*
 *	Main Entry: witness
 *	Pronunciation: 'wit-n&s
 *	Function: noun
 *	Etymology: Middle English witnesse, from Old English witnes knowledge,
 *	    testimony, witness, from 2wit
 *	Date: before 12th century
 *	1 : attestation of a fact or event : TESTIMONY
 *	2 : one that gives evidence; specifically : one who testifies in
 *	    a cause or before a judicial tribunal
 *	3 : one asked to be present at a transaction so as to be able to
 *	    testify to its having taken place
 *	4 : one who has personal knowledge of something
 *	5 a : something serving as evidence or proof : SIGN
 *	  b : public affirmation by word or example of usually
 *	      religious faith or conviction <the heroic witness to divine
 *	      life -- Pilot>
 *	6 capitalized : a member of the Jehovah's Witnesses
 */

/*
 * Special rules concerning Giant and lock orders:
 *
 * 1) Giant must be acquired before any other mutexes.  Stated another way,
 *    no other mutex may be held when Giant is acquired.
 *
 * 2) Giant must be released when blocking on a sleepable lock.
 *
 * This rule is less obvious, but is a result of Giant providing the same
 * semantics as spl().  Basically, when a thread sleeps, it must release
 * Giant.  When a thread blocks on a sleepable lock, it sleeps.  Hence rule
 * 2).
 *
 * 3) Giant may be acquired before or after sleepable locks.
 *
 * This rule is also not quite as obvious.  Giant may be acquired after
 * a sleepable lock because it is a non-sleepable lock and non-sleepable
 * locks may always be acquired while holding a sleepable lock.  The second
 * case, Giant before a sleepable lock, follows from rule 2) above.  Suppose
 * you have two threads T1 and T2 and a sleepable lock X.  Suppose that T1
 * acquires X and blocks on Giant.  Then suppose that T2 acquires Giant and
 * blocks on X.  When T2 blocks on X, T2 will release Giant allowing T1 to
 * execute.  Thus, acquiring Giant both before and after a sleepable lock
 * will not result in a lock order reversal.
 */

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

#include "opt_ddb.h"
#include "opt_witness.h"

#include <sys/param.h>
#include <sys/bus.h>
#include <sys/kdb.h>
#include <sys/kernel.h>
#include <sys/ktr.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/mutex.h>
#include <sys/proc.h>
#include <sys/sysctl.h>
#include <sys/systm.h>

#include <ddb/ddb.h>

#include <machine/stdarg.h>

/* Define this to check for blessed mutexes */
#undef BLESSING

#define WITNESS_COUNT 200
#define WITNESS_CHILDCOUNT (WITNESS_COUNT * 4)
/*
 * XXX: This is somewhat bogus, as we assume here that at most 1024 threads
 * will hold LOCK_NCHILDREN * 2 locks.  We handle failure ok, and we should
 * probably be safe for the most part, but it's still a SWAG.
 */
#define LOCK_CHILDCOUNT (MAXCPU + 1024) * 2

#define	WITNESS_NCHILDREN 6

struct witness_child_list_entry;

struct witness {
	const	char *w_name;
	struct	lock_class *w_class;
	STAILQ_ENTRY(witness) w_list;		/* List of all witnesses. */
	STAILQ_ENTRY(witness) w_typelist;	/* Witnesses of a type. */
	struct	witness_child_list_entry *w_children;	/* Great evilness... */
	const	char *w_file;
	int	w_line;
	u_int	w_level;
	u_int	w_refcount;
	u_char	w_Giant_squawked:1;
	u_char	w_other_squawked:1;
	u_char	w_same_squawked:1;
	u_char	w_displayed:1;
};

struct witness_child_list_entry {
	struct	witness_child_list_entry *wcl_next;
	struct	witness *wcl_children[WITNESS_NCHILDREN];
	u_int	wcl_count;
};

STAILQ_HEAD(witness_list, witness);

#ifdef BLESSING
struct witness_blessed {
	const	char *b_lock1;
	const	char *b_lock2;
};
#endif

struct witness_order_list_entry {
	const	char *w_name;
	struct	lock_class *w_class;
};

#ifdef BLESSING
static int	blessed(struct witness *, struct witness *);
#endif
static int	depart(struct witness *w);
static struct	witness *enroll(const char *description,
				struct lock_class *lock_class);
static int	insertchild(struct witness *parent, struct witness *child);
static int	isitmychild(struct witness *parent, struct witness *child);
static int	isitmydescendant(struct witness *parent, struct witness *child);
static int	itismychild(struct witness *parent, struct witness *child);
static int	rebalancetree(struct witness_list *list);
static void	removechild(struct witness *parent, struct witness *child);
static int	reparentchildren(struct witness *newparent,
		    struct witness *oldparent);
static int	sysctl_debug_witness_watch(SYSCTL_HANDLER_ARGS);
static void	witness_displaydescendants(void(*)(const char *fmt, ...),
					   struct witness *, int indent);
static const char *fixup_filename(const char *file);
static void	witness_leveldescendents(struct witness *parent, int level);
static void	witness_levelall(void);
static struct	witness *witness_get(void);
static void	witness_free(struct witness *m);
static struct	witness_child_list_entry *witness_child_get(void);
static void	witness_child_free(struct witness_child_list_entry *wcl);
static struct	lock_list_entry *witness_lock_list_get(void);
static void	witness_lock_list_free(struct lock_list_entry *lle);
static struct	lock_instance *find_instance(struct lock_list_entry *lock_list,
					     struct lock_object *lock);
static void	witness_list_lock(struct lock_instance *instance);
#ifdef DDB
static void	witness_list(struct thread *td);
static void	witness_display_list(void(*prnt)(const char *fmt, ...),
				     struct witness_list *list);
static void	witness_display(void(*)(const char *fmt, ...));
#endif

MALLOC_DEFINE(M_WITNESS, "witness", "witness structure");

/*
 * If set to 0, witness is disabled.  If set to 1, witness performs full lock
 * order checking for all locks.  If set to 2 or higher, then witness skips
 * the full lock order check if the lock being acquired is at a higher level
 * (i.e. farther down in the tree) than the current lock.  This last mode is
 * somewhat experimental and not considered fully safe.  At runtime, this
 * value may be set to 0 to turn off witness.  witness is not allowed be
 * turned on once it is turned off, however.
 */
static int witness_watch = 1;
TUNABLE_INT("debug.witness_watch", &witness_watch);
SYSCTL_PROC(_debug, OID_AUTO, witness_watch, CTLFLAG_RW | CTLTYPE_INT, NULL, 0,
    sysctl_debug_witness_watch, "I", "witness is watching lock operations");

#ifdef KDB
/*
 * When KDB is enabled and witness_kdb is set to 1, it will cause the system
 * to drop into kdebug() when:
 *	- a lock heirarchy violation occurs
 *	- locks are held when going to sleep.
 */
#ifdef WITNESS_KDB
int	witness_kdb = 1;
#else
int	witness_kdb = 0;
#endif
TUNABLE_INT("debug.witness_kdb", &witness_kdb);
SYSCTL_INT(_debug, OID_AUTO, witness_kdb, CTLFLAG_RW, &witness_kdb, 0, "");

/*
 * When KDB is enabled and witness_trace is set to 1, it will cause the system
 * to print a stack trace:
 *	- a lock heirarchy violation occurs
 *	- locks are held when going to sleep.
 */
int	witness_trace = 1;
TUNABLE_INT("debug.witness_trace", &witness_trace);
SYSCTL_INT(_debug, OID_AUTO, witness_trace, CTLFLAG_RW, &witness_trace, 0, "");
#endif /* KDB */

#ifdef WITNESS_SKIPSPIN
int	witness_skipspin = 1;
#else
int	witness_skipspin = 0;
#endif
TUNABLE_INT("debug.witness_skipspin", &witness_skipspin);
SYSCTL_INT(_debug, OID_AUTO, witness_skipspin, CTLFLAG_RDTUN, &witness_skipspin, 0,
    "");

static struct mtx w_mtx;
static struct witness_list w_free = STAILQ_HEAD_INITIALIZER(w_free);
static struct witness_list w_all = STAILQ_HEAD_INITIALIZER(w_all);
static struct witness_list w_spin = STAILQ_HEAD_INITIALIZER(w_spin);
static struct witness_list w_sleep = STAILQ_HEAD_INITIALIZER(w_sleep);
static struct witness_child_list_entry *w_child_free = NULL;
static struct lock_list_entry *w_lock_list_free = NULL;

static struct witness w_data[WITNESS_COUNT];
static struct witness_child_list_entry w_childdata[WITNESS_CHILDCOUNT];
static struct lock_list_entry w_locklistdata[LOCK_CHILDCOUNT];

static struct witness_order_list_entry order_lists[] = {
	{ "proctree", &lock_class_sx },
	{ "allproc", &lock_class_sx },
	{ "Giant", &lock_class_mtx_sleep },
	{ "filedesc structure", &lock_class_mtx_sleep },
	{ "pipe mutex", &lock_class_mtx_sleep },
	{ "sigio lock", &lock_class_mtx_sleep },
	{ "process group", &lock_class_mtx_sleep },
	{ "process lock", &lock_class_mtx_sleep },
	{ "session", &lock_class_mtx_sleep },
	{ "uidinfo hash", &lock_class_mtx_sleep },
	{ "uidinfo struct", &lock_class_mtx_sleep },
	{ "allprison", &lock_class_mtx_sleep },
	{ NULL, NULL },
	/*
	 * Sockets
	 */
	{ "filedesc structure", &lock_class_mtx_sleep },
	{ "accept", &lock_class_mtx_sleep },
	{ "so_snd", &lock_class_mtx_sleep },
	{ "so_rcv", &lock_class_mtx_sleep },
	{ "sellck", &lock_class_mtx_sleep },
	{ NULL, NULL },
	/*
	 * Routing
	 */
	{ "so_rcv", &lock_class_mtx_sleep },
	{ "radix node head", &lock_class_mtx_sleep },
	{ "rtentry", &lock_class_mtx_sleep },
	{ "ifaddr", &lock_class_mtx_sleep },
	{ NULL, NULL },
	/*
	 * UNIX Domain Sockets
	 */
	{ "unp", &lock_class_mtx_sleep },
	{ "so_snd", &lock_class_mtx_sleep },
	{ NULL, NULL },
	/*
	 * UDP/IP
	 */
	{ "udp", &lock_class_mtx_sleep },
	{ "udpinp", &lock_class_mtx_sleep },
	{ "so_snd", &lock_class_mtx_sleep },
	{ NULL, NULL },
	/*
	 * TCP/IP
	 */
	{ "tcp", &lock_class_mtx_sleep },
	{ "tcpinp", &lock_class_mtx_sleep },
	{ "so_snd", &lock_class_mtx_sleep },
	{ NULL, NULL },
	/*
	 * SLIP
	 */
	{ "slip_mtx", &lock_class_mtx_sleep },
	{ "slip sc_mtx", &lock_class_mtx_sleep },
	{ NULL, NULL },
	/*
	 * netatalk
	 */
	{ "ddp_list_mtx", &lock_class_mtx_sleep },
	{ "ddp_mtx", &lock_class_mtx_sleep },
	{ NULL, NULL },
	/*
	 * spin locks
	 */
#ifdef SMP
	{ "ap boot", &lock_class_mtx_spin },
#endif
	{ "sio", &lock_class_mtx_spin },
#ifdef __i386__
	{ "cy", &lock_class_mtx_spin },
#endif
	{ "uart_hwmtx", &lock_class_mtx_spin },
	{ "sabtty", &lock_class_mtx_spin },
	{ "zstty", &lock_class_mtx_spin },
	{ "ng_node", &lock_class_mtx_spin },
	{ "ng_worklist", &lock_class_mtx_spin },
	{ "taskqueue_fast", &lock_class_mtx_spin },
	{ "intr table", &lock_class_mtx_spin },
	{ "ithread table lock", &lock_class_mtx_spin },
	{ "sleepq chain", &lock_class_mtx_spin },
	{ "sched lock", &lock_class_mtx_spin },
	{ "turnstile chain", &lock_class_mtx_spin },
	{ "td_contested", &lock_class_mtx_spin },
	{ "callout", &lock_class_mtx_spin },
	{ "entropy harvest", &lock_class_mtx_spin },
	{ "entropy harvest buffers", &lock_class_mtx_spin },
	/*
	 * leaf locks
	 */
	{ "allpmaps", &lock_class_mtx_spin },
	{ "vm page queue free mutex", &lock_class_mtx_spin },
	{ "icu", &lock_class_mtx_spin },
#ifdef SMP
	{ "smp rendezvous", &lock_class_mtx_spin },
#if defined(__i386__) || defined(__amd64__)
	{ "tlb", &lock_class_mtx_spin },
	{ "lazypmap", &lock_class_mtx_spin },
#endif
#ifdef __sparc64__
	{ "ipi", &lock_class_mtx_spin },
#endif
#endif
	{ "clk", &lock_class_mtx_spin },
	{ "mutex profiling lock", &lock_class_mtx_spin },
	{ "kse zombie lock", &lock_class_mtx_spin },
	{ "ALD Queue", &lock_class_mtx_spin },
#ifdef __ia64__
	{ "MCA spin lock", &lock_class_mtx_spin },
#endif
#if defined(__i386__) || defined(__amd64__)
	{ "pcicfg", &lock_class_mtx_spin },
#endif
	{ NULL, NULL },
	{ NULL, NULL }
};

#ifdef BLESSING
/*
 * Pairs of locks which have been blessed
 * Don't complain about order problems with blessed locks
 */
static struct witness_blessed blessed_list[] = {
};
static int blessed_count =
	sizeof(blessed_list) / sizeof(struct witness_blessed);
#endif

/*
 * List of all locks in the system.
 */
TAILQ_HEAD(, lock_object) all_locks = TAILQ_HEAD_INITIALIZER(all_locks);

static struct mtx all_mtx = {
	{ &lock_class_mtx_sleep,	/* mtx_object.lo_class */
	  "All locks list",		/* mtx_object.lo_name */
	  "All locks list",		/* mtx_object.lo_type */
	  LO_INITIALIZED,		/* mtx_object.lo_flags */
	  { NULL, NULL },		/* mtx_object.lo_list */
	  NULL },			/* mtx_object.lo_witness */
	MTX_UNOWNED, 0			/* mtx_lock, mtx_recurse */
};

/*
 * This global is set to 0 once it becomes safe to use the witness code.
 */
static int witness_cold = 1;

/*
 * Global variables for book keeping.
 */
static int lock_cur_cnt;
static int lock_max_cnt;

/*
 * The WITNESS-enabled diagnostic code.
 */
static void
witness_initialize(void *dummy __unused)
{
	struct lock_object *lock;
	struct witness_order_list_entry *order;
	struct witness *w, *w1;
	int i;

	/*
	 * We have to release Giant before initializing its witness
	 * structure so that WITNESS doesn't get confused.
	 */
	mtx_unlock(&Giant);
	mtx_assert(&Giant, MA_NOTOWNED);

	CTR1(KTR_WITNESS, "%s: initializing witness", __func__);
	TAILQ_INSERT_HEAD(&all_locks, &all_mtx.mtx_object, lo_list);
	mtx_init(&w_mtx, "witness lock", NULL, MTX_SPIN | MTX_QUIET |
	    MTX_NOWITNESS);
	for (i = 0; i < WITNESS_COUNT; i++)
		witness_free(&w_data[i]);
	for (i = 0; i < WITNESS_CHILDCOUNT; i++)
		witness_child_free(&w_childdata[i]);
	for (i = 0; i < LOCK_CHILDCOUNT; i++)
		witness_lock_list_free(&w_locklistdata[i]);

	/* First add in all the specified order lists. */
	for (order = order_lists; order->w_name != NULL; order++) {
		w = enroll(order->w_name, order->w_class);
		if (w == NULL)
			continue;
		w->w_file = "order list";
		for (order++; order->w_name != NULL; order++) {
			w1 = enroll(order->w_name, order->w_class);
			if (w1 == NULL)
				continue;
			w1->w_file = "order list";
			if (!itismychild(w, w1))
				panic("Not enough memory for static orders!");
			w = w1;
		}
	}

	/* Iterate through all locks and add them to witness. */
	mtx_lock(&all_mtx);
	TAILQ_FOREACH(lock, &all_locks, lo_list) {
		if (lock->lo_flags & LO_WITNESS)
			lock->lo_witness = enroll(lock->lo_type,
			    lock->lo_class);
		else
			lock->lo_witness = NULL;
	}
	mtx_unlock(&all_mtx);

	/* Mark the witness code as being ready for use. */
	atomic_store_rel_int(&witness_cold, 0);

	mtx_lock(&Giant);
}
SYSINIT(witness_init, SI_SUB_WITNESS, SI_ORDER_FIRST, witness_initialize, NULL)

static int
sysctl_debug_witness_watch(SYSCTL_HANDLER_ARGS)
{
	int error, value;

	value = witness_watch;
	error = sysctl_handle_int(oidp, &value, 0, req);
	if (error != 0 || req->newptr == NULL)
		return (error);
	error = suser(req->td);
	if (error != 0)
		return (error);
	if (value == witness_watch)
		return (0);
	if (value != 0)
		return (EINVAL);
	witness_watch = 0;
	return (0);
}

void
witness_init(struct lock_object *lock)
{
	struct lock_class *class;

	class = lock->lo_class;
	if (lock->lo_flags & LO_INITIALIZED)
		panic("%s: lock (%s) %s is already initialized", __func__,
		    class->lc_name, lock->lo_name);
	if ((lock->lo_flags & LO_RECURSABLE) != 0 &&
	    (class->lc_flags & LC_RECURSABLE) == 0)
		panic("%s: lock (%s) %s can not be recursable", __func__,
		    class->lc_name, lock->lo_name);
	if ((lock->lo_flags & LO_SLEEPABLE) != 0 &&
	    (class->lc_flags & LC_SLEEPABLE) == 0)
		panic("%s: lock (%s) %s can not be sleepable", __func__,
		    class->lc_name, lock->lo_name);
	if ((lock->lo_flags & LO_UPGRADABLE) != 0 &&
	    (class->lc_flags & LC_UPGRADABLE) == 0)
		panic("%s: lock (%s) %s can not be upgradable", __func__,
		    class->lc_name, lock->lo_name);

	mtx_lock(&all_mtx);
	TAILQ_INSERT_TAIL(&all_locks, lock, lo_list);
	lock->lo_flags |= LO_INITIALIZED;
	lock_cur_cnt++;
	if (lock_cur_cnt > lock_max_cnt)
		lock_max_cnt = lock_cur_cnt;
	mtx_unlock(&all_mtx);
	if (!witness_cold && witness_watch != 0 && panicstr == NULL &&
	    (lock->lo_flags & LO_WITNESS) != 0)
		lock->lo_witness = enroll(lock->lo_type, class);
	else
		lock->lo_witness = NULL;
}

void
witness_destroy(struct lock_object *lock)
{
	struct witness *w;

	if (witness_cold)
		panic("lock (%s) %s destroyed while witness_cold",
		    lock->lo_class->lc_name, lock->lo_name);
	if ((lock->lo_flags & LO_INITIALIZED) == 0)
		panic("%s: lock (%s) %s is not initialized", __func__,
		    lock->lo_class->lc_name, lock->lo_name);

	/* XXX: need to verify that no one holds the lock */
	w = lock->lo_witness;
	if (w != NULL) {
		mtx_lock_spin(&w_mtx);
		MPASS(w->w_refcount > 0);
		w->w_refcount--;

		/*
		 * Lock is already released if we have an allocation failure
		 * and depart() fails.
		 */
		if (w->w_refcount != 0 || depart(w))
			mtx_unlock_spin(&w_mtx);
	}

	mtx_lock(&all_mtx);
	lock_cur_cnt--;
	TAILQ_REMOVE(&all_locks, lock, lo_list);
	lock->lo_flags &= ~LO_INITIALIZED;
	mtx_unlock(&all_mtx);
}

#ifdef DDB
static void
witness_display_list(void(*prnt)(const char *fmt, ...),
		     struct witness_list *list)
{
	struct witness *w;

	STAILQ_FOREACH(w, list, w_typelist) {
		if (w->w_file == NULL || w->w_level > 0)
			continue;
		/*
		 * This lock has no anscestors, display its descendants.
		 */
		witness_displaydescendants(prnt, w, 0);
	}
}

static void
witness_display(void(*prnt)(const char *fmt, ...))
{
	struct witness *w;

	KASSERT(!witness_cold, ("%s: witness_cold", __func__));
	witness_levelall();

	/* Clear all the displayed flags. */
	STAILQ_FOREACH(w, &w_all, w_list) {
		w->w_displayed = 0;
	}

	/*
	 * First, handle sleep locks which have been acquired at least
	 * once.
	 */
	prnt("Sleep locks:\n");
	witness_display_list(prnt, &w_sleep);

	/*
	 * Now do spin locks which have been acquired at least once.
	 */
	prnt("\nSpin locks:\n");
	witness_display_list(prnt, &w_spin);

	/*
	 * Finally, any locks which have not been acquired yet.
	 */
	prnt("\nLocks which were never acquired:\n");
	STAILQ_FOREACH(w, &w_all, w_list) {
		if (w->w_file != NULL || w->w_refcount == 0)
			continue;
		prnt("%s\n", w->w_name);
	}
}
#endif /* DDB */

/* Trim useless garbage from filenames. */
static const char *
fixup_filename(const char *file)
{

	if (file == NULL)
		return (NULL);
	while (strncmp(file, "../", 3) == 0)
		file += 3;
	return (file);
}

int
witness_defineorder(struct lock_object *lock1, struct lock_object *lock2)
{

	if (witness_watch == 0 || panicstr != NULL)
		return (0);

	/* Require locks that witness knows about. */
	if (lock1 == NULL || lock1->lo_witness == NULL || lock2 == NULL ||
	    lock2->lo_witness == NULL)
		return (EINVAL);

	MPASS(!mtx_owned(&w_mtx));
	mtx_lock_spin(&w_mtx);

	/*
	 * If we already have either an explicit or implied lock order that
	 * is the other way around, then return an error.
	 */
	if (isitmydescendant(lock2->lo_witness, lock1->lo_witness)) {
		mtx_unlock_spin(&w_mtx);
		return (EDOOFUS);
	}

	/* Try to add the new order. */
	CTR3(KTR_WITNESS, "%s: adding %s as a child of %s", __func__,
	    lock2->lo_type, lock1->lo_type);
	if (!itismychild(lock1->lo_witness, lock2->lo_witness))
		return (ENOMEM);
	mtx_unlock_spin(&w_mtx);
	return (0);
}

void
witness_checkorder(struct lock_object *lock, int flags, const char *file,
    int line)
{
	struct lock_list_entry **lock_list, *lle;
	struct lock_instance *lock1, *lock2;
	struct lock_class *class;
	struct witness *w, *w1;
	struct thread *td;
	int i, j;

	if (witness_cold || witness_watch == 0 || lock->lo_witness == NULL ||
	    panicstr != NULL)
		return;

	/*
	 * Try locks do not block if they fail to acquire the lock, thus
	 * there is no danger of deadlocks or of switching while holding a
	 * spin lock if we acquire a lock via a try operation.  This
	 * function shouldn't even be called for try locks, so panic if
	 * that happens.
	 */
	if (flags & LOP_TRYLOCK)
		panic("%s should not be called for try lock operations",
		    __func__);

	w = lock->lo_witness;
	class = lock->lo_class;
	td = curthread;
	file = fixup_filename(file);

	if (class->lc_flags & LC_SLEEPLOCK) {
		/*
		 * Since spin locks include a critical section, this check
		 * implicitly enforces a lock order of all sleep locks before
		 * all spin locks.
		 */
		if (td->td_critnest != 0)
			panic("blockable sleep lock (%s) %s @ %s:%d",
			    class->lc_name, lock->lo_name, file, line);

		/*
		 * If this is the first lock acquired then just return as
		 * no order checking is needed.
		 */
		if (td->td_sleeplocks == NULL)
			return;
		lock_list = &td->td_sleeplocks;
	} else {
		/*
		 * If this is the first lock, just return as no order
		 * checking is needed.  We check this in both if clauses
		 * here as unifying the check would require us to use a
		 * critical section to ensure we don't migrate while doing
		 * the check.  Note that if this is not the first lock, we
		 * are already in a critical section and are safe for the
		 * rest of the check.
		 */
		if (PCPU_GET(spinlocks) == NULL)
			return;
		lock_list = PCPU_PTR(spinlocks);
	}

	/*
	 * Check to see if we are recursing on a lock we already own.  If
	 * so, make sure that we don't mismatch exclusive and shared lock
	 * acquires.
	 */
	lock1 = find_instance(*lock_list, lock);
	if (lock1 != NULL) {
		if ((lock1->li_flags & LI_EXCLUSIVE) != 0 &&
		    (flags & LOP_EXCLUSIVE) == 0) {
			printf("shared lock of (%s) %s @ %s:%d\n",
			    class->lc_name, lock->lo_name, file, line);
			printf("while exclusively locked from %s:%d\n",
			    lock1->li_file, lock1->li_line);
			panic("share->excl");
		}
		if ((lock1->li_flags & LI_EXCLUSIVE) == 0 &&
		    (flags & LOP_EXCLUSIVE) != 0) {
			printf("exclusive lock of (%s) %s @ %s:%d\n",
			    class->lc_name, lock->lo_name, file, line);
			printf("while share locked from %s:%d\n",
			    lock1->li_file, lock1->li_line);
			panic("excl->share");
		}
		return;
	}

	/*
	 * Try locks do not block if they fail to acquire the lock, thus
	 * there is no danger of deadlocks or of switching while holding a
	 * spin lock if we acquire a lock via a try operation.
	 */
	if (flags & LOP_TRYLOCK)
		return;

	/*
	 * Check for duplicate locks of the same type.  Note that we only
	 * have to check for this on the last lock we just acquired.  Any
	 * other cases will be caught as lock order violations.
	 */
	lock1 = &(*lock_list)->ll_children[(*lock_list)->ll_count - 1];
	w1 = lock1->li_lock->lo_witness;
	if (w1 == w) {
		if (w->w_same_squawked || (lock->lo_flags & LO_DUPOK))
			return;
		w->w_same_squawked = 1;
		printf("acquiring duplicate lock of same type: \"%s\"\n",
			lock->lo_type);
		printf(" 1st %s @ %s:%d\n", lock1->li_lock->lo_name,
		    lock1->li_file, lock1->li_line);
		printf(" 2nd %s @ %s:%d\n", lock->lo_name, file, line);
#ifdef KDB
		goto debugger;
#else
		return;
#endif
	}
	MPASS(!mtx_owned(&w_mtx));
	mtx_lock_spin(&w_mtx);
	/*
	 * If we have a known higher number just say ok
	 */
	if (witness_watch > 1 && w->w_level > w1->w_level) {
		mtx_unlock_spin(&w_mtx);
		return;
	}
	/*
	 * If we know that the the lock we are acquiring comes after
	 * the lock we most recently acquired in the lock order tree,
	 * then there is no need for any further checks.
	 */
	if (isitmydescendant(w1, w)) {
		mtx_unlock_spin(&w_mtx);
		return;
	}
	for (j = 0, lle = *lock_list; lle != NULL; lle = lle->ll_next) {
		for (i = lle->ll_count - 1; i >= 0; i--, j++) {

			MPASS(j < WITNESS_COUNT);
			lock1 = &lle->ll_children[i];
			w1 = lock1->li_lock->lo_witness;

			/*
			 * If this lock doesn't undergo witness checking,
			 * then skip it.
			 */
			if (w1 == NULL) {
				KASSERT((lock1->li_lock->lo_flags & LO_WITNESS) == 0,
				    ("lock missing witness structure"));
				continue;
			}
			/*
			 * If we are locking Giant and this is a sleepable
			 * lock, then skip it.
			 */
			if ((lock1->li_lock->lo_flags & LO_SLEEPABLE) != 0 &&
			    lock == &Giant.mtx_object)
				continue;
			/*
			 * If we are locking a sleepable lock and this lock
			 * is Giant, then skip it.
			 */
			if ((lock->lo_flags & LO_SLEEPABLE) != 0 &&
			    lock1->li_lock == &Giant.mtx_object)
				continue;
			/*
			 * If we are locking a sleepable lock and this lock
			 * isn't sleepable, we want to treat it as a lock
			 * order violation to enfore a general lock order of
			 * sleepable locks before non-sleepable locks.
			 */
			if (!((lock->lo_flags & LO_SLEEPABLE) != 0 &&
			    (lock1->li_lock->lo_flags & LO_SLEEPABLE) == 0))
			    /*
			     * Check the lock order hierarchy for a reveresal.
			     */
			    if (!isitmydescendant(w, w1))
				continue;
			/*
			 * We have a lock order violation, check to see if it
			 * is allowed or has already been yelled about.
			 */
			mtx_unlock_spin(&w_mtx);
#ifdef BLESSING
			/*
			 * If the lock order is blessed, just bail.  We don't
			 * look for other lock order violations though, which
			 * may be a bug.
			 */
			if (blessed(w, w1))
				return;
#endif
			if (lock1->li_lock == &Giant.mtx_object) {
				if (w1->w_Giant_squawked)
					return;
				else
					w1->w_Giant_squawked = 1;
			} else {
				if (w1->w_other_squawked)
					return;
				else
					w1->w_other_squawked = 1;
			}
			/*
			 * Ok, yell about it.
			 */
			printf("lock order reversal\n");
			/*
			 * Try to locate an earlier lock with
			 * witness w in our list.
			 */
			do {
				lock2 = &lle->ll_children[i];
				MPASS(lock2->li_lock != NULL);
				if (lock2->li_lock->lo_witness == w)
					break;
				if (i == 0 && lle->ll_next != NULL) {
					lle = lle->ll_next;
					i = lle->ll_count - 1;
					MPASS(i >= 0 && i < LOCK_NCHILDREN);
				} else
					i--;
			} while (i >= 0);
			if (i < 0) {
				printf(" 1st %p %s (%s) @ %s:%d\n",
				    lock1->li_lock, lock1->li_lock->lo_name,
				    lock1->li_lock->lo_type, lock1->li_file,
				    lock1->li_line);
				printf(" 2nd %p %s (%s) @ %s:%d\n", lock,
				    lock->lo_name, lock->lo_type, file, line);
			} else {
				printf(" 1st %p %s (%s) @ %s:%d\n",
				    lock2->li_lock, lock2->li_lock->lo_name,
				    lock2->li_lock->lo_type, lock2->li_file,
				    lock2->li_line);
				printf(" 2nd %p %s (%s) @ %s:%d\n",
				    lock1->li_lock, lock1->li_lock->lo_name,
				    lock1->li_lock->lo_type, lock1->li_file,
				    lock1->li_line);
				printf(" 3rd %p %s (%s) @ %s:%d\n", lock,
				    lock->lo_name, lock->lo_type, file, line);
			}
#ifdef KDB
			goto debugger;
#else
			return;
#endif
		}
	}
	lock1 = &(*lock_list)->ll_children[(*lock_list)->ll_count - 1];
	/*
	 * If requested, build a new lock order.  However, don't build a new
	 * relationship between a sleepable lock and Giant if it is in the
	 * wrong direction.  The correct lock order is that sleepable locks
	 * always come before Giant.
	 */
	if (flags & LOP_NEWORDER &&
	    !(lock1->li_lock == &Giant.mtx_object &&
	    (lock->lo_flags & LO_SLEEPABLE) != 0)) {
		CTR3(KTR_WITNESS, "%s: adding %s as a child of %s", __func__,
		    lock->lo_type, lock1->li_lock->lo_type);
		if (!itismychild(lock1->li_lock->lo_witness, w))
			/* Witness is dead. */
			return;
	}
	mtx_unlock_spin(&w_mtx);
	return;

#ifdef KDB
debugger:
	if (witness_trace)
		kdb_backtrace();
	if (witness_kdb)
		kdb_enter(__func__);
#endif
}

void
witness_lock(struct lock_object *lock, int flags, const char *file, int line)
{
	struct lock_list_entry **lock_list, *lle;
	struct lock_instance *instance;
	struct witness *w;
	struct thread *td;

	if (witness_cold || witness_watch == 0 || lock->lo_witness == NULL ||
	    panicstr != NULL)
		return;
	w = lock->lo_witness;
	td = curthread;
	file = fixup_filename(file);

	/* Determine lock list for this lock. */
	if (lock->lo_class->lc_flags & LC_SLEEPLOCK)
		lock_list = &td->td_sleeplocks;
	else
		lock_list = PCPU_PTR(spinlocks);

	/* Check to see if we are recursing on a lock we already own. */
	instance = find_instance(*lock_list, lock);
	if (instance != NULL) {
		instance->li_flags++;
		CTR4(KTR_WITNESS, "%s: pid %d recursed on %s r=%d", __func__,
		    td->td_proc->p_pid, lock->lo_name,
		    instance->li_flags & LI_RECURSEMASK);
		instance->li_file = file;
		instance->li_line = line;
		return;
	}

	/* Update per-witness last file and line acquire. */
	w->w_file = file;
	w->w_line = line;

	/* Find the next open lock instance in the list and fill it. */
	lle = *lock_list;
	if (lle == NULL || lle->ll_count == LOCK_NCHILDREN) {
		lle = witness_lock_list_get();
		if (lle == NULL)
			return;
		lle->ll_next = *lock_list;
		CTR3(KTR_WITNESS, "%s: pid %d added lle %p", __func__,
		    td->td_proc->p_pid, lle);
		*lock_list = lle;
	}
	instance = &lle->ll_children[lle->ll_count++];
	instance->li_lock = lock;
	instance->li_line = line;
	instance->li_file = file;
	if ((flags & LOP_EXCLUSIVE) != 0)
		instance->li_flags = LI_EXCLUSIVE;
	else
		instance->li_flags = 0;
	CTR4(KTR_WITNESS, "%s: pid %d added %s as lle[%d]", __func__,
	    td->td_proc->p_pid, lock->lo_name, lle->ll_count - 1);
}

void
witness_upgrade(struct lock_object *lock, int flags, const char *file, int line)
{
	struct lock_instance *instance;
	struct lock_class *class;

	KASSERT(!witness_cold, ("%s: witness_cold", __func__));
	if (lock->lo_witness == NULL || witness_watch == 0 || panicstr != NULL)
		return;
	class = lock->lo_class;
	file = fixup_filename(file);
	if ((lock->lo_flags & LO_UPGRADABLE) == 0)
		panic("upgrade of non-upgradable lock (%s) %s @ %s:%d",
		    class->lc_name, lock->lo_name, file, line);
	if ((flags & LOP_TRYLOCK) == 0)
		panic("non-try upgrade of lock (%s) %s @ %s:%d", class->lc_name,
		    lock->lo_name, file, line);
	if ((lock->lo_class->lc_flags & LC_SLEEPLOCK) == 0)
		panic("upgrade of non-sleep lock (%s) %s @ %s:%d",
		    class->lc_name, lock->lo_name, file, line);
	instance = find_instance(curthread->td_sleeplocks, lock);
	if (instance == NULL)
		panic("upgrade of unlocked lock (%s) %s @ %s:%d",
		    class->lc_name, lock->lo_name, file, line);
	if ((instance->li_flags & LI_EXCLUSIVE) != 0)
		panic("upgrade of exclusive lock (%s) %s @ %s:%d",
		    class->lc_name, lock->lo_name, file, line);
	if ((instance->li_flags & LI_RECURSEMASK) != 0)
		panic("upgrade of recursed lock (%s) %s r=%d @ %s:%d",
		    class->lc_name, lock->lo_name,
		    instance->li_flags & LI_RECURSEMASK, file, line);
	instance->li_flags |= LI_EXCLUSIVE;
}

void
witness_downgrade(struct lock_object *lock, int flags, const char *file,
    int line)
{
	struct lock_instance *instance;
	struct lock_class *class;

	KASSERT(!witness_cold, ("%s: witness_cold", __func__));
	if (lock->lo_witness == NULL || witness_watch == 0 || panicstr != NULL)
		return;
	class = lock->lo_class;
	file = fixup_filename(file);
	if ((lock->lo_flags & LO_UPGRADABLE) == 0)
		panic("downgrade of non-upgradable lock (%s) %s @ %s:%d",
		    class->lc_name, lock->lo_name, file, line);
	if ((lock->lo_class->lc_flags & LC_SLEEPLOCK) == 0)
		panic("downgrade of non-sleep lock (%s) %s @ %s:%d",
		    class->lc_name, lock->lo_name, file, line);
	instance = find_instance(curthread->td_sleeplocks, lock);
	if (instance == NULL)
		panic("downgrade of unlocked lock (%s) %s @ %s:%d",
		    class->lc_name, lock->lo_name, file, line);
	if ((instance->li_flags & LI_EXCLUSIVE) == 0)
		panic("downgrade of shared lock (%s) %s @ %s:%d",
		    class->lc_name, lock->lo_name, file, line);
	if ((instance->li_flags & LI_RECURSEMASK) != 0)
		panic("downgrade of recursed lock (%s) %s r=%d @ %s:%d",
		    class->lc_name, lock->lo_name,
		    instance->li_flags & LI_RECURSEMASK, file, line);
	instance->li_flags &= ~LI_EXCLUSIVE;
}

void
witness_unlock(struct lock_object *lock, int flags, const char *file, int line)
{
	struct lock_list_entry **lock_list, *lle;
	struct lock_instance *instance;
	struct lock_class *class;
	struct thread *td;
	register_t s;
	int i, j;

	if (witness_cold || witness_watch == 0 || lock->lo_witness == NULL ||
	    panicstr != NULL)
		return;
	td = curthread;
	class = lock->lo_class;
	file = fixup_filename(file);

	/* Find lock instance associated with this lock. */
	if (class->lc_flags & LC_SLEEPLOCK)
		lock_list = &td->td_sleeplocks;
	else
		lock_list = PCPU_PTR(spinlocks);
	for (; *lock_list != NULL; lock_list = &(*lock_list)->ll_next)
		for (i = 0; i < (*lock_list)->ll_count; i++) {
			instance = &(*lock_list)->ll_children[i];
			if (instance->li_lock == lock)
				goto found;
		}
	panic("lock (%s) %s not locked @ %s:%d", class->lc_name, lock->lo_name,
	    file, line);
found:

	/* First, check for shared/exclusive mismatches. */
	if ((instance->li_flags & LI_EXCLUSIVE) != 0 &&
	    (flags & LOP_EXCLUSIVE) == 0) {
		printf("shared unlock of (%s) %s @ %s:%d\n", class->lc_name,
		    lock->lo_name, file, line);
		printf("while exclusively locked from %s:%d\n",
		    instance->li_file, instance->li_line);
		panic("excl->ushare");
	}
	if ((instance->li_flags & LI_EXCLUSIVE) == 0 &&
	    (flags & LOP_EXCLUSIVE) != 0) {
		printf("exclusive unlock of (%s) %s @ %s:%d\n", class->lc_name,
		    lock->lo_name, file, line);
		printf("while share locked from %s:%d\n", instance->li_file,
		    instance->li_line);
		panic("share->uexcl");
	}

	/* If we are recursed, unrecurse. */
	if ((instance->li_flags & LI_RECURSEMASK) > 0) {
		CTR4(KTR_WITNESS, "%s: pid %d unrecursed on %s r=%d", __func__,
		    td->td_proc->p_pid, instance->li_lock->lo_name,
		    instance->li_flags);
		instance->li_flags--;
		return;
	}

	/* Otherwise, remove this item from the list. */
	s = intr_disable();
	CTR4(KTR_WITNESS, "%s: pid %d removed %s from lle[%d]", __func__,
	    td->td_proc->p_pid, instance->li_lock->lo_name,
	    (*lock_list)->ll_count - 1);
	for (j = i; j < (*lock_list)->ll_count - 1; j++)
		(*lock_list)->ll_children[j] =
		    (*lock_list)->ll_children[j + 1];
	(*lock_list)->ll_count--;
	intr_restore(s);

	/* If this lock list entry is now empty, free it. */
	if ((*lock_list)->ll_count == 0) {
		lle = *lock_list;
		*lock_list = lle->ll_next;
		CTR3(KTR_WITNESS, "%s: pid %d removed lle %p", __func__,
		    td->td_proc->p_pid, lle);
		witness_lock_list_free(lle);
	}
}

/*
 * Warn if any locks other than 'lock' are held.  Flags can be passed in to
 * exempt Giant and sleepable locks from the checks as well.  If any
 * non-exempt locks are held, then a supplied message is printed to the
 * console along with a list of the offending locks.  If indicated in the
 * flags then a failure results in a panic as well.
 */
int
witness_warn(int flags, struct lock_object *lock, const char *fmt, ...)
{
	struct lock_list_entry *lle;
	struct lock_instance *lock1;
	struct thread *td;
	va_list ap;
	int i, n;

	if (witness_cold || witness_watch == 0 || panicstr != NULL)
		return (0);
	n = 0;
	td = curthread;
	for (lle = td->td_sleeplocks; lle != NULL; lle = lle->ll_next)
		for (i = lle->ll_count - 1; i >= 0; i--) {
			lock1 = &lle->ll_children[i];
			if (lock1->li_lock == lock)
				continue;
			if (flags & WARN_GIANTOK &&
			    lock1->li_lock == &Giant.mtx_object)
				continue;
			if (flags & WARN_SLEEPOK &&
			    (lock1->li_lock->lo_flags & LO_SLEEPABLE) != 0)
				continue;
			if (n == 0) {
				va_start(ap, fmt);
				vprintf(fmt, ap);
				va_end(ap);
				printf(" with the following");
				if (flags & WARN_SLEEPOK)
					printf(" non-sleepable");
				printf(" locks held:\n");
			}
			n++;
			witness_list_lock(lock1);
		}
	if (PCPU_GET(spinlocks) != NULL) {
		/*
		 * Since we already hold a spinlock preemption is
		 * already blocked.
		 */
		if (n == 0) {
			va_start(ap, fmt);
			vprintf(fmt, ap);
			va_end(ap);
			printf(" with the following");
			if (flags & WARN_SLEEPOK)
				printf(" non-sleepable");
			printf(" locks held:\n");
		}
		n += witness_list_locks(PCPU_PTR(spinlocks));
	}
	if (flags & WARN_PANIC && n)
		panic("witness_warn");
#ifdef KDB
	else if (witness_kdb && n)
		kdb_enter(__func__);
	else if (witness_trace && n)
		kdb_backtrace();
#endif
	return (n);
}

const char *
witness_file(struct lock_object *lock)
{
	struct witness *w;

	if (witness_cold || witness_watch == 0 || lock->lo_witness == NULL)
		return ("?");
	w = lock->lo_witness;
	return (w->w_file);
}

int
witness_line(struct lock_object *lock)
{
	struct witness *w;

	if (witness_cold || witness_watch == 0 || lock->lo_witness == NULL)
		return (0);
	w = lock->lo_witness;
	return (w->w_line);
}

static struct witness *
enroll(const char *description, struct lock_class *lock_class)
{
	struct witness *w;

	if (witness_watch == 0 || panicstr != NULL)
		return (NULL);
	if ((lock_class->lc_flags & LC_SPINLOCK) && witness_skipspin)
		return (NULL);
	mtx_lock_spin(&w_mtx);
	STAILQ_FOREACH(w, &w_all, w_list) {
		if (w->w_name == description || (w->w_refcount > 0 &&
		    strcmp(description, w->w_name) == 0)) {
			w->w_refcount++;
			mtx_unlock_spin(&w_mtx);
			if (lock_class != w->w_class)
				panic(
				"lock (%s) %s does not match earlier (%s) lock",
				    description, lock_class->lc_name,
				    w->w_class->lc_name);
			return (w);
		}
	}
	/*
	 * This isn't quite right, as witness_cold is still 0 while we
	 * enroll all the locks initialized before witness_initialize().
	 */
	if ((lock_class->lc_flags & LC_SPINLOCK) && !witness_cold) {
		mtx_unlock_spin(&w_mtx);
		panic("spin lock %s not in order list", description);
	}
	if ((w = witness_get()) == NULL)
		return (NULL);
	w->w_name = description;
	w->w_class = lock_class;
	w->w_refcount = 1;
	STAILQ_INSERT_HEAD(&w_all, w, w_list);
	if (lock_class->lc_flags & LC_SPINLOCK)
		STAILQ_INSERT_HEAD(&w_spin, w, w_typelist);
	else if (lock_class->lc_flags & LC_SLEEPLOCK)
		STAILQ_INSERT_HEAD(&w_sleep, w, w_typelist);
	else {
		mtx_unlock_spin(&w_mtx);
		panic("lock class %s is not sleep or spin",
		    lock_class->lc_name);
	}
	mtx_unlock_spin(&w_mtx);
	return (w);
}

/* Don't let the door bang you on the way out... */
static int
depart(struct witness *w)
{
	struct witness_child_list_entry *wcl, *nwcl;
	struct witness_list *list;
	struct witness *parent;

	MPASS(w->w_refcount == 0);
	if (w->w_class->lc_flags & LC_SLEEPLOCK)
		list = &w_sleep;
	else
		list = &w_spin;
	/*
	 * First, we run through the entire tree looking for any
	 * witnesses that the outgoing witness is a child of.  For
	 * each parent that we find, we reparent all the direct
	 * children of the outgoing witness to its parent.
	 */
	STAILQ_FOREACH(parent, list, w_typelist) {
		if (!isitmychild(parent, w))
			continue;
		removechild(parent, w);
		if (!reparentchildren(parent, w))
			return (0);
	}

	/*
	 * Now we go through and free up the child list of the
	 * outgoing witness.
	 */
	for (wcl = w->w_children; wcl != NULL; wcl = nwcl) {
		nwcl = wcl->wcl_next;
		witness_child_free(wcl);
	}

	/*
	 * Detach from various lists and free.
	 */
	STAILQ_REMOVE(list, w, witness, w_typelist);
	STAILQ_REMOVE(&w_all, w, witness, w_list);
	witness_free(w);

	/* Finally, fixup the tree. */
	return (rebalancetree(list));
}

/*
 * Prune an entire lock order tree.  We look for cases where a lock
 * is now both a descendant and a direct child of a given lock.  In
 * that case, we want to remove the direct child link from the tree.
 *
 * Returns false if insertchild() fails.
 */
static int
rebalancetree(struct witness_list *list)
{
	struct witness *child, *parent;

	STAILQ_FOREACH(child, list, w_typelist) {
		STAILQ_FOREACH(parent, list, w_typelist) {
			if (!isitmychild(parent, child))
				continue;
			removechild(parent, child);
			if (isitmydescendant(parent, child))
				continue;
			if (!insertchild(parent, child))
				return (0);
		}
	}
	witness_levelall();
	return (1);
}

/*
 * Add "child" as a direct child of "parent".  Returns false if
 * we fail due to out of memory.
 */
static int
insertchild(struct witness *parent, struct witness *child)
{
	struct witness_child_list_entry **wcl;

	MPASS(child != NULL && parent != NULL);

	/*
	 * Insert "child" after "parent"
	 */
	wcl = &parent->w_children;
	while (*wcl != NULL && (*wcl)->wcl_count == WITNESS_NCHILDREN)
		wcl = &(*wcl)->wcl_next;
	if (*wcl == NULL) {
		*wcl = witness_child_get();
		if (*wcl == NULL)
			return (0);
	}
	(*wcl)->wcl_children[(*wcl)->wcl_count++] = child;

	return (1);
}

/*
 * Make all the direct descendants of oldparent be direct descendants
 * of newparent.
 */
static int
reparentchildren(struct witness *newparent, struct witness *oldparent)
{
	struct witness_child_list_entry *wcl;
	int i;

	/* Avoid making a witness a child of itself. */
	MPASS(!isitmychild(oldparent, newparent));

	for (wcl = oldparent->w_children; wcl != NULL; wcl = wcl->wcl_next)
		for (i = 0; i < wcl->wcl_count; i++)
			if (!insertchild(newparent, wcl->wcl_children[i]))
				return (0);
	return (1);
}

static int
itismychild(struct witness *parent, struct witness *child)
{
	struct witness_list *list;

	MPASS(child != NULL && parent != NULL);
	if ((parent->w_class->lc_flags & (LC_SLEEPLOCK | LC_SPINLOCK)) !=
	    (child->w_class->lc_flags & (LC_SLEEPLOCK | LC_SPINLOCK)))
		panic(
		"%s: parent (%s) and child (%s) are not the same lock type",
		    __func__, parent->w_class->lc_name,
		    child->w_class->lc_name);

	if (!insertchild(parent, child))
		return (0);

	if (parent->w_class->lc_flags & LC_SLEEPLOCK)
		list = &w_sleep;
	else
		list = &w_spin;
	return (rebalancetree(list));
}

static void
removechild(struct witness *parent, struct witness *child)
{
	struct witness_child_list_entry **wcl, *wcl1;
	int i;

	for (wcl = &parent->w_children; *wcl != NULL; wcl = &(*wcl)->wcl_next)
		for (i = 0; i < (*wcl)->wcl_count; i++)
			if ((*wcl)->wcl_children[i] == child)
				goto found;
	return;
found:
	(*wcl)->wcl_count--;
	if ((*wcl)->wcl_count > i)
		(*wcl)->wcl_children[i] =
		    (*wcl)->wcl_children[(*wcl)->wcl_count];
	MPASS((*wcl)->wcl_children[i] != NULL);
	if ((*wcl)->wcl_count != 0)
		return;
	wcl1 = *wcl;
	*wcl = wcl1->wcl_next;
	witness_child_free(wcl1);
}

static int
isitmychild(struct witness *parent, struct witness *child)
{
	struct witness_child_list_entry *wcl;
	int i;

	for (wcl = parent->w_children; wcl != NULL; wcl = wcl->wcl_next) {
		for (i = 0; i < wcl->wcl_count; i++) {
			if (wcl->wcl_children[i] == child)
				return (1);
		}
	}
	return (0);
}

static int
isitmydescendant(struct witness *parent, struct witness *child)
{
	struct witness_child_list_entry *wcl;
	int i, j;

	if (isitmychild(parent, child))
		return (1);
	j = 0;
	for (wcl = parent->w_children; wcl != NULL; wcl = wcl->wcl_next) {
		MPASS(j < 1000);
		for (i = 0; i < wcl->wcl_count; i++) {
			if (isitmydescendant(wcl->wcl_children[i], child))
				return (1);
		}
		j++;
	}
	return (0);
}

static void
witness_levelall (void)
{
	struct witness_list *list;
	struct witness *w, *w1;

	/*
	 * First clear all levels.
	 */
	STAILQ_FOREACH(w, &w_all, w_list) {
		w->w_level = 0;
	}

	/*
	 * Look for locks with no parent and level all their descendants.
	 */
	STAILQ_FOREACH(w, &w_all, w_list) {
		/*
		 * This is just an optimization, technically we could get
		 * away just walking the all list each time.
		 */
		if (w->w_class->lc_flags & LC_SLEEPLOCK)
			list = &w_sleep;
		else
			list = &w_spin;
		STAILQ_FOREACH(w1, list, w_typelist) {
			if (isitmychild(w1, w))
				goto skip;
		}
		witness_leveldescendents(w, 0);
	skip:
		;	/* silence GCC 3.x */
	}
}

static void
witness_leveldescendents(struct witness *parent, int level)
{
	struct witness_child_list_entry *wcl;
	int i;

	if (parent->w_level < level)
		parent->w_level = level;
	level++;
	for (wcl = parent->w_children; wcl != NULL; wcl = wcl->wcl_next)
		for (i = 0; i < wcl->wcl_count; i++)
			witness_leveldescendents(wcl->wcl_children[i], level);
}

static void
witness_displaydescendants(void(*prnt)(const char *fmt, ...),
			   struct witness *parent, int indent)
{
	struct witness_child_list_entry *wcl;
	int i, level;

	level = parent->w_level;
	prnt("%-2d", level);
	for (i = 0; i < indent; i++)
		prnt(" ");
	if (parent->w_refcount > 0)
		prnt("%s", parent->w_name);
	else
		prnt("(dead)");
	if (parent->w_displayed) {
		prnt(" -- (already displayed)\n");
		return;
	}
	parent->w_displayed = 1;
	if (parent->w_refcount > 0) {
		if (parent->w_file != NULL)
			prnt(" -- last acquired @ %s:%d", parent->w_file,
			    parent->w_line);
	}
	prnt("\n");
	for (wcl = parent->w_children; wcl != NULL; wcl = wcl->wcl_next)
		for (i = 0; i < wcl->wcl_count; i++)
			    witness_displaydescendants(prnt,
				wcl->wcl_children[i], indent + 1);
}

#ifdef BLESSING
static int
blessed(struct witness *w1, struct witness *w2)
{
	int i;
	struct witness_blessed *b;

	for (i = 0; i < blessed_count; i++) {
		b = &blessed_list[i];
		if (strcmp(w1->w_name, b->b_lock1) == 0) {
			if (strcmp(w2->w_name, b->b_lock2) == 0)
				return (1);
			continue;
		}
		if (strcmp(w1->w_name, b->b_lock2) == 0)
			if (strcmp(w2->w_name, b->b_lock1) == 0)
				return (1);
	}
	return (0);
}
#endif

static struct witness *
witness_get(void)
{
	struct witness *w;

	if (witness_watch == 0) {
		mtx_unlock_spin(&w_mtx);
		return (NULL);
	}
	if (STAILQ_EMPTY(&w_free)) {
		witness_watch = 0;
		mtx_unlock_spin(&w_mtx);
		printf("%s: witness exhausted\n", __func__);
		return (NULL);
	}
	w = STAILQ_FIRST(&w_free);
	STAILQ_REMOVE_HEAD(&w_free, w_list);
	bzero(w, sizeof(*w));
	return (w);
}

static void
witness_free(struct witness *w)
{

	STAILQ_INSERT_HEAD(&w_free, w, w_list);
}

static struct witness_child_list_entry *
witness_child_get(void)
{
	struct witness_child_list_entry *wcl;

	if (witness_watch == 0) {
		mtx_unlock_spin(&w_mtx);
		return (NULL);
	}
	wcl = w_child_free;
	if (wcl == NULL) {
		witness_watch = 0;
		mtx_unlock_spin(&w_mtx);
		printf("%s: witness exhausted\n", __func__);
		return (NULL);
	}
	w_child_free = wcl->wcl_next;
	bzero(wcl, sizeof(*wcl));
	return (wcl);
}

static void
witness_child_free(struct witness_child_list_entry *wcl)
{

	wcl->wcl_next = w_child_free;
	w_child_free = wcl;
}

static struct lock_list_entry *
witness_lock_list_get(void)
{
	struct lock_list_entry *lle;

	if (witness_watch == 0)
		return (NULL);
	mtx_lock_spin(&w_mtx);
	lle = w_lock_list_free;
	if (lle == NULL) {
		witness_watch = 0;
		mtx_unlock_spin(&w_mtx);
		printf("%s: witness exhausted\n", __func__);
		return (NULL);
	}
	w_lock_list_free = lle->ll_next;
	mtx_unlock_spin(&w_mtx);
	bzero(lle, sizeof(*lle));
	return (lle);
}

static void
witness_lock_list_free(struct lock_list_entry *lle)
{

	mtx_lock_spin(&w_mtx);
	lle->ll_next = w_lock_list_free;
	w_lock_list_free = lle;
	mtx_unlock_spin(&w_mtx);
}

static struct lock_instance *
find_instance(struct lock_list_entry *lock_list, struct lock_object *lock)
{
	struct lock_list_entry *lle;
	struct lock_instance *instance;
	int i;

	for (lle = lock_list; lle != NULL; lle = lle->ll_next)
		for (i = lle->ll_count - 1; i >= 0; i--) {
			instance = &lle->ll_children[i];
			if (instance->li_lock == lock)
				return (instance);
		}
	return (NULL);
}

static void
witness_list_lock(struct lock_instance *instance)
{
	struct lock_object *lock;

	lock = instance->li_lock;
	printf("%s %s %s", (instance->li_flags & LI_EXCLUSIVE) != 0 ?
	    "exclusive" : "shared", lock->lo_class->lc_name, lock->lo_name);
	if (lock->lo_type != lock->lo_name)
		printf(" (%s)", lock->lo_type);
	printf(" r = %d (%p) locked @ %s:%d\n",
	    instance->li_flags & LI_RECURSEMASK, lock, instance->li_file,
	    instance->li_line);
}

int
witness_list_locks(struct lock_list_entry **lock_list)
{
	struct lock_list_entry *lle;
	int i, nheld;

	nheld = 0;
	for (lle = *lock_list; lle != NULL; lle = lle->ll_next)
		for (i = lle->ll_count - 1; i >= 0; i--) {
			witness_list_lock(&lle->ll_children[i]);
			nheld++;
		}
	return (nheld);
}

/*
 * This is a bit risky at best.  We call this function when we have timed
 * out acquiring a spin lock, and we assume that the other CPU is stuck
 * with this lock held.  So, we go groveling around in the other CPU's
 * per-cpu data to try to find the lock instance for this spin lock to
 * see when it was last acquired.
 */
void
witness_display_spinlock(struct lock_object *lock, struct thread *owner)
{
	struct lock_instance *instance;
	struct pcpu *pc;

	if (owner->td_critnest == 0 || owner->td_oncpu == NOCPU)
		return;
	pc = pcpu_find(owner->td_oncpu);
	instance = find_instance(pc->pc_spinlocks, lock);
	if (instance != NULL)
		witness_list_lock(instance);
}

void
witness_save(struct lock_object *lock, const char **filep, int *linep)
{
	struct lock_instance *instance;

	KASSERT(!witness_cold, ("%s: witness_cold", __func__));
	if (lock->lo_witness == NULL || witness_watch == 0 || panicstr != NULL)
		return;
	if ((lock->lo_class->lc_flags & LC_SLEEPLOCK) == 0)
		panic("%s: lock (%s) %s is not a sleep lock", __func__,
		    lock->lo_class->lc_name, lock->lo_name);
	instance = find_instance(curthread->td_sleeplocks, lock);
	if (instance == NULL)
		panic("%s: lock (%s) %s not locked", __func__,
		    lock->lo_class->lc_name, lock->lo_name);
	*filep = instance->li_file;
	*linep = instance->li_line;
}

void
witness_restore(struct lock_object *lock, const char *file, int line)
{
	struct lock_instance *instance;

	KASSERT(!witness_cold, ("%s: witness_cold", __func__));
	if (lock->lo_witness == NULL || witness_watch == 0 || panicstr != NULL)
		return;
	if ((lock->lo_class->lc_flags & LC_SLEEPLOCK) == 0)
		panic("%s: lock (%s) %s is not a sleep lock", __func__,
		    lock->lo_class->lc_name, lock->lo_name);
	instance = find_instance(curthread->td_sleeplocks, lock);
	if (instance == NULL)
		panic("%s: lock (%s) %s not locked", __func__,
		    lock->lo_class->lc_name, lock->lo_name);
	lock->lo_witness->w_file = file;
	lock->lo_witness->w_line = line;
	instance->li_file = file;
	instance->li_line = line;
}

void
witness_assert(struct lock_object *lock, int flags, const char *file, int line)
{
#ifdef INVARIANT_SUPPORT
	struct lock_instance *instance;

	if (lock->lo_witness == NULL || witness_watch == 0 || panicstr != NULL)
		return;
	if ((lock->lo_class->lc_flags & LC_SLEEPLOCK) != 0)
		instance = find_instance(curthread->td_sleeplocks, lock);
	else if ((lock->lo_class->lc_flags & LC_SPINLOCK) != 0)
		instance = find_instance(PCPU_GET(spinlocks), lock);
	else {
		panic("Lock (%s) %s is not sleep or spin!",
		    lock->lo_class->lc_name, lock->lo_name);
	}
	file = fixup_filename(file);
	switch (flags) {
	case LA_UNLOCKED:
		if (instance != NULL)
			panic("Lock (%s) %s locked @ %s:%d.",
			    lock->lo_class->lc_name, lock->lo_name, file, line);
		break;
	case LA_LOCKED:
	case LA_LOCKED | LA_RECURSED:
	case LA_LOCKED | LA_NOTRECURSED:
	case LA_SLOCKED:
	case LA_SLOCKED | LA_RECURSED:
	case LA_SLOCKED | LA_NOTRECURSED:
	case LA_XLOCKED:
	case LA_XLOCKED | LA_RECURSED:
	case LA_XLOCKED | LA_NOTRECURSED:
		if (instance == NULL) {
			panic("Lock (%s) %s not locked @ %s:%d.",
			    lock->lo_class->lc_name, lock->lo_name, file, line);
			break;
		}
		if ((flags & LA_XLOCKED) != 0 &&
		    (instance->li_flags & LI_EXCLUSIVE) == 0)
			panic("Lock (%s) %s not exclusively locked @ %s:%d.",
			    lock->lo_class->lc_name, lock->lo_name, file, line);
		if ((flags & LA_SLOCKED) != 0 &&
		    (instance->li_flags & LI_EXCLUSIVE) != 0)
			panic("Lock (%s) %s exclusively locked @ %s:%d.",
			    lock->lo_class->lc_name, lock->lo_name, file, line);
		if ((flags & LA_RECURSED) != 0 &&
		    (instance->li_flags & LI_RECURSEMASK) == 0)
			panic("Lock (%s) %s not recursed @ %s:%d.",
			    lock->lo_class->lc_name, lock->lo_name, file, line);
		if ((flags & LA_NOTRECURSED) != 0 &&
		    (instance->li_flags & LI_RECURSEMASK) != 0)
			panic("Lock (%s) %s recursed @ %s:%d.",
			    lock->lo_class->lc_name, lock->lo_name, file, line);
		break;
	default:
		panic("Invalid lock assertion at %s:%d.", file, line);

	}
#endif	/* INVARIANT_SUPPORT */
}

#ifdef DDB
static void
witness_list(struct thread *td)
{

	KASSERT(!witness_cold, ("%s: witness_cold", __func__));
	KASSERT(kdb_active, ("%s: not in the debugger", __func__));

	if (witness_watch == 0)
		return;

	witness_list_locks(&td->td_sleeplocks);

	/*
	 * We only handle spinlocks if td == curthread.  This is somewhat broken
	 * if td is currently executing on some other CPU and holds spin locks
	 * as we won't display those locks.  If we had a MI way of getting
	 * the per-cpu data for a given cpu then we could use
	 * td->td_oncpu to get the list of spinlocks for this thread
	 * and "fix" this.
	 *
	 * That still wouldn't really fix this unless we locked sched_lock
	 * or stopped the other CPU to make sure it wasn't changing the list
	 * out from under us.  It is probably best to just not try to handle
	 * threads on other CPU's for now.
	 */
	if (td == curthread && PCPU_GET(spinlocks) != NULL)
		witness_list_locks(PCPU_PTR(spinlocks));
}

DB_SHOW_COMMAND(locks, db_witness_list)
{
	struct thread *td;
	pid_t pid;
	struct proc *p;

	if (have_addr) {
		pid = (addr % 16) + ((addr >> 4) % 16) * 10 +
		    ((addr >> 8) % 16) * 100 + ((addr >> 12) % 16) * 1000 +
		    ((addr >> 16) % 16) * 10000;
		/* sx_slock(&allproc_lock); */
		FOREACH_PROC_IN_SYSTEM(p) {
			if (p->p_pid == pid)
				break;
		}
		/* sx_sunlock(&allproc_lock); */
		if (p == NULL) {
			db_printf("pid %d not found\n", pid);
			return;
		}
		FOREACH_THREAD_IN_PROC(p, td) {
			witness_list(td);
		}
	} else {
		td = curthread;
		witness_list(td);
	}
}

DB_SHOW_COMMAND(witness, db_witness_display)
{

	witness_display(db_printf);
}
#endif