Rethink the MD interfaces for libpthread to account for

archs that can (or are required to) have per-thread registers.

Tested on i386, amd64; marcel is testing on ia64 and will
have some follow-up commits.

Reviewed by:	davidxu

20 files changed, 983 insertions, 676 deletions

diff --git a/lib/libpthread/arch/amd64/Makefile.inc b/lib/libpthread/arch/amd64/Makefile.inc
index c0e4c47..c8b0362 100644
--- a/lib/libpthread/arch/amd64/Makefile.inc
+++ b/lib/libpthread/arch/amd64/Makefile.inc
@@ -2,4 +2,4 @@
 
 .PATH:	${.CURDIR}/arch/${MACHINE_ARCH}/${MACHINE_ARCH}
 
-SRCS+=	context.S enter_uts.S
+SRCS+=	context.S enter_uts.S pthread_md.c
diff --git a/lib/libpthread/arch/amd64/amd64/pthread_md.c b/lib/libpthread/arch/amd64/amd64/pthread_md.c
new file mode 100644
index 0000000..374b6d8
--- /dev/null
+++ b/lib/libpthread/arch/amd64/amd64/pthread_md.c
@@ -0,0 +1,72 @@
+/*
+ * Copyright (c) 2003 Daniel Eischen <deischen@freebsd.org>
+ * 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. Neither the name of the author nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
+ * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
+ * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
+ * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
+ * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ * $FreeBSD$
+ */
+
+#include <stdlib.h>
+#include <strings.h>
+#include "pthread_md.h"
+
+/*
+ * The constructors.
+ */
+struct tcb *
+_tcb_ctor(struct pthread *thread)
+{
+	struct tcb *tcb;
+
+	if ((tcb = malloc(sizeof(struct tcb))) != NULL) {
+		bzero(tcb, sizeof(struct tcb));
+		tcb->tcb_thread = thread;
+		/* Allocate TDV */
+	}
+	return (tcb);
+}
+
+void
+_tcb_dtor(struct tcb *tcb)
+{
+	/* Free TDV */
+	free(tcb);
+}
+
+struct kcb *
+_kcb_ctor(struct kse *kse)
+{
+	struct kcb *kcb;
+
+	if ((kcb = malloc(sizeof(struct kcb))) != NULL) {
+		bzero(kcb, sizeof(struct kcb));
+		kcb->kcb_kse = kse;
+	}
+	return (kcb);
+}
+
+void
+_kcb_dtor(struct kcb *kcb)
+{
+	free(kcb);
+}
diff --git a/lib/libpthread/arch/amd64/include/ksd.h b/lib/libpthread/arch/amd64/include/ksd.h
deleted file mode 100644
index 843d184..0000000
--- a/lib/libpthread/arch/amd64/include/ksd.h
+++ /dev/null
@@ -1,132 +0,0 @@
-/*-
- * Copyright (C) 2003 David Xu <davidxu@freebsd.org>
- * Copyright (c) 2001 Daniel Eischen <deischen@freebsd.org>
- * 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. Neither the name of the author nor the names of its contributors
- *    may be used to endorse or promote products derived from this software
- *    without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
- * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
- * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
- * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
- * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
- * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
- * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
- * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
- * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- */
-
-/*
- * $FreeBSD$
- */
-
-#ifndef	_KSD_H_
-#define	_KSD_H_
-
-#include <sys/types.h>
-#include <machine/sysarch.h>
-extern int sysarch(int, void *);
-
-struct kse;
-struct pthread;
-
-/*
- * Evaluates to the byte offset of the per-kse variable name.
- */
-#define	__ksd_offset(name)	__offsetof(struct kse, name)
-
-/*
- * Evaluates to the type of the per-kse variable name.
- */
-#define	__ksd_type(name)	__typeof(((struct kse *)0)->name)
-
-
-/*
- * Evaluates to the value of the per-kse variable name.
- */
-#define	KSD_GET64(name) ({					\
-	__ksd_type(name) __result;				\
-								\
-	u_long __i;						\
-	__asm __volatile("movq %%fs:%1, %0"			\
-	    : "=r" (__i)					\
-	    : "m" (*(u_long *)(__ksd_offset(name))));		\
-	__result = *(__ksd_type(name) *)&__i;			\
-								\
-	__result;						\
-})
-
-/*
- * Sets the value of the per-cpu variable name to value val.
- */
-#define	KSD_SET64(name, val) ({					\
-	__ksd_type(name) __val = (val);				\
-								\
-	u_long __i;						\
-	__i = *(u_long *)&__val;					\
-	__asm __volatile("movq %1,%%fs:%0"			\
-	    : "=m" (*(u_long *)(__ksd_offset(name)))		\
-	    : "r" (__i));					\
-})
-
-static __inline u_long
-__ksd_readandclear64(volatile u_long *addr)
-{
-	u_long result;
-
-	__asm __volatile (
-	    "	xorq	%0, %0;"
-	    "	xchgq	%%fs:%1, %0;"
-	    "# __ksd_readandclear64"
-	    : "=&r" (result)
-	    : "m" (*addr));
-	return (result);
-}
-
-#define	KSD_READANDCLEAR64(name) ({				\
-	__ksd_type(name) __result;				\
-								\
-	__result = (__ksd_type(name))				\
-	    __ksd_readandclear64((u_long *)__ksd_offset(name)); \
-	__result;						\
-})
-
-
-#define	_ksd_curkse()		((struct kse *)KSD_GET64(k_mbx.km_udata))
-#define	_ksd_curthread()	KSD_GET64(k_curthread)
-#define	_ksd_get_tmbx()		KSD_GET64(k_mbx.km_curthread)
-#define	_ksd_set_tmbx(value)	KSD_SET64(k_mbx.km_curthread, (void *)value);
-#define	_ksd_readandclear_tmbx() KSD_READANDCLEAR64(k_mbx.km_curthread)
-
-
-static __inline int
-_ksd_create(struct ksd *ksd, void *base, int size)
-{
-	ksd->base = base;
-	ksd->size = size;
-	return (0);
-}
-
-static __inline void
-_ksd_destroy(struct ksd *ksd)
-{
-	ksd->base = 0;
-	ksd->size = 0;
-}
-
-static __inline int
-_ksd_setprivate(struct ksd *ksd)
-{
-	return (sysarch(AMD64_SET_FSBASE, &ksd->base));
-}
-
-#endif
diff --git a/lib/libpthread/arch/amd64/include/pthread_md.h b/lib/libpthread/arch/amd64/include/pthread_md.h
index 3c7cd03..27a07cc 100644
--- a/lib/libpthread/arch/amd64/include/pthread_md.h
+++ b/lib/libpthread/arch/amd64/include/pthread_md.h
@@ -1,27 +1,28 @@
-/*
- * Copyright (c) 2003 Marcel Moolenaar
+/*-
+ * Copyright (C) 2003 David Xu <davidxu@freebsd.org>
+ * Copyright (c) 2001 Daniel Eischen <deischen@freebsd.org>
  * 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.
+ * 2. Neither the name of the author nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
  *
- * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
- * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
- * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
- * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
- * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
- * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
- * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
  *
  * $FreeBSD$
  */
@@ -31,52 +32,224 @@
 #ifndef _PTHREAD_MD_H_
 #define	_PTHREAD_MD_H_
 
+#include <sys/types.h>
+#include <sys/kse.h>
+#include <machine/sysarch.h>
 #include <ucontext.h>
 
+/* <machine/sysarch.h> should define this, but doesn't. */
+extern int sysarch(int, void *);
+
 #define	THR_GETCONTEXT(ucp)	\
 	(void)_amd64_save_context(&(ucp)->uc_mcontext)
 #define	THR_SETCONTEXT(ucp)	\
-	(void)_amd64_restore_context(&(ucp)->uc_mcontext, NULL, NULL)
+	(void)_amd64_restore_context(&(ucp)->uc_mcontext, 0, NULL)
+
 
-#define	THR_ALIGNBYTES	15
-#define	THR_ALIGN(td)	(((uintptr_t)(td) + THR_ALIGNBYTES) & ~THR_ALIGNBYTES)
+#define	PER_KSE
+#undef	PER_THREAD
+
+struct kse;
+struct pthread;
+struct tdv;
 
 /*
- * KSE Specific Data.
+ * %fs points to a struct kcb.
  */
-struct ksd {
-	void	*base;
-	long	size;
+struct kcb {
+	struct tcb		*kcb_curtcb;
+	struct kcb		*kcb_self;	/* self reference */
+	struct kse		*kcb_kse;
+	struct kse_mailbox	kcb_kmbx;
+};
+
+struct tcb {
+	struct tdv		*tcb_tdv;
+	struct pthread		*tcb_thread;
+	void			*tcb_spare[2];	/* align tcb_tmbx to 16 bytes */
+	struct kse_thr_mailbox	tcb_tmbx;
 };
 
+/*
+ * Evaluates to the byte offset of the per-kse variable name.
+ */
+#define	__kcb_offset(name)	__offsetof(struct kcb, name)
+
+/*
+ * Evaluates to the type of the per-kse variable name.
+ */
+#define	__kcb_type(name)	__typeof(((struct kcb *)0)->name)
+
+/*
+ * Evaluates to the value of the per-kse variable name.
+ */
+#define	KCB_GET64(name) ({					\
+	__kcb_type(name) __result;				\
+								\
+	u_long __i;						\
+	__asm __volatile("movq %%fs:%1, %0"			\
+	    : "=r" (__i)					\
+	    : "m" (*(u_long *)(__kcb_offset(name))));		\
+	__result = *(__kcb_type(name) *)&__i;			\
+								\
+	__result;						\
+})
+
+/*
+ * Sets the value of the per-kse variable name to value val.
+ */
+#define	KCB_SET64(name, val) ({					\
+	__kcb_type(name) __val = (val);				\
+								\
+	u_long __i;						\
+	__i = *(u_long *)&__val;				\
+	__asm __volatile("movq %1,%%fs:%0"			\
+	    : "=m" (*(u_long *)(__kcb_offset(name)))		\
+	    : "r" (__i));					\
+})
+
+static __inline u_long
+__kcb_readandclear64(volatile u_long *addr)
+{
+	u_long result;
+
+	__asm __volatile (
+	    "	xorq	%0, %0;"
+	    "	xchgq	%%fs:%1, %0;"
+	    "# __kcb_readandclear64"
+	    : "=&r" (result)
+	    : "m" (*addr));
+	return (result);
+}
+
+#define	KCB_READANDCLEAR64(name) ({				\
+	__kcb_type(name) __result;				\
+								\
+	__result = (__kcb_type(name))				\
+	    __kcb_readandclear64((u_long *)__kcb_offset(name)); \
+	__result;						\
+})
+
+
+#define	_kcb_curkcb()		KCB_GET64(kcb_self)
+#define	_kcb_curtcb()		KCB_GET64(kcb_curtcb)
+#define	_kcb_curkse()		((struct kse *)KCB_GET64(kcb_kmbx.km_udata))
+#define	_kcb_get_tmbx()		KCB_GET64(kcb_kmbx.km_curthread)
+#define	_kcb_set_tmbx(value)	KCB_SET64(kcb_kmbx.km_curthread, (void *)value)
+#define	_kcb_readandclear_tmbx() KCB_READANDCLEAR64(kcb_kmbx.km_curthread)
+
+/*
+ * The constructors.
+ */
+struct tcb	*_tcb_ctor(struct pthread *);
+void		_tcb_dtor(struct tcb *tcb);
+struct kcb	*_kcb_ctor(struct kse *);
+void		_kcb_dtor(struct kcb *);
+
+/* Called from the KSE to set its private data. */
+static __inline void
+_kcb_set(struct kcb *kcb)
+{
+	void *addr = kcb;
+
+	sysarch(AMD64_SET_FSBASE, &addr);
+}
+
+/* Get the current kcb. */
+static __inline struct kcb *
+_kcb_get(void)
+{
+	return (_kcb_curkcb());
+}
+
+static __inline struct kse_thr_mailbox *
+_kcb_critical_enter(void)
+{
+	struct kse_thr_mailbox *crit;
+
+	crit = _kcb_readandclear_tmbx();
+	return (crit);
+}
+
+static __inline void
+_kcb_critical_leave(struct kse_thr_mailbox *crit)
+{
+	_kcb_set_tmbx(crit);
+}
+
+static __inline int
+_kcb_in_critical(void)
+{
+	return (_kcb_get_tmbx() == NULL);
+}
+
+static __inline void
+_tcb_set(struct kcb *kcb, struct tcb *tcb)
+{
+	kcb->kcb_curtcb = tcb;
+}
+
+static __inline struct tcb *
+_tcb_get(void)
+{
+	return (_kcb_curtcb());
+}
+
+static __inline struct pthread *
+_get_curthread(void)
+{
+	struct tcb *tcb;
+
+	tcb = _kcb_curtcb();
+	if (tcb != NULL)
+		return (tcb->tcb_thread);
+	else
+		return (NULL);
+}
+
+static __inline struct kse *
+_get_curkse(void)
+{
+	return ((struct kse *)_kcb_curkse());
+}
+
 void _amd64_enter_uts(struct kse_mailbox *km, kse_func_t uts, void *stack,
     size_t stacksz);
 int _amd64_restore_context(mcontext_t *mc, intptr_t val, intptr_t *loc);
 int _amd64_save_context(mcontext_t *mc);
 
 static __inline int
-_thread_enter_uts(struct kse_thr_mailbox *tm, struct kse_mailbox *km)
+_thread_enter_uts(struct tcb *tcb, struct kcb *kcb)
 {
-	if (tm == NULL)
-		return (-1);
-	if (!_amd64_save_context(&tm->tm_context.uc_mcontext)) {
-		_amd64_enter_uts(km, km->km_func, km->km_stack.ss_sp,
-		    km->km_stack.ss_size);
+	int ret;
+
+	ret = _amd64_save_context(&tcb->tcb_tmbx.tm_context.uc_mcontext);
+	if (ret == 0) {
+		_amd64_enter_uts(&kcb->kcb_kmbx, kcb->kcb_kmbx.km_func,
+		    kcb->kcb_kmbx.km_stack.ss_sp,
+		    kcb->kcb_kmbx.km_stack.ss_size);
 		/* We should not reach here. */
 		return (-1);
 	}
+	else if (ret < 0)
+		return (-1);
 	return (0);
 }
 
 static __inline int
-_thread_switch(struct kse_thr_mailbox *tm, struct kse_thr_mailbox **thrp)
+_thread_switch(struct kcb *kcb, struct tcb *tcb, int setmbox)
 {
-	if (tm == NULL)
+	if ((kcb == NULL) || (tcb == NULL))
 		return (-1);
-	_amd64_restore_context(&tm->tm_context.uc_mcontext, (intptr_t)tm,
-	    (intptr_t*)thrp);
+	kcb->kcb_curtcb = tcb;
+	if (setmbox != 0)
+		_amd64_restore_context(&tcb->tcb_tmbx.tm_context.uc_mcontext,
+		    (intptr_t)&tcb->tcb_tmbx,
+		    (intptr_t *)&kcb->kcb_kmbx.km_curthread);
+	else
+		_amd64_restore_context(&tcb->tcb_tmbx.tm_context.uc_mcontext,
+		    0, NULL);
 	/* We should not reach here. */
 	return (-1);
 }
-
 #endif
diff --git a/lib/libpthread/arch/i386/Makefile.inc b/lib/libpthread/arch/i386/Makefile.inc
index c5cbe18..73a9a8a 100644
--- a/lib/libpthread/arch/i386/Makefile.inc
+++ b/lib/libpthread/arch/i386/Makefile.inc
@@ -2,4 +2,4 @@
 
 .PATH:	${.CURDIR}/arch/${MACHINE_ARCH}/${MACHINE_ARCH}
 
-SRCS+=	ksd.c thr_enter_uts.S thr_getcontext.S
+SRCS+=	thr_enter_uts.S thr_getcontext.S pthread_md.c
diff --git a/lib/libpthread/arch/i386/i386/ksd.c b/lib/libpthread/arch/i386/i386/pthread_md.c
index 4e95959..aeaa649 100644
--- a/lib/libpthread/arch/i386/i386/ksd.c
+++ b/lib/libpthread/arch/i386/i386/pthread_md.c
@@ -40,7 +40,6 @@ __FBSDID("$FreeBSD$");
 #include <ucontext.h>
 
 #include "pthread_md.h"
-#include "ksd.h"
 
 #define LDT_ENTRIES 8192
 #define LDT_WORDS   (8192/sizeof(unsigned int))
@@ -94,76 +93,79 @@ free_ldt_entry(u_int index)
 	ldt_mask[i] |= (1 << j);
 }
 
-/*
- * Initialize KSD.  This also includes setting up the LDT.
- */
-int
-_ksd_create(struct ksd *ksd, void *base, int size)
+struct tcb *
+_tcb_ctor(struct pthread *thread)
 {
-	union descriptor ldt;
+	struct tcb *tcb;
+	void *addr;
 
-	if (initialized == 0)
-		initialize();
-	ksd->ldt = alloc_ldt_entry();
-	if (ksd->ldt == 0)
-		return (-1);
-	ksd->base = base;
-	ksd->size = size;
-	ldt.sd.sd_hibase = (unsigned int)ksd->base >> 24;
-	ldt.sd.sd_lobase = (unsigned int)ksd->base & 0xFFFFFF;
-	ldt.sd.sd_hilimit = (size >> 16) & 0xF;
-	ldt.sd.sd_lolimit = ksd->size & 0xFFFF;
-	ldt.sd.sd_type = SDT_MEMRWA;
-	ldt.sd.sd_dpl = SEL_UPL;
-	ldt.sd.sd_p = 1;
-	ldt.sd.sd_xx = 0;
-	ldt.sd.sd_def32 = 1;
-	ldt.sd.sd_gran = 0;	/* no more than 1M */
-	if (i386_set_ldt(ksd->ldt, &ldt, 1) < 0) {
-		free_ldt_entry(ksd->ldt);
-		return (-1);
+	addr = malloc(sizeof(struct tcb) + 15);
+	if (addr == NULL)
+		tcb = NULL;
+	else {
+		tcb = (struct tcb *)(((uintptr_t)(addr) + 15) & ~15);
+		bzero(tcb, sizeof(struct tcb));
+		tcb->tcb_addr = addr;
+		tcb->tcb_thread = thread;
+		/* XXX - Allocate tdv/tls */
 	}
-	ksd->flags = KSDF_INITIALIZED;
-	return (0);
+	return (tcb);
 }
 
 void
-_ksd_destroy(struct ksd *ksd)
-{
-	if ((ksd->flags & KSDF_INITIALIZED) != 0) {
-		free_ldt_entry(ksd->ldt);
-	}
-}
-
-int
-_ksd_getprivate(struct ksd *ksd, void **base, int *size)
+_tcb_dtor(struct tcb *tcb)
 {
+	void *addr;
 
-	if ((ksd == NULL) || ((ksd->flags & KSDF_INITIALIZED) == 0))
-		return (-1);
-	else {
-		*base = ksd->base;
-		*size = ksd->size;
-		return (0);
-	}
+	addr = tcb->tcb_addr;
+	tcb->tcb_addr = NULL;
+	free(addr);
 }
 
 /*
- * This assumes that the LDT is already setup.  Just set %gs to
- * reference it.
+ * Initialize KSD.  This also includes setting up the LDT.
  */
-int
-_ksd_setprivate(struct ksd *ksd)
+struct kcb *
+_kcb_ctor(struct kse *kse)
 {
-	int val;
-	int ret;
+	union descriptor ldt;
+	struct kcb *kcb;
 
-	if ((ksd->flags & KSDF_INITIALIZED) == 0)
-		ret = -1;
-	else {
-		val = (ksd->ldt << 3) | 7;
-		__asm __volatile("movl %0, %%gs" : : "r" (val));
-		ret = 0;
+	if (initialized == 0)
+		initialize();
+	kcb = malloc(sizeof(struct kcb));
+	if (kcb != NULL) {
+		bzero(kcb, sizeof(struct kcb));
+		kcb->kcb_self = kcb;
+		kcb->kcb_kse = kse;
+		kcb->kcb_ldt = alloc_ldt_entry();
+		if (kcb->kcb_ldt == 0) {
+			free(kcb);
+			return (NULL);
+		}
+		ldt.sd.sd_hibase = (unsigned int)kcb >> 24;
+		ldt.sd.sd_lobase = (unsigned int)kcb & 0xFFFFFF;
+		ldt.sd.sd_hilimit = (sizeof(struct kcb) >> 16) & 0xF;
+		ldt.sd.sd_lolimit = sizeof(struct kcb) & 0xFFFF;
+		ldt.sd.sd_type = SDT_MEMRWA;
+		ldt.sd.sd_dpl = SEL_UPL;
+		ldt.sd.sd_p = 1;
+		ldt.sd.sd_xx = 0;
+		ldt.sd.sd_def32 = 1;
+		ldt.sd.sd_gran = 0;	/* no more than 1M */
+		if (i386_set_ldt(kcb->kcb_ldt, &ldt, 1) < 0) {
+			free_ldt_entry(kcb->kcb_ldt);
+			free(kcb);
+			return (NULL);
+		}
 	}
-	return (ret);
+	return (kcb);
+}
+
+void
+_kcb_dtor(struct kcb *kcb)
+{
+	if (kcb->kcb_ldt != -1)
+		free_ldt_entry(kcb->kcb_ldt);
+	free(kcb);
 }
diff --git a/lib/libpthread/arch/i386/i386/thr_getcontext.S b/lib/libpthread/arch/i386/i386/thr_getcontext.S
index 19afd06..7f58dce 100644
--- a/lib/libpthread/arch/i386/i386/thr_getcontext.S
+++ b/lib/libpthread/arch/i386/i386/thr_getcontext.S
@@ -74,7 +74,7 @@ ENTRY(__thr_setcontext)
 	movl	72(%edx), %esp		/* switch to context defined stack */
 	pushl	60(%edx)		/* push return address on stack */
 	pushl	44(%edx)		/* push ecx on stack */
-	push	48(%edx)		/* push eax on stack */
+	pushl	48(%edx)		/* push eax on stack */
 	/*
 	 * if (mc_fpowned == MC_OWNEDFP_FPU || mc_fpowned == MC_OWNEDFP_PCB) {
 	 *    if (mc_fpformat == MC_FPFMT_387)
diff --git a/lib/libpthread/arch/i386/include/ksd.h b/lib/libpthread/arch/i386/include/ksd.h
deleted file mode 100644
index f7f728f..0000000
--- a/lib/libpthread/arch/i386/include/ksd.h
+++ /dev/null
@@ -1,111 +0,0 @@
-/*-
- * Copyright (C) 2003 David Xu <davidxu@freebsd.org>
- * Copyright (c) 2001 Daniel Eischen <deischen@freebsd.org>
- * 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. Neither the name of the author nor the names of its contributors
- *    may be used to endorse or promote products derived from this software
- *    without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
- * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
- * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
- * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
- * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
- * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
- * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
- * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
- * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- */
-
-/*
- * $FreeBSD$
- */
-
-#ifndef	_KSD_H_
-#define	_KSD_H_
-
-#include <sys/types.h>
-
-struct kse;
-
-/*
- * Evaluates to the byte offset of the per-kse variable name.
- */
-#define	__ksd_offset(name)	__offsetof(struct kse, name)
-
-/*
- * Evaluates to the type of the per-kse variable name.
- */
-#define	__ksd_type(name)	__typeof(((struct kse *)0)->name)
-
-
-/*
- * Evaluates to the value of the per-kse variable name.
- */
-#define	KSD_GET32(name) ({					\
-	__ksd_type(name) __result;				\
-								\
-	u_int __i;						\
-	__asm __volatile("movl %%gs:%1, %0"			\
-	    : "=r" (__i)					\
-	    : "m" (*(u_int *)(__ksd_offset(name))));		\
-	__result = *(__ksd_type(name) *)&__i;			\
-								\
-	__result;						\
-})
-
-/*
- * Sets the value of the per-cpu variable name to value val.
- */
-#define	KSD_SET32(name, val) ({					\
-	__ksd_type(name) __val = (val);				\
-								\
-	u_int __i;						\
-	__i = *(u_int *)&__val;					\
-	__asm __volatile("movl %1,%%gs:%0"			\
-	    : "=m" (*(u_int *)(__ksd_offset(name)))		\
-	    : "r" (__i));					\
-})
-
-static __inline u_long
-__ksd_readandclear32(volatile u_long *addr)
-{
-	u_long result;
-
-	__asm __volatile (
-	    "	xorl	%0, %0;"
-	    "	xchgl	%%gs:%1, %0;"
-	    "# __ksd_readandclear32"
-	    : "=&r" (result)
-	    : "m" (*addr));
-	return (result);
-}
-
-#define	KSD_READANDCLEAR32(name) ({				\
-	__ksd_type(name) __result;				\
-								\
-	__result = (__ksd_type(name))				\
-	    __ksd_readandclear32((u_long *)__ksd_offset(name)); \
-	__result;						\
-})
-
-
-#define	_ksd_curkse()		((struct kse *)KSD_GET32(k_mbx.km_udata))
-#define	_ksd_curthread()	KSD_GET32(k_curthread)
-#define	_ksd_get_tmbx()		KSD_GET32(k_mbx.km_curthread)
-#define	_ksd_set_tmbx(value)	KSD_SET32(k_mbx.km_curthread, (void *)value);
-#define	_ksd_readandclear_tmbx() KSD_READANDCLEAR32(k_mbx.km_curthread)
-
-int	_ksd_create(struct ksd *ksd, void *base, int size);
-void	_ksd_destroy(struct ksd *ksd);
-int	_ksd_getprivate(struct ksd *ksd, void **base, int *size);
-int	_ksd_setprivate(struct ksd *ksd);
-#endif
diff --git a/lib/libpthread/arch/i386/include/pthread_md.h b/lib/libpthread/arch/i386/include/pthread_md.h
index 980680a..9718562 100644
--- a/lib/libpthread/arch/i386/include/pthread_md.h
+++ b/lib/libpthread/arch/i386/include/pthread_md.h
@@ -37,34 +37,193 @@
 extern int _thr_setcontext(mcontext_t *, intptr_t, intptr_t *);
 extern int _thr_getcontext(mcontext_t *);
 
-#define	THR_GETCONTEXT(ucp)	_thr_getcontext(&(ucp)->uc_mcontext);
-#define	THR_SETCONTEXT(ucp)	_thr_setcontext(&(ucp)->uc_mcontext, NULL, NULL);
+#define	THR_GETCONTEXT(ucp)	_thr_getcontext(&(ucp)->uc_mcontext)
+#define	THR_SETCONTEXT(ucp)	_thr_setcontext(&(ucp)->uc_mcontext, 0, NULL)
 
-#define	THR_ALIGNBYTES	15
-#define	THR_ALIGN(td)	(((unsigned)(td) + THR_ALIGNBYTES) & ~THR_ALIGNBYTES)
+#define	PER_KSE
+#undef	PER_THREAD
+
+struct kse;
+struct pthread;
+struct tdv;
 
 /*
- * KSE Specific Data.
+ * %gs points to a struct kcb.
  */
-struct ksd {
-	int	ldt;
-#define	KSDF_INITIALIZED	0x01
-	long	flags;
-	void	*base;
-	long	size;
+struct kcb {
+	struct tcb		*kcb_curtcb;
+	struct kcb		*kcb_self;	/* self reference */
+	int			kcb_ldt;
+	struct kse		*kcb_kse;
+	struct kse_mailbox	kcb_kmbx;
+};
+
+struct tcb {
+	struct tdv		*tcb_tdv;
+	struct pthread		*tcb_thread;
+	void			*tcb_addr;	/* allocated tcb address */
+	void			*tcb_spare;	/* align tcb_tmbx to 16 bytes */
+	struct kse_thr_mailbox	tcb_tmbx;
 };
 
-extern void _i386_enter_uts(struct kse_mailbox *, kse_func_t, void *, long);
+/*
+ * Evaluates to the byte offset of the per-kse variable name.
+ */
+#define	__kcb_offset(name)	__offsetof(struct kcb, name)
+
+/*
+ * Evaluates to the type of the per-kse variable name.
+ */
+#define	__kcb_type(name)	__typeof(((struct kcb *)0)->name)
+
+/*
+ * Evaluates to the value of the per-kse variable name.
+ */
+#define	KCB_GET32(name) ({					\
+	__kcb_type(name) __result;				\
+								\
+	u_int __i;						\
+	__asm __volatile("movl %%gs:%1, %0"			\
+	    : "=r" (__i)					\
+	    : "m" (*(u_int *)(__kcb_offset(name))));		\
+	__result = *(__kcb_type(name) *)&__i;			\
+								\
+	__result;						\
+})
+
+/*
+ * Sets the value of the per-kse variable name to value val.
+ */
+#define	KCB_SET32(name, val) ({					\
+	__kcb_type(name) __val = (val);				\
+								\
+	u_int __i;						\
+	__i = *(u_int *)&__val;					\
+	__asm __volatile("movl %1,%%gs:%0"			\
+	    : "=m" (*(u_int *)(__kcb_offset(name)))		\
+	    : "r" (__i));					\
+})
+
+static __inline u_long
+__kcb_readandclear32(volatile u_long *addr)
+{
+	u_long result;
+
+	__asm __volatile (
+	    "	xorl	%0, %0;"
+	    "	xchgl	%%gs:%1, %0;"
+	    "# __kcb_readandclear32"
+	    : "=&r" (result)
+	    : "m" (*addr));
+	return (result);
+}
+
+#define	KCB_READANDCLEAR32(name) ({				\
+	__kcb_type(name) __result;				\
+								\
+	__result = (__kcb_type(name))				\
+	    __kcb_readandclear32((u_long *)__kcb_offset(name)); \
+	__result;						\
+})
+
+
+#define	_kcb_curkcb()		KCB_GET32(kcb_self)
+#define	_kcb_curtcb()		KCB_GET32(kcb_curtcb)
+#define	_kcb_curkse()		((struct kse *)KCB_GET32(kcb_kmbx.km_udata))
+#define	_kcb_get_tmbx()		KCB_GET32(kcb_kmbx.km_curthread)
+#define	_kcb_set_tmbx(value)	KCB_SET32(kcb_kmbx.km_curthread, (void *)value)
+#define	_kcb_readandclear_tmbx() KCB_READANDCLEAR32(kcb_kmbx.km_curthread)
+
+
+/*
+ * The constructors.
+ */
+struct tcb	*_tcb_ctor(struct pthread *);
+void		_tcb_dtor(struct tcb *tcb);
+struct kcb	*_kcb_ctor(struct kse *);
+void		_kcb_dtor(struct kcb *);
+
+/* Called from the KSE to set its private data. */
+static __inline void
+_kcb_set(struct kcb *kcb)
+{
+	int val;
+
+	val = (kcb->kcb_ldt << 3) | 7;
+	__asm __volatile("movl %0, %%gs" : : "r" (val));
+}
+
+/* Get the current kcb. */
+static __inline struct kcb *
+_kcb_get(void)
+{
+	return (_kcb_curkcb());
+}
+
+static __inline struct kse_thr_mailbox *
+_kcb_critical_enter(void)
+{
+	struct kse_thr_mailbox *crit;
+
+	crit = _kcb_readandclear_tmbx();
+	return (crit);
+}
+
+static __inline void
+_kcb_critical_leave(struct kse_thr_mailbox *crit)
+{
+	_kcb_set_tmbx(crit);
+}
 
 static __inline int
-_thread_enter_uts(struct kse_thr_mailbox *tmbx, struct kse_mailbox *kmbx)
+_kcb_in_critical(void)
+{
+	return (_kcb_get_tmbx() == NULL);
+}
+
+static __inline void
+_tcb_set(struct kcb *kcb, struct tcb *tcb)
+{
+	kcb->kcb_curtcb = tcb;
+}
+
+static __inline struct tcb *
+_tcb_get(void)
+{
+	return (_kcb_curtcb());
+}
+
+static __inline struct pthread *
+_get_curthread(void)
+{
+	struct tcb *tcb;
+
+	tcb = _kcb_curtcb();
+	if (tcb != NULL)
+		return (tcb->tcb_thread);
+	else
+		return (NULL);
+}
+
+static __inline struct kse *
+_get_curkse(void)
+{
+	return ((struct kse *)_kcb_curkse());
+}
+
+void	_i386_enter_uts(struct kse_mailbox *km, kse_func_t uts, void *stack,
+    size_t stacksz);
+
+static __inline int
+_thread_enter_uts(struct tcb *tcb, struct kcb *kcb)
 {
 	int ret;
 
-	ret = _thr_getcontext(&tmbx->tm_context.uc_mcontext);
+	ret = _thr_getcontext(&tcb->tcb_tmbx.tm_context.uc_mcontext);
 	if (ret == 0) {
-		_i386_enter_uts(kmbx, kmbx->km_func,
-		    kmbx->km_stack.ss_sp, kmbx->km_stack.ss_size);
+		_i386_enter_uts(&kcb->kcb_kmbx, kcb->kcb_kmbx.km_func,
+		    kcb->kcb_kmbx.km_stack.ss_sp,
+		    kcb->kcb_kmbx.km_stack.ss_size);
 		/* We should not reach here. */
 		return (-1);
 	}
@@ -74,10 +233,17 @@ _thread_enter_uts(struct kse_thr_mailbox *tmbx, struct kse_mailbox *kmbx)
 }
 
 static __inline int
-_thread_switch(struct kse_thr_mailbox *tmbx, struct kse_thr_mailbox **loc)
+_thread_switch(struct kcb *kcb, struct tcb *tcb, int setmbox)
 {
-	_thr_setcontext(&tmbx->tm_context.uc_mcontext,
-	    (intptr_t)tmbx, (intptr_t *)loc);
+	if ((kcb == NULL) || (tcb == NULL))
+		return (-1);
+	kcb->kcb_curtcb = tcb;
+	if (setmbox != 0)
+		_thr_setcontext(&tcb->tcb_tmbx.tm_context.uc_mcontext,
+		    (intptr_t)&tcb->tcb_tmbx,
+		    (intptr_t *)&kcb->kcb_kmbx.km_curthread);
+	else
+		_thr_setcontext(&tcb->tcb_tmbx.tm_context.uc_mcontext, 0, NULL);
 	/* We should not reach here. */
 	return (-1);
 }
diff --git a/lib/libpthread/arch/ia64/Makefile.inc b/lib/libpthread/arch/ia64/Makefile.inc
index c0e4c47..c8b0362 100644
--- a/lib/libpthread/arch/ia64/Makefile.inc
+++ b/lib/libpthread/arch/ia64/Makefile.inc
@@ -2,4 +2,4 @@
 
 .PATH:	${.CURDIR}/arch/${MACHINE_ARCH}/${MACHINE_ARCH}
 
-SRCS+=	context.S enter_uts.S
+SRCS+=	context.S enter_uts.S pthread_md.c
diff --git a/lib/libpthread/arch/ia64/ia64/pthread_md.c b/lib/libpthread/arch/ia64/ia64/pthread_md.c
new file mode 100644
index 0000000..e8fd64e
--- /dev/null
+++ b/lib/libpthread/arch/ia64/ia64/pthread_md.c
@@ -0,0 +1,77 @@
+/*
+ * Copyright (c) 2003 Daniel Eischen <deischen@freebsd.org>
+ * 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. Neither the name of the author nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
+ * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
+ * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
+ * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
+ * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ * $FreeBSD$
+ */
+
+#include <stdlib.h>
+#include <strings.h>
+#include "pthread_md.h"
+
+/*
+ * The constructors.
+ */
+struct tcb *
+_tcb_ctor(struct pthread *thread)
+{
+	struct tcb *tcb;
+
+	if ((tcb = malloc(sizeof(struct tcb))) != NULL) {
+		bzero(tcb, sizeof(struct tcb));
+		tcb->tcb_thread = thread;
+		tcb->tcb_tp.tp_self = tcb;
+		/* Allocate TDV */
+	}
+	return (tcb);
+}
+
+void
+_tcb_dtor(struct tcb *tcb)
+{
+	/* Free TDV */
+	free(tcb);
+}
+
+struct kcb *
+_kcb_ctor(struct kse *kse)
+{
+	struct kcb *kcb;
+
+	if ((kcb = malloc(sizeof(struct kcb))) != NULL) {
+		bzero(kcb, sizeof(struct kcb));
+		kcb->kcb_faketcb.tcb_isfake = 1;
+		kcb->kcb_faketcb.tcb_tmbx.tm_flags = TMF_NOUPCALL;
+		kcb->kcb_faketcb.tcb_tp.tp_self = &kcb->kcb_faketcb;
+		kcb->kcb_curtcb = &kcb->kcb_faketcb;
+		kcb->kcb_kse = kse;
+	}
+	return (kcb);
+}
+
+void
+_kcb_dtor(struct kcb *kcb)
+{
+	free(kcb);
+}
diff --git a/lib/libpthread/arch/ia64/include/ksd.h b/lib/libpthread/arch/ia64/include/ksd.h
deleted file mode 100644
index 8878132..0000000
--- a/lib/libpthread/arch/ia64/include/ksd.h
+++ /dev/null
@@ -1,91 +0,0 @@
-/*
- * Copyright (c) 2003 Marcel Moolenaar
- * 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.
- *
- * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
- * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
- * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
- * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
- * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
- * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
- * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- *
- * $FreeBSD$
- */
-
-#ifndef	_KSD_H_
-#define	_KSD_H_
-
-struct kse;
-struct pthread;
-
-register struct kse *_tp __asm("%r13");
-
-static __inline int
-_ksd_create(struct ksd *ksd, void *base, int size)
-{
-	ksd->ksd_base = base;
-	ksd->ksd_size = size;
-	return (0);
-}
-
-static __inline struct kse *
-_ksd_curkse()
-{
-	/* XXX why not simply return _tp? */
-	return ((struct kse *)_tp->k_mbx.km_udata);
-}
-
-static __inline struct pthread *
-_ksd_curthread()
-{
-	return (_tp->k_curthread);
-}
-
-static __inline void
-_ksd_destroy(struct ksd *ksd)
-{
-}
-
-static __inline kse_critical_t
-_ksd_get_tmbx()
-{
-	return (_tp->k_mbx.km_curthread);
-}
-
-static __inline kse_critical_t
-_ksd_readandclear_tmbx()
-{
-	kse_critical_t crit;
-	__asm("xchg8    %0=[%1],r0" : "=r"(crit)
-	    : "r"(&_tp->k_mbx.km_curthread));
-	return (crit);
-}
-
-static __inline void
-_ksd_set_tmbx(kse_critical_t crit)
-{
-	_tp->k_mbx.km_curthread = crit;
-}
-
-static __inline int
-_ksd_setprivate(struct ksd *ksd)
-{
-	_tp = (struct kse *)ksd->ksd_base;
-	return (0);
-}
-
-#endif /* _KSD_H_ */
diff --git a/lib/libpthread/arch/ia64/include/pthread_md.h b/lib/libpthread/arch/ia64/include/pthread_md.h
index 6d30025..f7076d0 100644
--- a/lib/libpthread/arch/ia64/include/pthread_md.h
+++ b/lib/libpthread/arch/ia64/include/pthread_md.h
@@ -29,32 +29,190 @@
 #ifndef _PTHREAD_MD_H_
 #define	_PTHREAD_MD_H_
 
+#include <sys/kse.h>
+#include <ucontext.h>
+
 #define	THR_GETCONTEXT(ucp)	_ia64_save_context(&(ucp)->uc_mcontext)
 #define	THR_SETCONTEXT(ucp)	_ia64_restore_context(&(ucp)->uc_mcontext, \
 				    0, NULL)
 
-#define	THR_ALIGNBYTES	0
-#define	THR_ALIGN(td)	(td)
+#define	PER_THREAD
+
+struct kcb;
+struct kse;
+struct pthread;
+struct tcb;
+struct tdv;	/* We don't know what this is yet? */
+
+/*
+ * tp points to one of these.
+ */
+struct ia64_tp {
+	struct tdv		*tp_tdv;	/* dynamic TLS */
+	struct tcb		*tp_self;
+	char			tp_tls[0];	/* static TLS */
+};
+
+struct tcb {
+	struct kse_thr_mailbox	tcb_tmbx;
+	struct pthread		*tcb_thread;
+	struct kcb		*tcb_curkcb;
+	long			tcb_isfake;
+	struct ia64_tp		tcb_tp;
+};
 
-/* KSE Specific Data. */
-struct ksd {
-	void	*ksd_base;
-	int	ksd_size;
+struct kcb {
+	struct kse_mailbox	kcb_kmbx;
+	struct tcb		kcb_faketcb;
+	struct tcb		*kcb_curtcb;
+	struct kse		*kcb_kse;
 };
 
+register struct ia64_tp *_tp __asm("%r13");
+
+/*
+ * The kcb and tcb constructors.
+ */
+struct tcb	*_tcb_ctor(struct pthread *);
+void		_tcb_dtor(struct tcb *);
+struct kcb	*_kcb_ctor(struct kse *kse);
+void		_kcb_dtor(struct kcb *);
+
+/* Called from the KSE to set its private data. */
+static __inline void
+_kcb_set(struct kcb *kcb)
+{
+	/* There is no thread yet; use the fake tcb. */
+	_tp = &kcb->kcb_faketcb.tcb_tp;
+}
+
+/*
+ * Get the current kcb.
+ *
+ * This can only be called while in a critical region; don't
+ * worry about having the kcb changed out from under us.
+ */
+static __inline struct kcb *
+_kcb_get(void)
+{
+	return (_tp->tp_self->tcb_curkcb);
+}
+
+/*
+ * Enter a critical region.
+ *
+ * Read and clear km_curthread in the kse mailbox.
+ */
+static __inline struct kse_thr_mailbox *
+_kcb_critical_enter(void)
+{
+	struct kse_thr_mailbox *crit;
+	struct tcb *tcb;
+	uint32_t flags;
+
+	tcb = _tp->tp_self;
+	if (tcb->tcb_isfake != 0) {
+		/*
+		 * We already are in a critical region since
+		 * there is no current thread.
+		 */
+		crit = NULL;
+	} else {
+		flags = tcb->tcb_tmbx.tm_flags;
+		tcb->tcb_tmbx.tm_flags |= TMF_NOUPCALL;
+		crit = tcb->tcb_curkcb->kcb_kmbx.km_curthread;
+		tcb->tcb_curkcb->kcb_kmbx.km_curthread = NULL;
+		tcb->tcb_tmbx.tm_flags = flags;
+	}
+	return (crit);
+}
+
+static __inline void
+_kcb_critical_leave(struct kse_thr_mailbox *crit)
+{
+	struct tcb *tcb;
+
+	tcb = _tp->tp_self;
+	/* No need to do anything if this is a fake tcb. */
+	if (tcb->tcb_isfake == 0)
+		tcb->tcb_curkcb->kcb_kmbx.km_curthread = crit;
+}
+
+static __inline int
+_kcb_in_critical(void)
+{
+	struct tcb *tcb;
+	uint32_t flags;
+	int ret;
+
+	tcb = _tp->tp_self;
+	if (tcb->tcb_isfake != 0) {
+		/*
+		 * We are in a critical region since there is no
+		 * current thread.
+		 */
+		ret = 1;
+	} else {
+		flags = tcb->tcb_tmbx.tm_flags;
+		tcb->tcb_tmbx.tm_flags |= TMF_NOUPCALL;
+		ret = (tcb->tcb_curkcb->kcb_kmbx.km_curthread == NULL);
+		tcb->tcb_tmbx.tm_flags = flags;
+	}
+	return (ret);
+}
+
+static __inline void
+_tcb_set(struct kcb *kcb, struct tcb *tcb)
+{
+	if (tcb == NULL) {
+		kcb->kcb_curtcb = &kcb->kcb_faketcb;
+		_tp = &kcb->kcb_faketcb.tcb_tp;
+	}
+	else {
+		kcb->kcb_curtcb = tcb;
+		tcb->tcb_curkcb = kcb;
+		_tp = &tcb->tcb_tp;
+	}
+}
+
+static __inline struct tcb *
+_tcb_get(void)
+{
+	return (_tp->tp_self);
+}
+
+static __inline struct pthread *
+_get_curthread(void)
+{
+	return (_tp->tp_self->tcb_thread);
+}
+
+/*
+ * Get the current kse.
+ *
+ * Line _kcb_get(), this can only be called while in a critical region.
+ */
+static __inline struct kse *
+_get_curkse(void)
+{
+	return (_tp->tp_self->tcb_curkcb->kcb_kse);
+}
+
 void _ia64_enter_uts(kse_func_t uts, struct kse_mailbox *km, void *stack,
     size_t stacksz);
 int _ia64_restore_context(mcontext_t *mc, intptr_t val, intptr_t *loc);
 int _ia64_save_context(mcontext_t *mc);
 
 static __inline int
-_thread_enter_uts(struct kse_thr_mailbox *tm, struct kse_mailbox *km)
+_thread_enter_uts(struct tcb *tcb, struct kcb *kcb)
 {
-	if (tm == NULL)
-		return (-1);
-	if (!_ia64_save_context(&tm->tm_context.uc_mcontext)) {
-		_ia64_enter_uts(km->km_func, km, km->km_stack.ss_sp,
-		    km->km_stack.ss_size);
+	if (_ia64_save_context(&tcb->tcb_tmbx.tm_context.uc_mcontext) == 0) {
+		/* Make the fake tcb the current thread. */
+		kcb->kcb_curtcb = &kcb->kcb_faketcb;
+		_tp = &kcb->kcb_faketcb.tcb_tp;
+		_ia64_enter_uts(kcb->kcb_kmbx.km_func, &kcb->kcb_kmbx,
+		    kcb->kcb_kmbx.km_stack.ss_sp,
+		    kcb->kcb_kmbx.km_stack.ss_size);
 		/* We should not reach here. */
 		return (-1);
 	}
@@ -62,12 +220,18 @@ _thread_enter_uts(struct kse_thr_mailbox *tm, struct kse_mailbox *km)
 }
 
 static __inline int
-_thread_switch(struct kse_thr_mailbox *tm, struct kse_thr_mailbox **thrp)
+_thread_switch(struct kcb *kcb, struct tcb *tcb, int setmbox)
 {
-	if (tm == NULL)
-		return (-1);
-	_ia64_restore_context(&tm->tm_context.uc_mcontext, (intptr_t)tm,
-	    (intptr_t*)thrp);
+	kcb->kcb_curtcb = tcb;
+	tcb->tcb_curkcb = kcb;
+	_tp = &tcb->tcb_tp;
+	if (setmbox != 0)
+		_ia64_restore_context(&tcb->tcb_tmbx.tm_context.uc_mcontext,
+		    (intptr_t)&tcb->tcb_tmbx,
+		    (intptr_t *)&kcb->kcb_kmbx.km_curthread);
+	else
+		_ia64_restore_context(&tcb->tcb_tmbx.tm_context.uc_mcontext,
+		    0, NULL);
 	/* We should not reach here. */
 	return (-1);
 }
diff --git a/lib/libpthread/thread/thr_cancel.c b/lib/libpthread/thread/thr_cancel.c
index 581b719..8190af6 100644
--- a/lib/libpthread/thread/thr_cancel.c
+++ b/lib/libpthread/thread/thr_cancel.c
@@ -111,7 +111,7 @@ _pthread_cancel(pthread_t pthread)
 			if ((pthread->cancelflags & THR_AT_CANCEL_POINT) &&
 			    (pthread->blocked != 0 ||
 			     pthread->attr.flags & PTHREAD_SCOPE_SYSTEM))
-				kse_thr_interrupt(&pthread->tmbx,
+				kse_thr_interrupt(&pthread->tcb->tcb_tmbx,
 					KSE_INTR_INTERRUPT, 0);
 		}
 
diff --git a/lib/libpthread/thread/thr_concurrency.c b/lib/libpthread/thread/thr_concurrency.c
index 8eaa6c1..694255a 100644
--- a/lib/libpthread/thread/thr_concurrency.c
+++ b/lib/libpthread/thread/thr_concurrency.c
@@ -105,7 +105,7 @@ _thr_setconcurrency(int new_level)
 		newkse->k_kseg->kg_ksecount++;
 		newkse->k_flags |= KF_STARTED;
 		KSE_SCHED_UNLOCK(curthread->kse, newkse->k_kseg);
-		if (kse_create(&newkse->k_mbx, 0) != 0) {
+		if (kse_create(&newkse->k_kcb->kcb_kmbx, 0) != 0) {
 			KSE_SCHED_LOCK(curthread->kse, newkse->k_kseg);
 			TAILQ_REMOVE(&newkse->k_kseg->kg_kseq,
 			    newkse, k_kgqe);
diff --git a/lib/libpthread/thread/thr_create.c b/lib/libpthread/thread/thr_create.c
index 9f04136..fcc4ac5 100644
--- a/lib/libpthread/thread/thr_create.c
+++ b/lib/libpthread/thread/thr_create.c
@@ -49,7 +49,10 @@ int _thread_next_offset			= OFF(tle.tqe_next);
 int _thread_uniqueid_offset		= OFF(uniqueid);
 int _thread_state_offset		= OFF(state);
 int _thread_name_offset			= OFF(name);
-int _thread_ctx_offset			= OFF(tmbx.tm_context);
+void *_thread_tcb_offset		= OFF(tcb);
+#undef OFF
+#define OFF(f)	offsetof(struct tcb, f)
+int _thread_ctx_offset			= OFF(tcb_tmbx.tm_context);
 #undef OFF
 
 int _thread_PS_RUNNING_value		= PS_RUNNING;
@@ -95,7 +98,6 @@ _pthread_create(pthread_t * thread, const pthread_attr_t * attr,
 	struct pthread *curthread, *new_thread;
 	struct kse *kse = NULL;
 	struct kse_group *kseg = NULL;
-	void *p;
 	kse_critical_t crit;
 	int i;
 	int ret = 0;
@@ -121,11 +123,6 @@ _pthread_create(pthread_t * thread, const pthread_attr_t * attr,
 		/* Insufficient memory to create a thread: */
 		ret = EAGAIN;
 	} else {
-		/* Initialize the thread structure: */
-		p = new_thread->alloc_addr;
-		memset(new_thread, 0, sizeof(struct pthread));
-		new_thread->alloc_addr = p;
-
 		/* Check if default thread attributes are required: */
 		if (attr == NULL || *attr == NULL)
 			/* Use the default thread attributes: */
@@ -146,7 +143,7 @@ _pthread_create(pthread_t * thread, const pthread_attr_t * attr,
 			/* Insufficient memory to create a new KSE/KSEG: */
 			ret = EAGAIN;
 			if (kse != NULL) {
-				kse->k_mbx.km_flags |= KMF_DONE;
+				kse->k_kcb->kcb_kmbx.km_flags |= KMF_DONE;
 				_kse_free(curthread, kse);
 			}
 			free_stack(&new_thread->attr);
@@ -183,18 +180,19 @@ _pthread_create(pthread_t * thread, const pthread_attr_t * attr,
 			 * Enter a critical region to get consistent context.
 			 */
 			crit = _kse_critical_enter();
-			THR_GETCONTEXT(&new_thread->tmbx.tm_context);
+			THR_GETCONTEXT(&new_thread->tcb->tcb_tmbx.tm_context);
 			/* Initialize the thread for signals: */
 			new_thread->sigmask = curthread->sigmask;
 			_kse_critical_leave(crit);
-			new_thread->tmbx.tm_udata = new_thread;
-			new_thread->tmbx.tm_context.uc_sigmask =
+
+			new_thread->tcb->tcb_tmbx.tm_udata = new_thread;
+			new_thread->tcb->tcb_tmbx.tm_context.uc_sigmask =
 			    new_thread->sigmask;
-			new_thread->tmbx.tm_context.uc_stack.ss_size =
+			new_thread->tcb->tcb_tmbx.tm_context.uc_stack.ss_size =
 			    new_thread->attr.stacksize_attr;
-			new_thread->tmbx.tm_context.uc_stack.ss_sp =
+			new_thread->tcb->tcb_tmbx.tm_context.uc_stack.ss_sp =
 			    new_thread->attr.stackaddr_attr;
-			makecontext(&new_thread->tmbx.tm_context,
+			makecontext(&new_thread->tcb->tcb_tmbx.tm_context,
 			    (void (*)(void))thread_start, 4, new_thread,
 			    start_routine, arg);
 			/*
@@ -274,8 +272,8 @@ _pthread_create(pthread_t * thread, const pthread_attr_t * attr,
 				kse->k_kseg->kg_flags |= KGF_SINGLE_THREAD;
 				new_thread->kse = kse;
 				new_thread->kseg = kse->k_kseg;
-				kse->k_mbx.km_udata = kse;
-				kse->k_mbx.km_curthread = NULL;
+				kse->k_kcb->kcb_kmbx.km_udata = kse;
+				kse->k_kcb->kcb_kmbx.km_curthread = NULL;
 			}
 
 			/*
diff --git a/lib/libpthread/thread/thr_init.c b/lib/libpthread/thread/thr_init.c
index 5c31ae7..f9e0d9b 100644
--- a/lib/libpthread/thread/thr_init.c
+++ b/lib/libpthread/thread/thr_init.c
@@ -68,7 +68,6 @@
 
 #include "libc_private.h"
 #include "thr_private.h"
-#include "ksd.h"
 
 int	__pthread_cond_wait(pthread_cond_t *, pthread_mutex_t *);
 int	__pthread_mutex_lock(pthread_mutex_t *);
@@ -306,12 +305,10 @@ _libpthread_init(struct pthread *curthread)
 	KSEG_THRQ_ADD(_kse_initial->k_kseg, _thr_initial);
 
 	/* Setup the KSE/thread specific data for the current KSE/thread. */
-	if (_ksd_setprivate(&_thr_initial->kse->k_ksd) != 0)
-		PANIC("Can't set initial KSE specific data");
-	_set_curkse(_thr_initial->kse);
 	_thr_initial->kse->k_curthread = _thr_initial;
+	_kcb_set(_thr_initial->kse->k_kcb);
+	_tcb_set(_thr_initial->kse->k_kcb, _thr_initial->tcb);
 	_thr_initial->kse->k_flags |= KF_INITIALIZED;
-	_kse_initial->k_curthread = _thr_initial;
 
 	_thr_rtld_init();
 }
@@ -323,14 +320,8 @@ _libpthread_init(struct pthread *curthread)
 static void
 init_main_thread(struct pthread *thread)
 {
-	void *p;
 	int i;
 
-	/* Zero the initial thread structure. */
-	p = thread->alloc_addr;
-	memset(thread, 0, sizeof(struct pthread));
-	thread->alloc_addr = p;
-
 	/* Setup the thread attributes. */
 	thread->attr = _pthread_attr_default;
 #ifdef SYSTEM_SCOPE_ONLY
@@ -381,9 +372,11 @@ init_main_thread(struct pthread *thread)
 	 * Set up the thread mailbox.  The threads saved context
 	 * is also in the mailbox.
 	 */
-	thread->tmbx.tm_udata = thread;
-	thread->tmbx.tm_context.uc_stack.ss_size = thread->attr.stacksize_attr;
-	thread->tmbx.tm_context.uc_stack.ss_sp = thread->attr.stackaddr_attr;
+	thread->tcb->tcb_tmbx.tm_udata = thread;
+	thread->tcb->tcb_tmbx.tm_context.uc_stack.ss_size =
+	    thread->attr.stacksize_attr;
+	thread->tcb->tcb_tmbx.tm_context.uc_stack.ss_sp =
+	    thread->attr.stackaddr_attr;
 
 	/* Default the priority of the initial thread: */
 	thread->base_priority = THR_DEFAULT_PRIORITY;
diff --git a/lib/libpthread/thread/thr_kern.c b/lib/libpthread/thread/thr_kern.c
index a08c677..085c079 100644
--- a/lib/libpthread/thread/thr_kern.c
+++ b/lib/libpthread/thread/thr_kern.c
@@ -54,7 +54,6 @@ __FBSDID("$FreeBSD$");
 #include "atomic_ops.h"
 #include "thr_private.h"
 #include "libc_private.h"
-#include "ksd.h"
 
 /*#define DEBUG_THREAD_KERN */
 #ifdef DEBUG_THREAD_KERN
@@ -79,7 +78,7 @@ __FBSDID("$FreeBSD$");
 #define	KSE_STACKSIZE		16384
 
 #define	KSE_SET_MBOX(kse, thrd) \
-	(kse)->k_mbx.km_curthread = &(thrd)->tmbx
+	(kse)->k_kcb->kcb_kmbx.km_curthread = &(thrd)->tcb->tcb_tmbx
 
 #define	KSE_SET_EXITED(kse)	(kse)->k_flags |= KF_EXITED
 
@@ -132,8 +131,8 @@ static void	kse_check_completed(struct kse *kse);
 static void	kse_check_waitq(struct kse *kse);
 static void	kse_fini(struct kse *curkse);
 static void	kse_reinit(struct kse *kse, int sys_scope);
-static void	kse_sched_multi(struct kse *curkse);
-static void	kse_sched_single(struct kse *curkse);
+static void	kse_sched_multi(struct kse_mailbox *kmbx);
+static void	kse_sched_single(struct kse_mailbox *kmbx);
 static void	kse_switchout_thread(struct kse *kse, struct pthread *thread);
 static void	kse_wait(struct kse *kse, struct pthread *td_wait, int sigseq);
 static void	kse_free_unlocked(struct kse *kse);
@@ -152,20 +151,30 @@ static int	thr_timedout(struct pthread *thread, struct timespec *curtime);
 static void	thr_unlink(struct pthread *thread);
 
 
+static __inline void
+kse_set_curthread(struct kse *kse, struct pthread *td)
+{
+	kse->k_curthread = td;
+	if (td != NULL)
+		_tcb_set(kse->k_kcb, td->tcb);
+	else
+		_tcb_set(kse->k_kcb, NULL);
+}
+
 static void __inline
 thr_accounting(struct pthread *thread)
 {
 	if ((thread->slice_usec != -1) &&
 	    (thread->slice_usec <= TIMESLICE_USEC) &&
 	    (thread->attr.sched_policy != SCHED_FIFO)) {
-		thread->slice_usec += (thread->tmbx.tm_uticks
-		    + thread->tmbx.tm_sticks) * _clock_res_usec;
+		thread->slice_usec += (thread->tcb->tcb_tmbx.tm_uticks
+		    + thread->tcb->tcb_tmbx.tm_sticks) * _clock_res_usec;
 		/* Check for time quantum exceeded: */
 		if (thread->slice_usec > TIMESLICE_USEC)
 			thread->slice_usec = -1;
 	}
-	thread->tmbx.tm_uticks = 0;
-	thread->tmbx.tm_sticks = 0;
+	thread->tcb->tcb_tmbx.tm_uticks = 0;
+	thread->tcb->tcb_tmbx.tm_sticks = 0;
 }
 
 /*
@@ -246,7 +255,7 @@ _kse_single_thread(struct pthread *curthread)
 			_lockuser_destroy(&kse->k_lockusers[i]);
 		}
 		_lock_destroy(&kse->k_lock);
-		_ksd_destroy(&kse->k_ksd);
+		_kcb_dtor(kse->k_kcb);
 		if (kse->k_stack.ss_sp != NULL)
 			free(kse->k_stack.ss_sp);
 		free(kse);
@@ -341,7 +350,7 @@ _kse_single_thread(struct pthread *curthread)
 #else
 	if (__isthreaded)
 		_thr_signal_deinit();
-	_ksd_set_tmbx(NULL);
+	curthread->kse->k_kcb->kcb_kmbx.km_curthread = NULL;
 	__isthreaded   = 0;
 	active_threads = 0;
 #endif
@@ -409,11 +418,12 @@ _kse_setthreaded(int threaded)
 		 * For bound thread, kernel reads mailbox pointer once,
 		 * we'd set it here before calling kse_create
 		 */
+		_tcb_set(_kse_initial->k_kcb, _thr_initial->tcb);
 		KSE_SET_MBOX(_kse_initial, _thr_initial);
-		_kse_initial->k_mbx.km_flags |= KMF_BOUND;
+		_kse_initial->k_kcb->kcb_kmbx.km_flags |= KMF_BOUND;
 #endif
 
-		if (kse_create(&_kse_initial->k_mbx, 0) != 0) {
+		if (kse_create(&_kse_initial->k_kcb->kcb_kmbx, 0) != 0) {
 			_kse_initial->k_flags &= ~KF_STARTED;
 			__isthreaded = 0;
 			PANIC("kse_create() failed\n");
@@ -422,6 +432,7 @@ _kse_setthreaded(int threaded)
 
 #ifndef SYSTEM_SCOPE_ONLY
 		/* Set current thread to initial thread */
+		_tcb_set(_kse_initial->k_kcb, _thr_initial->tcb);
 		KSE_SET_MBOX(_kse_initial, _thr_initial);
 		_thr_start_sig_daemon();
 		_thr_setmaxconcurrency();
@@ -450,7 +461,7 @@ _kse_lock_wait(struct lock *lock, struct lockuser *lu)
 	struct timespec ts;
 	int saved_flags;
 
-	if (curkse->k_mbx.km_curthread != NULL)
+	if (curkse->k_kcb->kcb_kmbx.km_curthread != NULL)
 		PANIC("kse_lock_wait does not disable upcall.\n");
 	/*
 	 * Enter a loop to wait until we get the lock.
@@ -462,10 +473,11 @@ _kse_lock_wait(struct lock *lock, struct lockuser *lu)
 		 * Yield the kse and wait to be notified when the lock
 		 * is granted.
 		 */
-		saved_flags = curkse->k_mbx.km_flags;
-		curkse->k_mbx.km_flags |= KMF_NOUPCALL | KMF_NOCOMPLETED;
+		saved_flags = curkse->k_kcb->kcb_kmbx.km_flags;
+		curkse->k_kcb->kcb_kmbx.km_flags |= KMF_NOUPCALL |
+		    KMF_NOCOMPLETED;
 		kse_release(&ts);
-		curkse->k_mbx.km_flags = saved_flags;
+		curkse->k_kcb->kcb_kmbx.km_flags = saved_flags;
 	}
 }
 
@@ -482,7 +494,7 @@ _kse_lock_wakeup(struct lock *lock, struct lockuser *lu)
 	if (kse == curkse)
 		PANIC("KSE trying to wake itself up in lock");
 	else {
-		mbx = &kse->k_mbx;
+		mbx = &kse->k_kcb->kcb_kmbx;
 		_lock_grant(lock, lu);
 		/*
 		 * Notify the owning kse that it has the lock.
@@ -534,8 +546,7 @@ _kse_critical_enter(void)
 {
 	kse_critical_t crit;
 
-	crit = _ksd_get_tmbx();
-	_ksd_set_tmbx(NULL);
+	crit = (kse_critical_t)_kcb_critical_enter();
 	return (crit);
 }
 
@@ -544,7 +555,7 @@ _kse_critical_leave(kse_critical_t crit)
 {
 	struct pthread *curthread;
 
-	_ksd_set_tmbx(crit);
+	_kcb_critical_leave((struct kse_thr_mailbox *)crit);
 	if ((crit != NULL) && ((curthread = _get_curthread()) != NULL))
 		THR_YIELD_CHECK(curthread);
 }
@@ -552,7 +563,7 @@ _kse_critical_leave(kse_critical_t crit)
 int
 _kse_in_critical(void)
 {
-	return (_ksd_get_tmbx() == NULL);
+	return (_kcb_in_critical());
 }
 
 void
@@ -629,17 +640,17 @@ _thr_sched_switch_unlocked(struct pthread *curthread)
 	 *     we don't bother checking for that.
 	 */
 	if (curthread->attr.flags & PTHREAD_SCOPE_SYSTEM)
-		kse_sched_single(curkse);
+		kse_sched_single(&curkse->k_kcb->kcb_kmbx);
 	else if ((curthread->state == PS_DEAD) ||
 	    (((td = KSE_RUNQ_FIRST(curkse)) == NULL) &&
 	    (curthread->state != PS_RUNNING)) ||
 	    ((td != NULL) && (td->lock_switch == 0))) {
 		curkse->k_switch = 1;
-		_thread_enter_uts(&curthread->tmbx, &curkse->k_mbx);
+		_thread_enter_uts(curthread->tcb, curkse->k_kcb);
 	}
 	else {
 		uts_once = 0;
-		THR_GETCONTEXT(&curthread->tmbx.tm_context);
+		THR_GETCONTEXT(&curthread->tcb->tcb_tmbx.tm_context);
 		if (uts_once == 0) {
 			uts_once = 1;
 
@@ -649,7 +660,7 @@ _thr_sched_switch_unlocked(struct pthread *curthread)
 		 	/* Choose another thread to run. */
 			td = KSE_RUNQ_FIRST(curkse);
 			KSE_RUNQ_REMOVE(curkse, td);
-			curkse->k_curthread = td;
+			kse_set_curthread(curkse, td);
 
 			/*
 			 * Make sure the current thread's kse points to
@@ -674,7 +685,7 @@ _thr_sched_switch_unlocked(struct pthread *curthread)
 			/*
 			 * Continue the thread at its current frame:
 			 */
-			ret = _thread_switch(&td->tmbx, NULL);
+			ret = _thread_switch(curkse->k_kcb, td->tcb, 0);
 			/* This point should not be reached. */
 			if (ret != 0)
 				PANIC("Bad return from _thread_switch");
@@ -701,7 +712,7 @@ _thr_sched_switch_unlocked(struct pthread *curthread)
 
 		curthread->lock_switch = 0;
 		KSE_SCHED_UNLOCK(curkse, curkse->k_kseg);
-		_kse_critical_leave(&curthread->tmbx);
+		_kse_critical_leave(&curthread->tcb->tcb_tmbx);
 	}
 	/*
 	 * This thread is being resumed; check for cancellations.
@@ -728,17 +739,21 @@ _thr_sched_switch_unlocked(struct pthread *curthread)
  */
 
 static void
-kse_sched_single(struct kse *curkse)
+kse_sched_single(struct kse_mailbox *kmbx)
 {
-	struct pthread *curthread = curkse->k_curthread;
+	struct kse *curkse;
+	struct pthread *curthread;
 	struct timespec ts;
 	sigset_t sigmask;
 	int i, sigseqno, level, first = 0;
 
+	curkse = (struct kse *)kmbx->km_udata;
+	curthread = curkse->k_curthread;
+
 	if ((curkse->k_flags & KF_INITIALIZED) == 0) {
 		/* Setup this KSEs specific data. */
-		_ksd_setprivate(&curkse->k_ksd);
-		_set_curkse(curkse);
+		_kcb_set(curkse->k_kcb);
+		_tcb_set(curkse->k_kcb, curthread->tcb);
 		curkse->k_flags |= KF_INITIALIZED;
 		first = 1;
 		curthread->active = 1;
@@ -750,7 +765,7 @@ kse_sched_single(struct kse *curkse)
 		 * It is used to let other code work, those code want mailbox
 		 * to be cleared.
 		 */
-		_kse_critical_enter();
+		(void)_kse_critical_enter();
  	}
 
 	curthread->critical_yield = 0;
@@ -875,7 +890,7 @@ kse_sched_single(struct kse *curkse)
 
 	DBG_MSG("Continuing bound thread %p\n", curthread);
 	if (first) {
-		_kse_critical_leave(&curthread->tmbx);
+		_kse_critical_leave(&curthread->tcb->tcb_tmbx);
 		pthread_exit(curthread->start_routine(curthread->arg));
 	}
 }
@@ -898,20 +913,21 @@ dump_queues(struct kse *curkse)
  * This is the scheduler for a KSE which runs multiple threads.
  */
 static void
-kse_sched_multi(struct kse *curkse)
+kse_sched_multi(struct kse_mailbox *kmbx)
 {
+	struct kse *curkse;
 	struct pthread *curthread, *td_wait;
 	struct pthread_sigframe *curframe;
 	int ret;
 
-	THR_ASSERT(curkse->k_mbx.km_curthread == NULL,
+	curkse = (struct kse *)kmbx->km_udata;
+	THR_ASSERT(curkse->k_kcb->kcb_kmbx.km_curthread == NULL,
 	    "Mailbox not null in kse_sched_multi");
 
 	/* Check for first time initialization: */
 	if ((curkse->k_flags & KF_INITIALIZED) == 0) {
 		/* Setup this KSEs specific data. */
-		_ksd_setprivate(&curkse->k_ksd);
-		_set_curkse(curkse);
+		_kcb_set(curkse->k_kcb);
 
 		/* Set this before grabbing the context. */
 		curkse->k_flags |= KF_INITIALIZED;
@@ -928,6 +944,12 @@ kse_sched_multi(struct kse *curkse)
 		KSE_SCHED_LOCK(curkse, curkse->k_kseg);
 	curkse->k_switch = 0;
 
+	/*
+	 * Now that the scheduler lock is held, get the current
+	 * thread.  The KSE's current thread cannot be safely
+	 * examined without the lock because it could have returned
+	 * as completed on another KSE.  See kse_check_completed().
+	 */
 	curthread = curkse->k_curthread;
 
 	if (KSE_IS_IDLE(curkse)) {
@@ -975,20 +997,19 @@ kse_sched_multi(struct kse *curkse)
 		curthread->active = 1;
 		if ((curthread->flags & THR_FLAGS_IN_RUNQ) != 0)
 			KSE_RUNQ_REMOVE(curkse, curthread);
-		curkse->k_curthread = curthread;
+		kse_set_curthread(curkse, curthread);
 		curthread->kse = curkse;
 		DBG_MSG("Continuing thread %p in critical region\n",
 		    curthread);
 		kse_wakeup_multi(curkse);
 		KSE_SCHED_UNLOCK(curkse, curkse->k_kseg);
-		ret = _thread_switch(&curthread->tmbx,
-		    &curkse->k_mbx.km_curthread);
+		ret = _thread_switch(curkse->k_kcb, curthread->tcb, 1);
 		if (ret != 0)
 			PANIC("Can't resume thread in critical region\n");
 	}
 	else if ((curthread->flags & THR_FLAGS_IN_RUNQ) == 0)
 		kse_switchout_thread(curkse, curthread);
-	curkse->k_curthread = NULL;
+	kse_set_curthread(curkse, NULL);
 
 	kse_wakeup_multi(curkse);
 
@@ -1034,7 +1055,7 @@ kse_sched_multi(struct kse *curkse)
 	/*
 	 * Make the selected thread the current thread.
 	 */
-	curkse->k_curthread = curthread;
+	kse_set_curthread(curkse, curthread);
 
 	/*
 	 * Make sure the current thread's kse points to this kse.
@@ -1069,13 +1090,13 @@ kse_sched_multi(struct kse *curkse)
 	    (((curthread->cancelflags & THR_AT_CANCEL_POINT) == 0) &&
 	     ((curthread->cancelflags & PTHREAD_CANCEL_ASYNCHRONOUS) != 0))) &&
 	     !THR_IN_CRITICAL(curthread))
-		signalcontext(&curthread->tmbx.tm_context, 0,
+		signalcontext(&curthread->tcb->tcb_tmbx.tm_context, 0,
 		    (__sighandler_t *)thr_resume_wrapper);
 #else
 	if ((curframe == NULL) && (curthread->state == PS_RUNNING) &&
 	    (curthread->check_pending != 0) && !THR_IN_CRITICAL(curthread)) {
 		curthread->check_pending = 0;
-		signalcontext(&curthread->tmbx.tm_context, 0,
+		signalcontext(&curthread->tcb->tcb_tmbx.tm_context, 0,
 		    (__sighandler_t *)thr_resume_wrapper);
 	}
 #endif
@@ -1087,12 +1108,11 @@ kse_sched_multi(struct kse *curkse)
 		 * This thread came from a scheduler switch; it will
 		 * unlock the scheduler lock and set the mailbox.
 		 */
-		ret = _thread_switch(&curthread->tmbx, NULL);
+		ret = _thread_switch(curkse->k_kcb, curthread->tcb, 0);
 	} else {
 		/* This thread won't unlock the scheduler lock. */
 		KSE_SCHED_UNLOCK(curkse, curkse->k_kseg);
-		ret = _thread_switch(&curthread->tmbx,
-		    &curkse->k_mbx.km_curthread);
+		ret = _thread_switch(curkse->k_kcb, curthread->tcb, 1);
 	}
 	if (ret != 0)
 		PANIC("Thread has returned from _thread_switch");
@@ -1114,9 +1134,9 @@ thr_resume_wrapper(int sig, siginfo_t *siginfo, ucontext_t *ucp)
 	thr_resume_check(curthread, ucp, NULL);
 	_kse_critical_enter();
 	curkse = _get_curkse();
-	curthread->tmbx.tm_context = *ucp;
+	curthread->tcb->tcb_tmbx.tm_context = *ucp;
 	curthread->error = err_save;
-	ret = _thread_switch(&curthread->tmbx, &curkse->k_mbx.km_curthread);
+	ret = _thread_switch(curkse->k_kcb, curthread->tcb, 1);
 	if (ret != 0)
 		PANIC("thr_resume_wrapper: thread has returned "
 		      "from _thread_switch");
@@ -1242,7 +1262,7 @@ _thr_gc(struct pthread *curthread)
 		if ((td->flags & THR_FLAGS_GC_SAFE) == 0)
 			continue;
 		else if (((td->attr.flags & PTHREAD_SCOPE_SYSTEM) != 0) &&
-		    ((td->kse->k_mbx.km_flags & KMF_DONE) == 0)) {
+		    ((td->kse->k_kcb->kcb_kmbx.km_flags & KMF_DONE) == 0)) {
 			/*
 			 * The thread and KSE are operating on the same
 			 * stack.  Wait for the KSE to exit before freeing
@@ -1319,9 +1339,9 @@ _thr_schedule_add(struct pthread *curthread, struct pthread *newthread)
 	 */
 	if ((newthread->attr.flags & PTHREAD_SCOPE_SYSTEM) != 0) {
 		/* We use the thread's stack as the KSE's stack. */
-		newthread->kse->k_mbx.km_stack.ss_sp =
+		newthread->kse->k_kcb->kcb_kmbx.km_stack.ss_sp =
 		    newthread->attr.stackaddr_attr;
-		newthread->kse->k_mbx.km_stack.ss_size =
+		newthread->kse->k_kcb->kcb_kmbx.km_stack.ss_size =
 		    newthread->attr.stacksize_attr;
 
 		/*
@@ -1331,10 +1351,10 @@ _thr_schedule_add(struct pthread *curthread, struct pthread *newthread)
 		KSEG_THRQ_ADD(newthread->kseg, newthread);
 		/* this thread never gives up kse */
 		newthread->active = 1;
-		newthread->kse->k_curthread = newthread;
-		newthread->kse->k_mbx.km_flags = KMF_BOUND;
-		newthread->kse->k_mbx.km_func = (kse_func_t *)kse_sched_single;
-		newthread->kse->k_mbx.km_quantum = 0;
+		kse_set_curthread(newthread->kse, newthread);
+		newthread->kse->k_kcb->kcb_kmbx.km_flags = KMF_BOUND;
+		newthread->kse->k_kcb->kcb_kmbx.km_func = (kse_func_t *)kse_sched_single;
+		newthread->kse->k_kcb->kcb_kmbx.km_quantum = 0;
 		KSE_SET_MBOX(newthread->kse, newthread);
 		/*
 		 * This thread needs a new KSE and KSEG.
@@ -1342,7 +1362,7 @@ _thr_schedule_add(struct pthread *curthread, struct pthread *newthread)
 		newthread->kse->k_flags &= ~KF_INITIALIZED;
 		newthread->kse->k_flags |= KF_STARTED;
 		/* Fire up! */
-		ret = kse_create(&newthread->kse->k_mbx, 1);
+		ret = kse_create(&newthread->kse->k_kcb->kcb_kmbx, 1);
 		if (ret != 0)
 			ret = errno;
 	}
@@ -1363,10 +1383,10 @@ _thr_schedule_add(struct pthread *curthread, struct pthread *newthread)
 			 * outside of holding the lock.
 			 */
 			newthread->kse->k_flags |= KF_STARTED;
-			newthread->kse->k_mbx.km_func =
+			newthread->kse->k_kcb->kcb_kmbx.km_func =
 			    (kse_func_t *)kse_sched_multi;
-			newthread->kse->k_mbx.km_flags = 0;
-			kse_create(&newthread->kse->k_mbx, 0);
+			newthread->kse->k_kcb->kcb_kmbx.km_flags = 0;
+			kse_create(&newthread->kse->k_kcb->kcb_kmbx, 0);
 		 } else if ((newthread->state == PS_RUNNING) &&
 		     KSE_IS_IDLE(newthread->kse)) {
 			/*
@@ -1418,8 +1438,8 @@ kse_check_completed(struct kse *kse)
 	struct kse_thr_mailbox *completed;
 	int sig;
 
-	if ((completed = kse->k_mbx.km_completed) != NULL) {
-		kse->k_mbx.km_completed = NULL;
+	if ((completed = kse->k_kcb->kcb_kmbx.km_completed) != NULL) {
+		kse->k_kcb->kcb_kmbx.km_completed = NULL;
 		while (completed != NULL) {
 			thread = completed->tm_udata;
 			DBG_MSG("Found completed thread %p, name %s\n",
@@ -1434,17 +1454,23 @@ kse_check_completed(struct kse *kse)
 					KSE_RUNQ_INSERT_TAIL(kse, thread);
 				if ((thread->kse != kse) &&
 				    (thread->kse->k_curthread == thread)) {
-					thread->kse->k_curthread = NULL;
+					/*
+					 * Remove this thread from its
+					 * previous KSE so that it (the KSE)
+					 * doesn't think it is still active.
+					 */
+					kse_set_curthread(thread->kse, NULL);
 					thread->active = 0;
 				}
 			}
-			if ((sig = thread->tmbx.tm_syncsig.si_signo) != 0) {
+			if ((sig = thread->tcb->tcb_tmbx.tm_syncsig.si_signo)
+			    != 0) {
 				if (SIGISMEMBER(thread->sigmask, sig))
 					SIGADDSET(thread->sigpend, sig);
 				else
 					(void)_thr_sig_add(thread, sig,
-					    &thread->tmbx.tm_syncsig);
-				thread->tmbx.tm_syncsig.si_signo = 0;
+					    &thread->tcb->tcb_tmbx.tm_syncsig);
+				thread->tcb->tcb_tmbx.tm_syncsig.si_signo = 0;
 			}
 			completed = completed->tm_next;
 		}
@@ -1567,7 +1593,7 @@ kse_switchout_thread(struct kse *kse, struct pthread *thread)
 				if (SIGISMEMBER(thread->sigpend, i) &&
 				    !SIGISMEMBER(thread->sigmask, i)) {
 					restart = _thread_sigact[1 - 1].sa_flags & SA_RESTART;
-					kse_thr_interrupt(&thread->tmbx,
+					kse_thr_interrupt(&thread->tcb->tcb_tmbx,
 					    restart ? KSE_INTR_RESTART : KSE_INTR_INTERRUPT, 0);
 					break;
 				}
@@ -1584,6 +1610,7 @@ kse_switchout_thread(struct kse *kse, struct pthread *thread)
 			 */
 			thread->active = 0;
 			thread->need_switchout = 0;
+			thread->lock_switch = 0;
 			thr_cleanup(kse, thread);
 			return;
 			break;
@@ -1705,10 +1732,10 @@ kse_wait(struct kse *kse, struct pthread *td_wait, int sigseqno)
 		    (kse->k_sigseqno != sigseqno))
 			; /* don't sleep */
 		else {
-			saved_flags = kse->k_mbx.km_flags;
-			kse->k_mbx.km_flags |= KMF_NOUPCALL;
+			saved_flags = kse->k_kcb->kcb_kmbx.km_flags;
+			kse->k_kcb->kcb_kmbx.km_flags |= KMF_NOUPCALL;
 			kse_release(&ts_sleep);
-			kse->k_mbx.km_flags = saved_flags;
+			kse->k_kcb->kcb_kmbx.km_flags = saved_flags;
 		}
 		KSE_SCHED_LOCK(kse, kse->k_kseg);
 		if (KSE_IS_IDLE(kse)) {
@@ -1781,7 +1808,7 @@ kse_fini(struct kse *kse)
 		KSE_SCHED_UNLOCK(kse, kse->k_kseg);
 		ts.tv_sec = 120;
 		ts.tv_nsec = 0;
-		kse->k_mbx.km_flags = 0;
+		kse->k_kcb->kcb_kmbx.km_flags = 0;
 		kse_release(&ts);
 		/* Never reach */
 	}
@@ -1898,13 +1925,13 @@ kse_wakeup_one(struct pthread *thread)
 	if (KSE_IS_IDLE(thread->kse)) {
 		KSE_CLEAR_IDLE(thread->kse);
 		thread->kseg->kg_idle_kses--;
-		return (&thread->kse->k_mbx);
+		return (&thread->kse->k_kcb->kcb_kmbx);
 	} else {
 		TAILQ_FOREACH(ke, &thread->kseg->kg_kseq, k_kgqe) {
 			if (KSE_IS_IDLE(ke)) {
 				KSE_CLEAR_IDLE(ke);
 				ke->k_kseg->kg_idle_kses--;
-				return (&ke->k_mbx);
+				return (&ke->k_kcb->kcb_kmbx);
 			}
 		}
 	}
@@ -1930,25 +1957,6 @@ kse_wakeup_multi(struct kse *curkse)
 	}
 }
 
-struct pthread *
-_get_curthread(void)
-{
-	return (_ksd_curthread());
-}
-
-/* This assumes the caller has disabled upcalls. */
-struct kse *
-_get_curkse(void)
-{
-	return (_ksd_curkse());
-}
-
-void
-_set_curkse(struct kse *kse)
-{
-	_ksd_setprivate(&kse->k_ksd);
-}
-
 /*
  * Allocate a new KSEG.
  *
@@ -2048,8 +2056,8 @@ struct kse *
 _kse_alloc(struct pthread *curthread, int sys_scope)
 {
 	struct kse *kse = NULL;
+	char *stack;
 	kse_critical_t crit;
-	int need_ksd = 0;
 	int i;
 
 	if ((curthread != NULL) && (free_kse_count > 0)) {
@@ -2058,7 +2066,7 @@ _kse_alloc(struct pthread *curthread, int sys_scope)
 		/* Search for a finished KSE. */
 		kse = TAILQ_FIRST(&free_kseq);
 		while ((kse != NULL) &&
-		    ((kse->k_mbx.km_flags & KMF_DONE) == 0)) {
+		    ((kse->k_kcb->kcb_kmbx.km_flags & KMF_DONE) == 0)) {
 			kse = TAILQ_NEXT(kse, k_qe);
 		}
 		if (kse != NULL) {
@@ -2075,8 +2083,22 @@ _kse_alloc(struct pthread *curthread, int sys_scope)
 	}
 	if ((kse == NULL) &&
 	    ((kse = (struct kse *)malloc(sizeof(*kse))) != NULL)) {
+		if (sys_scope != 0)
+			stack = NULL;
+		else if ((stack = malloc(KSE_STACKSIZE)) == NULL) {
+			free(kse);
+			return (NULL);
+		}
 		bzero(kse, sizeof(*kse));
 
+		/* Initialize KCB without the lock. */
+		if ((kse->k_kcb = _kcb_ctor(kse)) == NULL) {
+			if (stack != NULL)
+				free(stack);
+			free(kse);
+			return (NULL);
+		}
+
 		/* Initialize the lockusers. */
 		for (i = 0; i < MAX_KSE_LOCKLEVEL; i++) {
 			_lockuser_init(&kse->k_lockusers[i], (void *)kse);
@@ -2084,65 +2106,39 @@ _kse_alloc(struct pthread *curthread, int sys_scope)
 		}
 		/* _lock_init(kse->k_lock, ...) */
 
-		/* We had to malloc a kse; mark it as needing a new ID.*/
-		need_ksd = 1;
-
+		if (curthread != NULL) {
+			crit = _kse_critical_enter();
+			KSE_LOCK_ACQUIRE(curthread->kse, &kse_lock);
+		}
+		kse->k_flags = 0;
+		TAILQ_INSERT_TAIL(&active_kseq, kse, k_qe);
+		active_kse_count++;
+		if (curthread != NULL) {
+			KSE_LOCK_RELEASE(curthread->kse, &kse_lock);
+			_kse_critical_leave(crit);
+		}
 		/*
 		 * Create the KSE context.
 		 * Scope system threads (one thread per KSE) are not required
 		 * to have a stack for an unneeded kse upcall.
 		 */
 		if (!sys_scope) {
-			kse->k_mbx.km_func = (kse_func_t *)kse_sched_multi;
-			kse->k_stack.ss_sp = (char *) malloc(KSE_STACKSIZE);
+			kse->k_kcb->kcb_kmbx.km_func = (kse_func_t *)kse_sched_multi;
+			kse->k_stack.ss_sp = stack;
 			kse->k_stack.ss_size = KSE_STACKSIZE;
 		} else {
-			kse->k_mbx.km_func = (kse_func_t *)kse_sched_single;
+			kse->k_kcb->kcb_kmbx.km_func = (kse_func_t *)kse_sched_single;
+			kse->k_stack.ss_sp = NULL;
+			kse->k_stack.ss_size = 0;
 		}
-		kse->k_mbx.km_udata = (void *)kse;
-		kse->k_mbx.km_quantum = 20000;
+		kse->k_kcb->kcb_kmbx.km_udata = (void *)kse;
+		kse->k_kcb->kcb_kmbx.km_quantum = 20000;
 		/*
 		 * We need to keep a copy of the stack in case it
 		 * doesn't get used; a KSE running a scope system
 		 * thread will use that thread's stack.
 		 */
-		kse->k_mbx.km_stack = kse->k_stack;
-		if (!sys_scope && kse->k_stack.ss_sp == NULL) {
-			for (i = 0; i < MAX_KSE_LOCKLEVEL; i++) {
-				_lockuser_destroy(&kse->k_lockusers[i]);
-			}
-			/* _lock_destroy(&kse->k_lock); */
-			free(kse);
-			kse = NULL;
-		}
-	}
-	if ((kse != NULL) && (need_ksd != 0)) {
-		/* This KSE needs initialization. */
-		if (curthread != NULL) {
-			crit = _kse_critical_enter();
-			KSE_LOCK_ACQUIRE(curthread->kse, &kse_lock);
-		}
-		/* Initialize KSD inside of the lock. */
-		if (_ksd_create(&kse->k_ksd, (void *)kse, sizeof(*kse)) != 0) {
-			if (curthread != NULL) {
-				KSE_LOCK_RELEASE(curthread->kse, &kse_lock);
-				_kse_critical_leave(crit);
-			}
-			if (kse->k_stack.ss_sp)
-				free(kse->k_stack.ss_sp);
-			for (i = 0; i < MAX_KSE_LOCKLEVEL; i++) {
-				_lockuser_destroy(&kse->k_lockusers[i]);
-			}
-			free(kse);
-			return (NULL);
-		}
-		kse->k_flags = 0;
-		TAILQ_INSERT_TAIL(&active_kseq, kse, k_qe);
-		active_kse_count++;
-		if (curthread != NULL) {
-			KSE_LOCK_RELEASE(curthread->kse, &kse_lock);
-			_kse_critical_leave(crit);
-		}
+		kse->k_kcb->kcb_kmbx.km_stack = kse->k_stack;
 	}
 	return (kse);
 }
@@ -2151,26 +2147,26 @@ static void
 kse_reinit(struct kse *kse, int sys_scope)
 {
 	if (!sys_scope) {
-		kse->k_mbx.km_func = (kse_func_t *)kse_sched_multi;
+		kse->k_kcb->kcb_kmbx.km_func = (kse_func_t *)kse_sched_multi;
 		if (kse->k_stack.ss_sp == NULL) {
 			/* XXX check allocation failure */
 			kse->k_stack.ss_sp = (char *) malloc(KSE_STACKSIZE);
 			kse->k_stack.ss_size = KSE_STACKSIZE;
 		}
-		kse->k_mbx.km_quantum = 20000;
+		kse->k_kcb->kcb_kmbx.km_quantum = 20000;
 	} else {
-		kse->k_mbx.km_func = (kse_func_t *)kse_sched_single;
+		kse->k_kcb->kcb_kmbx.km_func = (kse_func_t *)kse_sched_single;
 		if (kse->k_stack.ss_sp)
 			free(kse->k_stack.ss_sp);
 		kse->k_stack.ss_sp = NULL;
 		kse->k_stack.ss_size = 0;
-		kse->k_mbx.km_quantum = 0;
+		kse->k_kcb->kcb_kmbx.km_quantum = 0;
 	}
-	kse->k_mbx.km_stack = kse->k_stack;
-	kse->k_mbx.km_udata = (void *)kse;
-	kse->k_mbx.km_curthread = NULL;
-	kse->k_mbx.km_flags = 0;
-	kse->k_curthread = 0;
+	kse->k_kcb->kcb_kmbx.km_stack = kse->k_stack;
+	kse->k_kcb->kcb_kmbx.km_udata = (void *)kse;
+	kse->k_kcb->kcb_kmbx.km_curthread = NULL;
+	kse->k_kcb->kcb_kmbx.km_flags = 0;
+	kse->k_curthread = NULL;
 	kse->k_kseg = 0;
 	kse->k_schedq = 0;
 	kse->k_locklevel = 0;
@@ -2193,9 +2189,10 @@ kse_free_unlocked(struct kse *kse)
 	TAILQ_REMOVE(&active_kseq, kse, k_qe);
 	active_kse_count--;
 	kse->k_kseg = NULL;
-	kse->k_mbx.km_quantum = 20000;
+	kse->k_kcb->kcb_kmbx.km_quantum = 20000;
 	kse->k_flags = 0;
 	TAILQ_INSERT_HEAD(&free_kseq, kse, k_qe);
+	_kcb_dtor(kse->k_kcb);
 	free_kse_count++;
 }
 
@@ -2239,7 +2236,6 @@ struct pthread *
 _thr_alloc(struct pthread *curthread)
 {
 	kse_critical_t crit;
-	void *p;
 	struct pthread *thread = NULL;
 
 	if (curthread != NULL) {
@@ -2256,11 +2252,12 @@ _thr_alloc(struct pthread *curthread)
 			_kse_critical_leave(crit);
 		}
 	}
-	if (thread == NULL) {
-		p = malloc(sizeof(struct pthread) + THR_ALIGNBYTES);
-		if (p != NULL) {
-			thread = (struct pthread *)THR_ALIGN(p);
-			thread->alloc_addr = p;
+	if ((thread == NULL) &&
+	    ((thread = malloc(sizeof(struct pthread))) != NULL)) {
+		bzero(thread, sizeof(struct pthread));
+		if ((thread->tcb = _tcb_ctor(thread)) == NULL) {
+			free(thread);
+			thread = NULL;
 		}
 	}
 	return (thread);
@@ -2278,9 +2275,16 @@ _thr_free(struct pthread *curthread, struct pthread *thread)
 			_lockuser_destroy(&thread->lockusers[i]);
 		}
 		_lock_destroy(&thread->lock);
-		free(thread->alloc_addr);
+		_tcb_dtor(thread->tcb);
+		free(thread);
 	}
 	else {
+		/* Reinitialize any important fields here. */
+		thread->lock_switch = 0;
+		sigemptyset(&thread->sigpend);
+		thread->check_pending = 0;
+
+		/* Add the thread to the free thread list. */
 		crit = _kse_critical_enter();
 		KSE_LOCK_ACQUIRE(curthread->kse, &thread_lock);
 		TAILQ_INSERT_TAIL(&free_threadq, thread, tle);
diff --git a/lib/libpthread/thread/thr_private.h b/lib/libpthread/thread/thr_private.h
index 25a47a7..0664aca 100644
--- a/lib/libpthread/thread/thr_private.h
+++ b/lib/libpthread/thread/thr_private.h
@@ -173,15 +173,14 @@ struct kse_group;
 
 #define	MAX_KSE_LOCKLEVEL	5	
 struct kse {
-	struct kse_mailbox	k_mbx;		/* kernel kse mailbox */
 	/* -- location and order specific items for gdb -- */
+	struct kcb		*k_kcb;
 	struct pthread		*k_curthread;	/* current thread */
 	struct kse_group	*k_kseg;	/* parent KSEG */
 	struct sched_queue	*k_schedq;	/* scheduling queue */
 	/* -- end of location and order specific items -- */
 	TAILQ_ENTRY(kse)	k_qe;		/* KSE list link entry */
 	TAILQ_ENTRY(kse)	k_kgqe;		/* KSEG's KSE list entry */
-	struct ksd		k_ksd;		/* KSE specific data */
 	/*
 	 * Items that are only modified by the kse, or that otherwise
 	 * don't need to be locked when accessed
@@ -300,7 +299,7 @@ do { \
 #define	KSE_CLEAR_WAIT(kse) 	atomic_store_rel_int(&(kse)->k_waiting, 0)
 
 #define	KSE_WAITING(kse)	(kse)->k_waiting != 0
-#define	KSE_WAKEUP(kse)		kse_wakeup(&(kse)->k_mbx)
+#define	KSE_WAKEUP(kse)		kse_wakeup(&(kse)->k_kcb->kcb_kmbx)
 
 #define	KSE_SET_IDLE(kse)	((kse)->k_idle = 1)
 #define	KSE_CLEAR_IDLE(kse)	((kse)->k_idle = 0)
@@ -509,7 +508,7 @@ struct pthread_attr {
  */
 #define	KSE_GET_TOD(curkse, tsp) \
 do {							\
-	*tsp = (curkse)->k_mbx.km_timeofday;		\
+	*tsp = (curkse)->k_kcb->kcb_kmbx.km_timeofday;	\
 	if ((tsp)->tv_sec == 0)				\
 		clock_gettime(CLOCK_REALTIME, tsp);	\
 } while (0)
@@ -601,8 +600,7 @@ struct pthread {
 	/*
 	 * Thread mailbox is first so it cal be aligned properly.
 	 */
-	struct kse_thr_mailbox	tmbx;
-	void			*alloc_addr;	/* real address (unaligned) */
+	struct tcb		*tcb;
 
 	/*
 	 * Magic value to help recognize a valid thread structure
@@ -1049,9 +1047,6 @@ SCLASS int		_thr_debug_flags	SCLASS_PRESET(0);
 __BEGIN_DECLS
 int	_cond_reinit(pthread_cond_t *);
 void	_cond_wait_backout(struct pthread *);
-struct pthread *_get_curthread(void);
-struct kse *_get_curkse(void);
-void	_set_curkse(struct kse *);
 struct kse *_kse_alloc(struct pthread *, int sys_scope);
 kse_critical_t _kse_critical_enter(void);
 void	_kse_critical_leave(kse_critical_t);
@@ -1098,8 +1093,6 @@ int	_pthread_rwlock_destroy (pthread_rwlock_t *);
 struct pthread *_pthread_self(void);
 int	_pthread_setspecific(pthread_key_t, const void *);
 struct pthread *_thr_alloc(struct pthread *);
-int	_thread_enter_uts(struct kse_thr_mailbox *, struct kse_mailbox *);
-int	_thread_switch(struct kse_thr_mailbox *, struct kse_thr_mailbox **);
 void	_thr_exit(char *, int, char *);
 void	_thr_exit_cleanup(void);
 void	_thr_lock_wait(struct lock *lock, struct lockuser *lu);
diff --git a/lib/libpthread/thread/thr_sig.c b/lib/libpthread/thread/thr_sig.c
index 6f16318..77654de 100644
--- a/lib/libpthread/thread/thr_sig.c
+++ b/lib/libpthread/thread/thr_sig.c
@@ -41,7 +41,6 @@
 #include <string.h>
 #include <pthread.h>
 #include "thr_private.h"
-#include "pthread_md.h"
 
 /* Prototypes: */
 static void	build_siginfo(siginfo_t *info, int signo);
@@ -212,11 +211,11 @@ sig_daemon(void *arg /* Unused */)
 		}
 		ts.tv_sec = 30;
 		ts.tv_nsec = 0;
-		curkse->k_mbx.km_flags =
+		curkse->k_kcb->kcb_kmbx.km_flags =
 		    KMF_NOUPCALL | KMF_NOCOMPLETED | KMF_WAITSIGEVENT;
 		kse_release(&ts);
-		curkse->k_mbx.km_flags = 0;
-		set = curkse->k_mbx.km_sigscaught;
+		curkse->k_kcb->kcb_kmbx.km_flags = 0;
+		set = curkse->k_kcb->kcb_kmbx.km_sigscaught;
 	}
 	return (0);
 }
@@ -355,7 +354,7 @@ _thr_sig_handler(int sig, siginfo_t *info, ucontext_t *ucp)
 		 * so kse_release will return from kernel immediately.
 		 */
 		if (KSE_IS_IDLE(curkse))
-			kse_wakeup(&curkse->k_mbx);
+			kse_wakeup(&curkse->k_kcb->kcb_kmbx);
 		return;
 	}
 
@@ -377,7 +376,7 @@ _thr_sig_handler(int sig, siginfo_t *info, ucontext_t *ucp)
 		__sys_sigaction(sig, NULL, &_thread_sigact[sig - 1]);
 	}
 	KSE_LOCK_RELEASE(curkse, &_thread_signal_lock);
-	_kse_critical_leave(&curthread->tmbx);
+	_kse_critical_leave(&curthread->tcb->tcb_tmbx);
 
 	/* Now invoke real handler */
 	if (((__sighandler_t *)sigfunc != SIG_DFL) &&
@@ -403,7 +402,7 @@ _thr_sig_handler(int sig, siginfo_t *info, ucontext_t *ucp)
 	curthread->interrupted = intr_save;
 	_kse_critical_enter();
 	curthread->sigmask = ucp->uc_sigmask;
-	_kse_critical_leave(&curthread->tmbx);
+	_kse_critical_leave(&curthread->tcb->tcb_tmbx);
 	DBG_MSG("<<< _thr_sig_handler(%d)\n", sig);
 }
 
@@ -446,7 +445,7 @@ thr_sig_invoke_handler(struct pthread *curthread, int sig, siginfo_t *info,
 	}
 	KSE_LOCK_RELEASE(curkse, &_thread_signal_lock);
 	KSE_SCHED_UNLOCK(curkse, curkse->k_kseg);
-	_kse_critical_leave(&curthread->tmbx);
+	_kse_critical_leave(&curthread->tcb->tcb_tmbx);
 	/*
 	 * We are processing buffered signals, synchronize working
 	 * signal mask into kernel.
@@ -737,7 +736,7 @@ _thr_sig_rundown(struct pthread *curthread, ucontext_t *ucp,
 	KSE_SCHED_UNLOCK(curkse, curkse->k_kseg);
 	if (curthread->attr.flags & PTHREAD_SCOPE_SYSTEM)
 		__sys_sigprocmask(SIG_SETMASK, &curthread->sigmask, NULL);
-	_kse_critical_leave(&curthread->tmbx);
+	_kse_critical_leave(&curthread->tcb->tcb_tmbx);
 
 	curthread->interrupted = interrupted;
 	curthread->timeout = timeout;
@@ -860,7 +859,7 @@ _thr_sig_add(struct pthread *pthread, int sig, siginfo_t *info)
 			if (!(pthread->attr.flags & PTHREAD_SCOPE_SYSTEM) &&
 			    (pthread->blocked != 0) && 
 			    !THR_IN_CRITICAL(pthread))
-				kse_thr_interrupt(&pthread->tmbx,
+				kse_thr_interrupt(&pthread->tcb->tcb_tmbx,
 				    restart ? KSE_INTR_RESTART : KSE_INTR_INTERRUPT, 0);
 		}
 	}
@@ -983,7 +982,7 @@ _thr_sig_send(struct pthread *pthread, int sig)
 	struct kse_mailbox *kmbx;
 
 	if (pthread->attr.flags & PTHREAD_SCOPE_SYSTEM) {
-		kse_thr_interrupt(&pthread->tmbx, KSE_INTR_SENDSIG, sig);
+		kse_thr_interrupt(&pthread->tcb->tcb_tmbx, KSE_INTR_SENDSIG, sig);
 		return;
 	}