20060303a

143 files changed, 13871 insertions, 0 deletions

diff --git a/libkern/NOTICE b/libkern/NOTICE
new file mode 100644
index 00000000..d2099067
--- /dev/null
+++ b/libkern/NOTICE
@@ -0,0 +1,29 @@
+This copyright NOTICE applies to all files in this directory and
+subdirectories, unless another copyright notice appears in a given
+file or subdirectory.  If you take substantial code from this software to use in
+other programs, you must somehow include with it an appropriate
+copyright notice that includes the copyright notice and the other
+notices below.  It is fine (and often tidier) to do that in a separate
+file such as NOTICE, LICENCE or COPYING.
+
+	Copyright © 1994-1999 Lucent Technologies Inc.  All rights reserved.
+	Revisions Copyright © 2000-2006 Vita Nuova Holdings Limited (www.vitanuova.com).  All rights reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.
+
diff --git a/libkern/abort.c b/libkern/abort.c
new file mode 100644
index 00000000..f71c01c3
--- /dev/null
+++ b/libkern/abort.c
@@ -0,0 +1,7 @@
+#include <lib9.h>
+void
+abort(void)
+{
+	while(*(int*)0)
+		;
+}
diff --git a/libkern/abs.c b/libkern/abs.c
new file mode 100644
index 00000000..6bb51107
--- /dev/null
+++ b/libkern/abs.c
@@ -0,0 +1,17 @@
+#include <lib9.h>
+
+int
+abs(int a)
+{
+	if(a < 0)
+		return -a;
+	return a;
+}
+
+long
+labs(long a)
+{
+	if(a < 0)
+		return -a;
+	return a;
+}
diff --git a/libkern/atol.c b/libkern/atol.c
new file mode 100644
index 00000000..18a34b77
--- /dev/null
+++ b/libkern/atol.c
@@ -0,0 +1,47 @@
+#include <lib9.h>
+
+long
+atol(char *s)
+{
+	long n;
+	int f;
+
+	n = 0;
+	f = 0;
+	while(*s == ' ' || *s == '\t')
+		s++;
+	if(*s == '-' || *s == '+') {
+		if(*s++ == '-')
+			f = 1;
+		while(*s == ' ' || *s == '\t')
+			s++;
+	}
+	if(s[0]=='0' && s[1]){
+		if(s[1]=='x' || s[1]=='X'){
+			s += 2;
+			for(;;){
+				if(*s >= '0' && *s <= '9')
+					n = n*16 + *s++ - '0';
+				else if(*s >= 'a' && *s <= 'f')
+					n = n*16 + *s++ - 'a' + 10;
+				else if(*s >= 'A' && *s <= 'F')
+					n = n*16 + *s++ - 'A' + 10;
+				else
+					break;
+			}
+		} else
+			while(*s >= '0' && *s <= '7')
+				n = n*8 + *s++ - '0';
+	} else
+		while(*s >= '0' && *s <= '9')
+			n = n*10 + *s++ - '0';
+	if(f)
+		n = -n;
+	return n;
+}
+
+atoi(char *s)
+{
+
+	return atol(s);
+}
diff --git a/libkern/charstod.c b/libkern/charstod.c
new file mode 100644
index 00000000..203e5379
--- /dev/null
+++ b/libkern/charstod.c
@@ -0,0 +1,68 @@
+#include <lib9.h>
+
+/*
+ * Reads a floating-point number by interpreting successive characters
+ * returned by (*f)(vp).  The last call it makes to f terminates the
+ * scan, so is not a character in the number.  It may therefore be
+ * necessary to back up the input stream up one byte after calling charstod.
+ */
+
+double
+charstod(int(*f)(void*), void *vp)
+{
+	double num, dem;
+	int neg, eneg, dig, exp, c;
+
+	num = 0;
+	neg = 0;
+	dig = 0;
+	exp = 0;
+	eneg = 0;
+
+	c = (*f)(vp);
+	while(c == ' ' || c == '\t')
+		c = (*f)(vp);
+	if(c == '-' || c == '+'){
+		if(c == '-')
+			neg = 1;
+		c = (*f)(vp);
+	}
+	while(c >= '0' && c <= '9'){
+		num = num*10 + c-'0';
+		c = (*f)(vp);
+	}
+	if(c == '.')
+		c = (*f)(vp);
+	while(c >= '0' && c <= '9'){
+		num = num*10 + c-'0';
+		dig++;
+		c = (*f)(vp);
+	}
+	if(c == 'e' || c == 'E'){
+		c = (*f)(vp);
+		if(c == '-' || c == '+'){
+			if(c == '-'){
+				dig = -dig;
+				eneg = 1;
+			}
+			c = (*f)(vp);
+		}
+		while(c >= '0' && c <= '9'){
+			exp = exp*10 + c-'0';
+			c = (*f)(vp);
+		}
+	}
+	exp -= dig;
+	if(exp < 0){
+		exp = -exp;
+		eneg = !eneg;
+	}
+	dem = pow10(exp);
+	if(eneg)
+		num /= dem;
+	else
+		num *= dem;
+	if(neg)
+		return -num;
+	return num;
+}
diff --git a/libkern/cistrcmp.c b/libkern/cistrcmp.c
new file mode 100644
index 00000000..fc7d84e4
--- /dev/null
+++ b/libkern/cistrcmp.c
@@ -0,0 +1,25 @@
+#include "lib9.h"
+
+int
+cistrcmp(char *s1, char *s2)
+{
+	int c1, c2;
+
+	while(*s1){
+		c1 = *(uchar*)s1++;
+		c2 = *(uchar*)s2++;
+
+		if(c1 == c2)
+			continue;
+
+		if(c1 >= 'A' && c1 <= 'Z')
+			c1 -= 'A' - 'a';
+
+		if(c2 >= 'A' && c2 <= 'Z')
+			c2 -= 'A' - 'a';
+
+		if(c1 != c2)
+			return c1 - c2;
+	}
+	return -*s2;
+}
diff --git a/libkern/cistrncmp.c b/libkern/cistrncmp.c
new file mode 100644
index 00000000..1d55d71b
--- /dev/null
+++ b/libkern/cistrncmp.c
@@ -0,0 +1,27 @@
+#include "lib9.h"
+
+int
+cistrncmp(char *s1, char *s2, int n)
+{
+	int c1, c2;
+
+	while(*s1 && n-- > 0){
+		c1 = *(uchar*)s1++;
+		c2 = *(uchar*)s2++;
+
+		if(c1 == c2)
+			continue;
+
+		if(c1 >= 'A' && c1 <= 'Z')
+			c1 -= 'A' - 'a';
+
+		if(c2 >= 'A' && c2 <= 'Z')
+			c2 -= 'A' - 'a';
+
+		if(c1 != c2)
+			return c1 - c2;
+	}
+	if(n <= 0)
+		return 0;
+	return -*s2;
+}
diff --git a/libkern/cistrstr.c b/libkern/cistrstr.c
new file mode 100644
index 00000000..20d3205d
--- /dev/null
+++ b/libkern/cistrstr.c
@@ -0,0 +1,22 @@
+#include "lib9.h"
+
+char*
+cistrstr(char *s, char *sub)
+{
+	int c, csub, n;
+
+	csub = *sub;
+	if(csub == '\0')
+		return s;
+	if(csub >= 'A' && csub <= 'Z')
+		csub -= 'A' - 'a';
+	sub++;
+	n = strlen(sub);
+	for(; c = *s; s++){
+		if(c >= 'A' && c <= 'Z')
+			c -= 'A' - 'a';
+		if(c == csub && cistrncmp(s+1, sub, n) == 0)
+			return s;
+	}
+	return nil;
+}
diff --git a/libkern/cleanname.c b/libkern/cleanname.c
new file mode 100644
index 00000000..fb8f370a
--- /dev/null
+++ b/libkern/cleanname.c
@@ -0,0 +1,51 @@
+#include "lib9.h"
+
+/*
+ * In place, rewrite name to compress multiple /, eliminate ., and process ..
+ */
+#define SEP(x)	((x)=='/' || (x) == 0)
+char*
+cleanname(char *name)
+{
+	char *p, *q, *dotdot;
+	int rooted;
+
+	rooted = name[0] == '/';
+
+	/*
+	 * invariants:
+	 *	p points at beginning of path element we're considering.
+	 *	q points just past the last path element we wrote (no slash).
+	 *	dotdot points just past the point where .. cannot backtrack
+	 *		any further (no slash).
+	 */
+	p = q = dotdot = name+rooted;
+	while(*p) {
+		if(p[0] == '/')	/* null element */
+			p++;
+		else if(p[0] == '.' && SEP(p[1]))
+			p += 1;	/* don't count the separator in case it is nul */
+		else if(p[0] == '.' && p[1] == '.' && SEP(p[2])) {
+			p += 2;
+			if(q > dotdot) {	/* can backtrack */
+				while(--q > dotdot && *q != '/')
+					;
+			} else if(!rooted) {	/* /.. is / but ./../ is .. */
+				if(q != name)
+					*q++ = '/';
+				*q++ = '.';
+				*q++ = '.';
+				dotdot = q;
+			}
+		} else {	/* real path element */
+			if(q != name+rooted)
+				*q++ = '/';
+			while((*q = *p) != '/' && *q != 0)
+				p++, q++;
+		}
+	}
+	if(q == name)	/* empty string is really ``.'' */
+		*q++ = '.';
+	*q = '\0';
+	return name;
+}
diff --git a/libkern/convD2M.c b/libkern/convD2M.c
new file mode 100644
index 00000000..29f0cd55
--- /dev/null
+++ b/libkern/convD2M.c
@@ -0,0 +1,94 @@
+#include	"lib9.h"
+#include	"fcall.h"
+
+uint
+sizeD2M(Dir *d)
+{
+	char *sv[4];
+	int i, ns;
+
+	sv[0] = d->name;
+	sv[1] = d->uid;
+	sv[2] = d->gid;
+	sv[3] = d->muid;
+
+	ns = 0;
+	for(i = 0; i < 4; i++)
+		if(sv[i])
+			ns += strlen(sv[i]);
+
+	return STATFIXLEN + ns;
+}
+
+uint
+convD2M(Dir *d, uchar *buf, uint nbuf)
+{
+	uchar *p, *ebuf;
+	char *sv[4];
+	int i, ns, nsv[4], ss;
+
+	if(nbuf < BIT16SZ)
+		return 0;
+
+	p = buf;
+	ebuf = buf + nbuf;
+
+	sv[0] = d->name;
+	sv[1] = d->uid;
+	sv[2] = d->gid;
+	sv[3] = d->muid;
+
+	ns = 0;
+	for(i = 0; i < 4; i++){
+		if(sv[i])
+			nsv[i] = strlen(sv[i]);
+		else
+			nsv[i] = 0;
+		ns += nsv[i];
+	}
+
+	ss = STATFIXLEN + ns;
+
+	/* set size befor erroring, so user can know how much is needed */
+	/* note that length excludes count field itself */
+	PBIT16(p, ss-BIT16SZ);
+	p += BIT16SZ;
+
+	if(ss > nbuf)
+		return BIT16SZ;
+
+	PBIT16(p, d->type);
+	p += BIT16SZ;
+	PBIT32(p, d->dev);
+	p += BIT32SZ;
+	PBIT8(p, d->qid.type);
+	p += BIT8SZ;
+	PBIT32(p, d->qid.vers);
+	p += BIT32SZ;
+	PBIT64(p, d->qid.path);
+	p += BIT64SZ;
+	PBIT32(p, d->mode);
+	p += BIT32SZ;
+	PBIT32(p, d->atime);
+	p += BIT32SZ;
+	PBIT32(p, d->mtime);
+	p += BIT32SZ;
+	PBIT64(p, d->length);
+	p += BIT64SZ;
+
+	for(i = 0; i < 4; i++){
+		ns = nsv[i];
+		if(p + ns + BIT16SZ > ebuf)
+			return 0;
+		PBIT16(p, ns);
+		p += BIT16SZ;
+		if(ns)
+			memmove(p, sv[i], ns);
+		p += ns;
+	}
+
+	if(ss != p - buf)
+		return 0;
+
+	return p - buf;
+}
diff --git a/libkern/convM2D.c b/libkern/convM2D.c
new file mode 100644
index 00000000..156bd763
--- /dev/null
+++ b/libkern/convM2D.c
@@ -0,0 +1,93 @@
+#include	"lib9.h"
+#include	"fcall.h"
+
+int
+statcheck(uchar *buf, uint nbuf)
+{
+	uchar *ebuf;
+	int i;
+
+	ebuf = buf + nbuf;
+
+	if(nbuf < STATFIXLEN || nbuf != BIT16SZ + GBIT16(buf))
+		return -1;
+
+	buf += STATFIXLEN - 4 * BIT16SZ;
+
+	for(i = 0; i < 4; i++){
+		if(buf + BIT16SZ > ebuf)
+			return -1;
+		buf += BIT16SZ + GBIT16(buf);
+	}
+
+	if(buf != ebuf)
+		return -1;
+
+	return 0;
+}
+
+static char nullstring[] = "";
+
+uint
+convM2D(uchar *buf, uint nbuf, Dir *d, char *strs)
+{
+	uchar *p, *ebuf;
+	char *sv[4];
+	int i, ns;
+
+	if(nbuf < STATFIXLEN)
+		return 0; 
+
+	p = buf;
+	ebuf = buf + nbuf;
+
+	p += BIT16SZ;	/* ignore size */
+	d->type = GBIT16(p);
+	p += BIT16SZ;
+	d->dev = GBIT32(p);
+	p += BIT32SZ;
+	d->qid.type = GBIT8(p);
+	p += BIT8SZ;
+	d->qid.vers = GBIT32(p);
+	p += BIT32SZ;
+	d->qid.path = GBIT64(p);
+	p += BIT64SZ;
+	d->mode = GBIT32(p);
+	p += BIT32SZ;
+	d->atime = GBIT32(p);
+	p += BIT32SZ;
+	d->mtime = GBIT32(p);
+	p += BIT32SZ;
+	d->length = GBIT64(p);
+	p += BIT64SZ;
+
+	for(i = 0; i < 4; i++){
+		if(p + BIT16SZ > ebuf)
+			return 0;
+		ns = GBIT16(p);
+		p += BIT16SZ;
+		if(p + ns > ebuf)
+			return 0;
+		if(strs){
+			sv[i] = strs;
+			memmove(strs, p, ns);
+			strs += ns;
+			*strs++ = '\0';
+		}
+		p += ns;
+	}
+
+	if(strs){
+		d->name = sv[0];
+		d->uid = sv[1];
+		d->gid = sv[2];
+		d->muid = sv[3];
+	}else{
+		d->name = nullstring;
+		d->uid = nullstring;
+		d->gid = nullstring;
+		d->muid = nullstring;
+	}
+	
+	return p - buf;
+}
diff --git a/libkern/convM2S.c b/libkern/convM2S.c
new file mode 100644
index 00000000..443d1957
--- /dev/null
+++ b/libkern/convM2S.c
@@ -0,0 +1,314 @@
+#include	"lib9.h"
+#include	"fcall.h"
+
+static
+uchar*
+gstring(uchar *p, uchar *ep, char **s)
+{
+	uint n;
+
+	if(p+BIT16SZ > ep)
+		return nil;
+	n = GBIT16(p);
+	p += BIT16SZ - 1;
+	if(p+n+1 > ep)
+		return nil;
+	/* move it down, on top of count, to make room for '\0' */
+	memmove(p, p + 1, n);
+	p[n] = '\0';
+	*s = (char*)p;
+	p += n+1;
+	return p;
+}
+
+static
+uchar*
+gqid(uchar *p, uchar *ep, Qid *q)
+{
+	if(p+QIDSZ > ep)
+		return nil;
+	q->type = GBIT8(p);
+	p += BIT8SZ;
+	q->vers = GBIT32(p);
+	p += BIT32SZ;
+	q->path = GBIT64(p);
+	p += BIT64SZ;
+	return p;
+}
+
+/*
+ * no syntactic checks.
+ * three causes for error:
+ *  1. message size field is incorrect
+ *  2. input buffer too short for its own data (counts too long, etc.)
+ *  3. too many names or qids
+ * gqid() and gstring() return nil if they would reach beyond buffer.
+ * main switch statement checks range and also can fall through
+ * to test at end of routine.
+ */
+uint
+convM2S(uchar *ap, uint nap, Fcall *f)
+{
+	uchar *p, *ep;
+	uint i, size;
+
+	p = ap;
+	ep = p + nap;
+
+	if(p+BIT32SZ+BIT8SZ+BIT16SZ > ep)
+		return 0;
+	size = GBIT32(p);
+	p += BIT32SZ;
+
+	if(size < BIT32SZ+BIT8SZ+BIT16SZ)
+		return 0;
+
+	f->type = GBIT8(p);
+	p += BIT8SZ;
+	f->tag = GBIT16(p);
+	p += BIT16SZ;
+
+	switch(f->type)
+	{
+	default:
+		return 0;
+
+	case Tversion:
+		if(p+BIT32SZ > ep)
+			return 0;
+		f->msize = GBIT32(p);
+		p += BIT32SZ;
+		p = gstring(p, ep, &f->version);
+		break;
+
+	case Tflush:
+		if(p+BIT16SZ > ep)
+			return 0;
+		f->oldtag = GBIT16(p);
+		p += BIT16SZ;
+		break;
+
+	case Tauth:
+		if(p+BIT32SZ > ep)
+			return 0;
+		f->afid = GBIT32(p);
+		p += BIT32SZ;
+		p = gstring(p, ep, &f->uname);
+		if(p == nil)
+			break;
+		p = gstring(p, ep, &f->aname);
+		if(p == nil)
+			break;
+		break;
+
+	case Tattach:
+		if(p+BIT32SZ > ep)
+			return 0;
+		f->fid = GBIT32(p);
+		p += BIT32SZ;
+		if(p+BIT32SZ > ep)
+			return 0;
+		f->afid = GBIT32(p);
+		p += BIT32SZ;
+		p = gstring(p, ep, &f->uname);
+		if(p == nil)
+			break;
+		p = gstring(p, ep, &f->aname);
+		if(p == nil)
+			break;
+		break;
+
+	case Twalk:
+		if(p+BIT32SZ+BIT32SZ+BIT16SZ > ep)
+			return 0;
+		f->fid = GBIT32(p);
+		p += BIT32SZ;
+		f->newfid = GBIT32(p);
+		p += BIT32SZ;
+		f->nwname = GBIT16(p);
+		p += BIT16SZ;
+		if(f->nwname > MAXWELEM)
+			return 0;
+		for(i=0; i<f->nwname; i++){
+			p = gstring(p, ep, &f->wname[i]);
+			if(p == nil)
+				break;
+		}
+		break;
+
+	case Topen:
+		if(p+BIT32SZ+BIT8SZ > ep)
+			return 0;
+		f->fid = GBIT32(p);
+		p += BIT32SZ;
+		f->mode = GBIT8(p);
+		p += BIT8SZ;
+		break;
+
+	case Tcreate:
+		if(p+BIT32SZ > ep)
+			return 0;
+		f->fid = GBIT32(p);
+		p += BIT32SZ;
+		p = gstring(p, ep, &f->name);
+		if(p == nil)
+			break;
+		if(p+BIT32SZ+BIT8SZ > ep)
+			return 0;
+		f->perm = GBIT32(p);
+		p += BIT32SZ;
+		f->mode = GBIT8(p);
+		p += BIT8SZ;
+		break;
+
+	case Tread:
+		if(p+BIT32SZ+BIT64SZ+BIT32SZ > ep)
+			return 0;
+		f->fid = GBIT32(p);
+		p += BIT32SZ;
+		f->offset = GBIT64(p);
+		p += BIT64SZ;
+		f->count = GBIT32(p);
+		p += BIT32SZ;
+		break;
+
+	case Twrite:
+		if(p+BIT32SZ+BIT64SZ+BIT32SZ > ep)
+			return 0;
+		f->fid = GBIT32(p);
+		p += BIT32SZ;
+		f->offset = GBIT64(p);
+		p += BIT64SZ;
+		f->count = GBIT32(p);
+		p += BIT32SZ;
+		if(p+f->count > ep)
+			return 0;
+		f->data = (char*)p;
+		p += f->count;
+		break;
+
+	case Tclunk:
+	case Tremove:
+		if(p+BIT32SZ > ep)
+			return 0;
+		f->fid = GBIT32(p);
+		p += BIT32SZ;
+		break;
+
+	case Tstat:
+		if(p+BIT32SZ > ep)
+			return 0;
+		f->fid = GBIT32(p);
+		p += BIT32SZ;
+		break;
+
+	case Twstat:
+		if(p+BIT32SZ+BIT16SZ > ep)
+			return 0;
+		f->fid = GBIT32(p);
+		p += BIT32SZ;
+		f->nstat = GBIT16(p);
+		p += BIT16SZ;
+		if(p+f->nstat > ep)
+			return 0;
+		f->stat = p;
+		p += f->nstat;
+		break;
+
+/*
+ */
+	case Rversion:
+		if(p+BIT32SZ > ep)
+			return 0;
+		f->msize = GBIT32(p);
+		p += BIT32SZ;
+		p = gstring(p, ep, &f->version);
+		break;
+
+	case Rerror:
+		p = gstring(p, ep, &f->ename);
+		break;
+
+	case Rflush:
+		break;
+
+	case Rauth:
+		p = gqid(p, ep, &f->aqid);
+		if(p == nil)
+			break;
+		break;
+
+	case Rattach:
+		p = gqid(p, ep, &f->qid);
+		if(p == nil)
+			break;
+		break;
+
+	case Rwalk:
+		if(p+BIT16SZ > ep)
+			return 0;
+		f->nwqid = GBIT16(p);
+		p += BIT16SZ;
+		if(f->nwqid > MAXWELEM)
+			return 0;
+		for(i=0; i<f->nwqid; i++){
+			p = gqid(p, ep, &f->wqid[i]);
+			if(p == nil)
+				break;
+		}
+		break;
+
+	case Ropen:
+	case Rcreate:
+		p = gqid(p, ep, &f->qid);
+		if(p == nil)
+			break;
+		if(p+BIT32SZ > ep)
+			return 0;
+		f->iounit = GBIT32(p);
+		p += BIT32SZ;
+		break;
+
+	case Rread:
+		if(p+BIT32SZ > ep)
+			return 0;
+		f->count = GBIT32(p);
+		p += BIT32SZ;
+		if(p+f->count > ep)
+			return 0;
+		f->data = (char*)p;
+		p += f->count;
+		break;
+
+	case Rwrite:
+		if(p+BIT32SZ > ep)
+			return 0;
+		f->count = GBIT32(p);
+		p += BIT32SZ;
+		break;
+
+	case Rclunk:
+	case Rremove:
+		break;
+
+	case Rstat:
+		if(p+BIT16SZ > ep)
+			return 0;
+		f->nstat = GBIT16(p);
+		p += BIT16SZ;
+		if(p+f->nstat > ep)
+			return 0;
+		f->stat = p;
+		p += f->nstat;
+		break;
+
+	case Rwstat:
+		break;
+	}
+
+	if(p==nil || p>ep)
+		return 0;
+	if(ap+size == p)
+		return size;
+	return 0;
+}
diff --git a/libkern/convS2M.c b/libkern/convS2M.c
new file mode 100644
index 00000000..7035373f
--- /dev/null
+++ b/libkern/convS2M.c
@@ -0,0 +1,385 @@
+#include	"lib9.h"
+#include	"fcall.h"
+
+static
+uchar*
+pstring(uchar *p, char *s)
+{
+	uint n;
+
+	if(s == nil){
+		PBIT16(p, 0);
+		p += BIT16SZ;
+		return p;
+	}
+
+	n = strlen(s);
+	PBIT16(p, n);
+	p += BIT16SZ;
+	memmove(p, s, n);
+	p += n;
+	return p;
+}
+
+static
+uchar*
+pqid(uchar *p, Qid *q)
+{
+	PBIT8(p, q->type);
+	p += BIT8SZ;
+	PBIT32(p, q->vers);
+	p += BIT32SZ;
+	PBIT64(p, q->path);
+	p += BIT64SZ;
+	return p;
+}
+
+static
+uint
+stringsz(char *s)
+{
+	if(s == nil)
+		return BIT16SZ;
+
+	return BIT16SZ+strlen(s);
+}
+
+uint
+sizeS2M(Fcall *f)
+{
+	uint n;
+	int i;
+
+	n = 0;
+	n += BIT32SZ;	/* size */
+	n += BIT8SZ;	/* type */
+	n += BIT16SZ;	/* tag */
+
+	switch(f->type)
+	{
+	default:
+		return 0;
+
+	case Tversion:
+		n += BIT32SZ;
+		n += stringsz(f->version);
+		break;
+
+	case Tflush:
+		n += BIT16SZ;
+		break;
+
+	case Tauth:
+		n += BIT32SZ;
+		n += stringsz(f->uname);
+		n += stringsz(f->aname);
+		break;
+
+	case Tattach:
+		n += BIT32SZ;
+		n += BIT32SZ;
+		n += stringsz(f->uname);
+		n += stringsz(f->aname);
+		break;
+
+	case Twalk:
+		n += BIT32SZ;
+		n += BIT32SZ;
+		n += BIT16SZ;
+		for(i=0; i<f->nwname; i++)
+			n += stringsz(f->wname[i]);
+		break;
+
+	case Topen:
+		n += BIT32SZ;
+		n += BIT8SZ;
+		break;
+
+	case Tcreate:
+		n += BIT32SZ;
+		n += stringsz(f->name);
+		n += BIT32SZ;
+		n += BIT8SZ;
+		break;
+
+	case Tread:
+		n += BIT32SZ;
+		n += BIT64SZ;
+		n += BIT32SZ;
+		break;
+
+	case Twrite:
+		n += BIT32SZ;
+		n += BIT64SZ;
+		n += BIT32SZ;
+		n += f->count;
+		break;
+
+	case Tclunk:
+	case Tremove:
+		n += BIT32SZ;
+		break;
+
+	case Tstat:
+		n += BIT32SZ;
+		break;
+
+	case Twstat:
+		n += BIT32SZ;
+		n += BIT16SZ;
+		n += f->nstat;
+		break;
+/*
+ */
+
+	case Rversion:
+		n += BIT32SZ;
+		n += stringsz(f->version);
+		break;
+
+	case Rerror:
+		n += stringsz(f->ename);
+		break;
+
+	case Rflush:
+		break;
+
+	case Rauth:
+		n += QIDSZ;
+		break;
+
+	case Rattach:
+		n += QIDSZ;
+		break;
+
+	case Rwalk:
+		n += BIT16SZ;
+		n += f->nwqid*QIDSZ;
+		break;
+
+	case Ropen:
+	case Rcreate:
+		n += QIDSZ;
+		n += BIT32SZ;
+		break;
+
+	case Rread:
+		n += BIT32SZ;
+		n += f->count;
+		break;
+
+	case Rwrite:
+		n += BIT32SZ;
+		break;
+
+	case Rclunk:
+		break;
+
+	case Rremove:
+		break;
+
+	case Rstat:
+		n += BIT16SZ;
+		n += f->nstat;
+		break;
+
+	case Rwstat:
+		break;
+	}
+	return n;
+}
+
+uint
+convS2M(Fcall *f, uchar *ap, uint nap)
+{
+	uchar *p;
+	uint i, size;
+
+	size = sizeS2M(f);
+	if(size == 0)
+		return 0;
+	if(size > nap)
+		return 0;
+
+	p = (uchar*)ap;
+
+	PBIT32(p, size);
+	p += BIT32SZ;
+	PBIT8(p, f->type);
+	p += BIT8SZ;
+	PBIT16(p, f->tag);
+	p += BIT16SZ;
+
+	switch(f->type)
+	{
+	default:
+		return 0;
+
+	case Tversion:
+		PBIT32(p, f->msize);
+		p += BIT32SZ;
+		p = pstring(p, f->version);
+		break;
+
+	case Tflush:
+		PBIT16(p, f->oldtag);
+		p += BIT16SZ;
+		break;
+
+	case Tauth:
+		PBIT32(p, f->afid);
+		p += BIT32SZ;
+		p  = pstring(p, f->uname);
+		p  = pstring(p, f->aname);
+		break;
+
+	case Tattach:
+		PBIT32(p, f->fid);
+		p += BIT32SZ;
+		PBIT32(p, f->afid);
+		p += BIT32SZ;
+		p  = pstring(p, f->uname);
+		p  = pstring(p, f->aname);
+		break;
+
+	case Twalk:
+		PBIT32(p, f->fid);
+		p += BIT32SZ;
+		PBIT32(p, f->newfid);
+		p += BIT32SZ;
+		PBIT16(p, f->nwname);
+		p += BIT16SZ;
+		if(f->nwname > MAXWELEM)
+			return 0;
+		for(i=0; i<f->nwname; i++)
+			p = pstring(p, f->wname[i]);
+		break;
+
+	case Topen:
+		PBIT32(p, f->fid);
+		p += BIT32SZ;
+		PBIT8(p, f->mode);
+		p += BIT8SZ;
+		break;
+
+	case Tcreate:
+		PBIT32(p, f->fid);
+		p += BIT32SZ;
+		p = pstring(p, f->name);
+		PBIT32(p, f->perm);
+		p += BIT32SZ;
+		PBIT8(p, f->mode);
+		p += BIT8SZ;
+		break;
+
+	case Tread:
+		PBIT32(p, f->fid);
+		p += BIT32SZ;
+		PBIT64(p, f->offset);
+		p += BIT64SZ;
+		PBIT32(p, f->count);
+		p += BIT32SZ;
+		break;
+
+	case Twrite:
+		PBIT32(p, f->fid);
+		p += BIT32SZ;
+		PBIT64(p, f->offset);
+		p += BIT64SZ;
+		PBIT32(p, f->count);
+		p += BIT32SZ;
+		memmove(p, f->data, f->count);
+		p += f->count;
+		break;
+
+	case Tclunk:
+	case Tremove:
+		PBIT32(p, f->fid);
+		p += BIT32SZ;
+		break;
+
+	case Tstat:
+		PBIT32(p, f->fid);
+		p += BIT32SZ;
+		break;
+
+	case Twstat:
+		PBIT32(p, f->fid);
+		p += BIT32SZ;
+		PBIT16(p, f->nstat);
+		p += BIT16SZ;
+		memmove(p, f->stat, f->nstat);
+		p += f->nstat;
+		break;
+/*
+ */
+
+	case Rversion:
+		PBIT32(p, f->msize);
+		p += BIT32SZ;
+		p = pstring(p, f->version);
+		break;
+
+	case Rerror:
+		p = pstring(p, f->ename);
+		break;
+
+	case Rflush:
+		break;
+
+	case Rauth:
+		p = pqid(p, &f->aqid);
+		break;
+
+	case Rattach:
+		p = pqid(p, &f->qid);
+		break;
+
+	case Rwalk:
+		PBIT16(p, f->nwqid);
+		p += BIT16SZ;
+		if(f->nwqid > MAXWELEM)
+			return 0;
+		for(i=0; i<f->nwqid; i++)
+			p = pqid(p, &f->wqid[i]);
+		break;
+
+	case Ropen:
+	case Rcreate:
+		p = pqid(p, &f->qid);
+		PBIT32(p, f->iounit);
+		p += BIT32SZ;
+		break;
+
+	case Rread:
+		PBIT32(p, f->count);
+		p += BIT32SZ;
+		memmove(p, f->data, f->count);
+		p += f->count;
+		break;
+
+	case Rwrite:
+		PBIT32(p, f->count);
+		p += BIT32SZ;
+		break;
+
+	case Rclunk:
+		break;
+
+	case Rremove:
+		break;
+
+	case Rstat:
+		PBIT16(p, f->nstat);
+		p += BIT16SZ;
+		memmove(p, f->stat, f->nstat);
+		p += f->nstat;
+		break;
+
+	case Rwstat:
+		break;
+	}
+	if(size != p-ap)
+		return 0;
+	return size;
+}
diff --git a/libkern/div-arm.s b/libkern/div-arm.s
new file mode 100644
index 00000000..a08289f9
--- /dev/null
+++ b/libkern/div-arm.s
@@ -0,0 +1,119 @@
+Q	= 0
+N	= 1
+D	= 2
+CC	= 3
+TMP	= 11
+
+TEXT	save<>(SB), 1, $0
+	MOVW	R(Q), 0(FP)
+	MOVW	R(N), 4(FP)
+	MOVW	R(D), 8(FP)
+	MOVW	R(CC), 12(FP)
+
+	MOVW	R(TMP), R(Q)		/* numerator */
+	MOVW	20(FP), R(D)		/* denominator */
+	CMP	$0, R(D)
+	BNE	s1
+	SWI		0
+/*	MOVW	-1(R(D)), R(TMP)	/* divide by zero fault */
+s1:	RET
+
+TEXT	rest<>(SB), 1, $0
+	MOVW	0(FP), R(Q)
+	MOVW	4(FP), R(N)
+	MOVW	8(FP), R(D)
+	MOVW	12(FP), R(CC)
+/*
+ * return to caller
+ * of rest<>
+ */
+	MOVW	0(R13), R14
+	ADD	$20, R13
+	B	(R14)
+
+TEXT	div<>(SB), 1, $0
+	MOVW	$32, R(CC)
+/*
+ * skip zeros 8-at-a-time
+ */
+e1:
+	AND.S	$(0xff<<24),R(Q), R(N)
+	BNE	e2
+	SLL	$8, R(Q)
+	SUB.S	$8, R(CC)
+	BNE	e1
+	RET
+e2:
+	MOVW	$0, R(N)
+
+loop:
+/*
+ * shift R(N||Q) left one
+ */
+	SLL	$1, R(N)
+	CMP	$0, R(Q)
+	ORR.LT	$1, R(N)
+	SLL	$1, R(Q)
+
+/*
+ * compare numerator to denominator
+ * if less, subtract and set quotent bit
+ */
+	CMP	R(D), R(N)
+	ORR.HS	$1, R(Q)
+	SUB.HS	R(D), R(N)
+	SUB.S	$1, R(CC)
+	BNE	loop
+	RET
+
+TEXT	_div(SB), 1, $16
+	BL	save<>(SB)
+	CMP	$0, R(Q)
+	BGE	d1
+	RSB	$0, R(Q), R(Q)
+	CMP	$0, R(D)
+	BGE	d2
+	RSB	$0, R(D), R(D)
+d0:
+	BL	div<>(SB)		/* none/both neg */
+	MOVW	R(Q), R(TMP)
+	B	out
+d1:
+	CMP	$0, R(D)
+	BGE	d0
+	RSB	$0, R(D), R(D)
+d2:
+	BL	div<>(SB)		/* one neg */
+	RSB	$0, R(Q), R(TMP)
+	B	out
+
+TEXT	_mod(SB), 1, $16
+	BL	save<>(SB)
+	CMP	$0, R(D)
+	RSB.LT	$0, R(D), R(D)
+	CMP	$0, R(Q)
+	BGE	m1
+	RSB	$0, R(Q), R(Q)
+	BL	div<>(SB)		/* neg numerator */
+	RSB	$0, R(N), R(TMP)
+	B	out
+m1:
+	BL	div<>(SB)		/* pos numerator */
+	MOVW	R(N), R(TMP)
+	B	out
+
+TEXT	_divu(SB), 1, $16
+	BL	save<>(SB)
+	BL	div<>(SB)
+	MOVW	R(Q), R(TMP)
+	B	out
+
+TEXT	_modu(SB), 1, $16
+	BL	save<>(SB)
+	BL	div<>(SB)
+	MOVW	R(N), R(TMP)
+	B	out
+
+out:
+	BL	rest<>(SB)
+	B	out
diff --git a/libkern/dofmt.c b/libkern/dofmt.c
new file mode 100644
index 00000000..0f47c85d
--- /dev/null
+++ b/libkern/dofmt.c
@@ -0,0 +1,532 @@
+/*
+ * The authors of this software are Rob Pike and Ken Thompson.
+ *              Copyright (c) 2002 by Lucent Technologies.
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose without fee is hereby granted, provided that this entire notice
+ * is included in all copies of any software which is or includes a copy
+ * or modification of this software and in all copies of the supporting
+ * documentation for such software.
+ * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR IMPLIED
+ * WARRANTY.  IN PARTICULAR, NEITHER THE AUTHORS NOR LUCENT TECHNOLOGIES MAKE ANY
+ * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE MERCHANTABILITY
+ * OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR PURPOSE.
+ */
+#include "lib9.h"
+#include "fmtdef.h"
+
+/* format the output into f->to and return the number of characters fmted  */
+int
+dofmt(Fmt *f, char *fmt)
+{
+	Rune rune, *rt, *rs;
+	int r;
+	char *t, *s;
+	int n, nfmt;
+
+	nfmt = f->nfmt;
+	for(;;){
+		if(f->runes){
+			rt = f->to;
+			rs = f->stop;
+			while((r = *(uchar*)fmt) && r != '%'){
+				if(r < Runeself)
+					fmt++;
+				else{
+					fmt += chartorune(&rune, fmt);
+					r = rune;
+				}
+				FMTRCHAR(f, rt, rs, r);
+			}
+			fmt++;
+			f->nfmt += rt - (Rune *)f->to;
+			f->to = rt;
+			if(!r)
+				return f->nfmt - nfmt;
+			f->stop = rs;
+		}else{
+			t = f->to;
+			s = f->stop;
+			while((r = *(uchar*)fmt) && r != '%'){
+				if(r < Runeself){
+					FMTCHAR(f, t, s, r);
+					fmt++;
+				}else{
+					n = chartorune(&rune, fmt);
+					if(t + n > s){
+						t = _fmtflush(f, t, n);
+						if(t != nil)
+							s = f->stop;
+						else
+							return -1;
+					}
+					while(n--)
+						*t++ = *fmt++;
+				}
+			}
+			fmt++;
+			f->nfmt += t - (char *)f->to;
+			f->to = t;
+			if(!r)
+				return f->nfmt - nfmt;
+			f->stop = s;
+		}
+
+		fmt = _fmtdispatch(f, fmt, 0);
+		if(fmt == nil)
+			return -1;
+	}
+	return 0;	/* not reached */
+}
+
+void *
+_fmtflush(Fmt *f, void *t, int len)
+{
+	if(f->runes)
+		f->nfmt += (Rune*)t - (Rune*)f->to;
+	else
+		f->nfmt += (char*)t - (char *)f->to;
+	f->to = t;
+	if(f->flush == 0 || (*f->flush)(f) == 0 || (char*)f->to + len > (char*)f->stop){
+		f->stop = f->to;
+		return nil;
+	}
+	return f->to;
+}
+
+/*
+ * put a formatted block of memory sz bytes long of n runes into the output buffer,
+ * left/right justified in a field of at least f->width charactes
+ */
+int
+_fmtpad(Fmt *f, int n)
+{
+	char *t, *s;
+	int i;
+
+	t = f->to;
+	s = f->stop;
+	for(i = 0; i < n; i++)
+		FMTCHAR(f, t, s, ' ');
+	f->nfmt += t - (char *)f->to;
+	f->to = t;
+	return 0;
+}
+
+int
+_rfmtpad(Fmt *f, int n)
+{
+	Rune *t, *s;
+	int i;
+
+	t = f->to;
+	s = f->stop;
+	for(i = 0; i < n; i++)
+		FMTRCHAR(f, t, s, ' ');
+	f->nfmt += t - (Rune *)f->to;
+	f->to = t;
+	return 0;
+}
+
+int
+_fmtcpy(Fmt *f, void *vm, int n, int sz)
+{
+	Rune *rt, *rs, r;
+	char *t, *s, *m, *me;
+	ulong fl;
+	int nc, w;
+
+	m = vm;
+	me = m + sz;
+	w = f->width;
+	fl = f->flags;
+	if((fl & FmtPrec) && n > f->prec)
+		n = f->prec;
+	if(f->runes){
+		if(!(fl & FmtLeft) && _rfmtpad(f, w - n) < 0)
+			return -1;
+		rt = f->to;
+		rs = f->stop;
+		for(nc = n; nc > 0; nc--){
+			r = *(uchar*)m;
+			if(r < Runeself)
+				m++;
+			else if((me - m) >= UTFmax || fullrune(m, me-m))
+				m += chartorune(&r, m);
+			else
+				break;
+			FMTRCHAR(f, rt, rs, r);
+		}
+		f->nfmt += rt - (Rune *)f->to;
+		f->to = rt;
+		if(m < me)
+			return -1;
+		if(fl & FmtLeft && _rfmtpad(f, w - n) < 0)
+			return -1;
+	}else{
+		if(!(fl & FmtLeft) && _fmtpad(f, w - n) < 0)
+			return -1;
+		t = f->to;
+		s = f->stop;
+		for(nc = n; nc > 0; nc--){
+			r = *(uchar*)m;
+			if(r < Runeself)
+				m++;
+			else if((me - m) >= UTFmax || fullrune(m, me-m))
+				m += chartorune(&r, m);
+			else
+				break;
+			FMTRUNE(f, t, s, r);
+		}
+		f->nfmt += t - (char *)f->to;
+		f->to = t;
+		if(fl & FmtLeft && _fmtpad(f, w - n) < 0)
+			return -1;
+	}
+	return 0;
+}
+
+int
+_fmtrcpy(Fmt *f, void *vm, int n)
+{
+	Rune r, *m, *me, *rt, *rs;
+	char *t, *s;
+	ulong fl;
+	int w;
+
+	m = vm;
+	w = f->width;
+	fl = f->flags;
+	if((fl & FmtPrec) && n > f->prec)
+		n = f->prec;
+	if(f->runes){
+		if(!(fl & FmtLeft) && _rfmtpad(f, w - n) < 0)
+			return -1;
+		rt = f->to;
+		rs = f->stop;
+		for(me = m + n; m < me; m++)
+			FMTRCHAR(f, rt, rs, *m);
+		f->nfmt += rt - (Rune *)f->to;
+		f->to = rt;
+		if(fl & FmtLeft && _rfmtpad(f, w - n) < 0)
+			return -1;
+	}else{
+		if(!(fl & FmtLeft) && _fmtpad(f, w - n) < 0)
+			return -1;
+		t = f->to;
+		s = f->stop;
+		for(me = m + n; m < me; m++){
+			r = *m;
+			FMTRUNE(f, t, s, r);
+		}
+		f->nfmt += t - (char *)f->to;
+		f->to = t;
+		if(fl & FmtLeft && _fmtpad(f, w - n) < 0)
+			return -1;
+	}
+	return 0;
+}
+
+/* fmt out one character */
+int
+_charfmt(Fmt *f)
+{
+	char x[1];
+
+	x[0] = va_arg(f->args, int);
+	f->prec = 1;
+	return _fmtcpy(f, x, 1, 1);
+}
+
+/* fmt out one rune */
+int
+_runefmt(Fmt *f)
+{
+	Rune x[1];
+
+	x[0] = va_arg(f->args, int);
+	return _fmtrcpy(f, x, 1);
+}
+
+/* public helper routine: fmt out a null terminated string already in hand */
+int
+fmtstrcpy(Fmt *f, char *s)
+{
+	int p, i;
+	if(!s)
+		return _fmtcpy(f, "<nil>", 5, 5);
+	/* if precision is specified, make sure we don't wander off the end */
+	if(f->flags & FmtPrec){
+		p = f->prec;
+		for(i = 0; i < p; i++)
+			if(s[i] == 0)
+				break;
+		return _fmtcpy(f, s, utfnlen(s, i), i);	/* BUG?: won't print a partial rune at end */
+	}
+
+	return _fmtcpy(f, s, utflen(s), strlen(s));
+}
+
+/* fmt out a null terminated utf string */
+int
+_strfmt(Fmt *f)
+{
+	char *s;
+
+	s = va_arg(f->args, char *);
+	return fmtstrcpy(f, s);
+}
+
+/* public helper routine: fmt out a null terminated rune string already in hand */
+int
+fmtrunestrcpy(Fmt *f, Rune *s)
+{
+	Rune *e;
+	int n, p;
+
+	if(!s)
+		return _fmtcpy(f, "<nil>", 5, 5);
+	/* if precision is specified, make sure we don't wander off the end */
+	if(f->flags & FmtPrec){
+		p = f->prec;
+		for(n = 0; n < p; n++)
+			if(s[n] == 0)
+				break;
+	}else{
+		for(e = s; *e; e++)
+			;
+		n = e - s;
+	}
+	return _fmtrcpy(f, s, n);
+}
+
+/* fmt out a null terminated rune string */
+int
+_runesfmt(Fmt *f)
+{
+	Rune *s;
+
+	s = va_arg(f->args, Rune *);
+	return fmtrunestrcpy(f, s);
+}
+
+/* fmt a % */
+int
+_percentfmt(Fmt *f)
+{
+	Rune x[1];
+
+	x[0] = f->r;
+	f->prec = 1;
+	return _fmtrcpy(f, x, 1);
+}
+
+/* fmt an integer */
+int
+_ifmt(Fmt *f)
+{
+	char buf[70], *p, *conv;
+	uvlong vu;
+	ulong u;
+	int neg, base, i, n, fl, w, isv;
+
+	neg = 0;
+	fl = f->flags;
+	isv = 0;
+	vu = 0;
+	u = 0;
+	if(f->r == 'p'){
+		u = (ulong)va_arg(f->args, void*);
+		f->r = 'x';
+		fl |= FmtUnsigned;
+	}else if(fl & FmtVLong){
+		isv = 1;
+		if(fl & FmtUnsigned)
+			vu = va_arg(f->args, uvlong);
+		else
+			vu = va_arg(f->args, vlong);
+	}else if(fl & FmtLong){
+		if(fl & FmtUnsigned)
+			u = va_arg(f->args, ulong);
+		else
+			u = va_arg(f->args, long);
+	}else if(fl & FmtByte){
+		if(fl & FmtUnsigned)
+			u = (uchar)va_arg(f->args, int);
+		else
+			u = (char)va_arg(f->args, int);
+	}else if(fl & FmtShort){
+		if(fl & FmtUnsigned)
+			u = (ushort)va_arg(f->args, int);
+		else
+			u = (short)va_arg(f->args, int);
+	}else{
+		if(fl & FmtUnsigned)
+			u = va_arg(f->args, uint);
+		else
+			u = va_arg(f->args, int);
+	}
+	conv = "0123456789abcdef";
+	switch(f->r){
+	case 'd':
+		base = 10;
+		break;
+	case 'x':
+		base = 16;
+		break;
+	case 'X':
+		base = 16;
+		conv = "0123456789ABCDEF";
+		break;
+	case 'b':
+		base = 2;
+		break;
+	case 'o':
+		base = 8;
+		break;
+	default:
+		return -1;
+	}
+	if(!(fl & FmtUnsigned)){
+		if(isv && (vlong)vu < 0){
+			vu = -(vlong)vu;
+			neg = 1;
+		}else if(!isv && (long)u < 0){
+			u = -(long)u;
+			neg = 1;
+		}
+	}
+	p = buf + sizeof buf - 1;
+	n = 0;
+	if(isv){
+		while(vu){
+			i = vu % base;
+			vu /= base;
+			if((fl & FmtComma) && n % 4 == 3){
+				*p-- = ',';
+				n++;
+			}
+			*p-- = conv[i];
+			n++;
+		}
+	}else{
+		while(u){
+			i = u % base;
+			u /= base;
+			if((fl & FmtComma) && n % 4 == 3){
+				*p-- = ',';
+				n++;
+			}
+			*p-- = conv[i];
+			n++;
+		}
+	}
+	if(n == 0){
+		*p-- = '0';
+		n = 1;
+	}
+	for(w = f->prec; n < w && p > buf+3; n++)
+		*p-- = '0';
+	if(neg || (fl & (FmtSign|FmtSpace)))
+		n++;
+	if(fl & FmtSharp){
+		if(base == 16)
+			n += 2;
+		else if(base == 8){
+			if(p[1] == '0')
+				fl &= ~FmtSharp;
+			else
+				n++;
+		}
+	}
+	if((fl & FmtZero) && !(fl & FmtLeft)){
+		for(w = f->width; n < w && p > buf+3; n++)
+			*p-- = '0';
+		f->width = 0;
+	}
+	if(fl & FmtSharp){
+		if(base == 16)
+			*p-- = f->r;
+		if(base == 16 || base == 8)
+			*p-- = '0';
+	}
+	if(neg)
+		*p-- = '-';
+	else if(fl & FmtSign)
+		*p-- = '+';
+	else if(fl & FmtSpace)
+		*p-- = ' ';
+	f->flags &= ~FmtPrec;
+	return _fmtcpy(f, p + 1, n, n);
+}
+
+int
+_countfmt(Fmt *f)
+{
+	void *p;
+	ulong fl;
+
+	fl = f->flags;
+	p = va_arg(f->args, void*);
+	if(fl & FmtVLong){
+		*(vlong*)p = f->nfmt;
+	}else if(fl & FmtLong){
+		*(long*)p = f->nfmt;
+	}else if(fl & FmtByte){
+		*(char*)p = f->nfmt;
+	}else if(fl & FmtShort){
+		*(short*)p = f->nfmt;
+	}else{
+		*(int*)p = f->nfmt;
+	}
+	return 0;
+}
+
+int
+_flagfmt(Fmt *f)
+{
+	switch(f->r){
+	case ',':
+		f->flags |= FmtComma;
+		break;
+	case '-':
+		f->flags |= FmtLeft;
+		break;
+	case '+':
+		f->flags |= FmtSign;
+		break;
+	case '#':
+		f->flags |= FmtSharp;
+		break;
+	case ' ':
+		f->flags |= FmtSpace;
+		break;
+	case 'u':
+		f->flags |= FmtUnsigned;
+		break;
+	case 'h':
+		if(f->flags & FmtShort)
+			f->flags |= FmtByte;
+		f->flags |= FmtShort;
+		break;
+	case 'l':
+		if(f->flags & FmtLong)
+			f->flags |= FmtVLong;
+		f->flags |= FmtLong;
+		break;
+	}
+	return 1;
+}
+
+/* default error format */
+int
+_badfmt(Fmt *f)
+{
+	char x[3];
+
+	x[0] = '%';
+	x[1] = f->r;
+	x[2] = '%';
+	f->prec = 3;
+	_fmtcpy(f, x, 3, 3);
+	return 0;
+}
diff --git a/libkern/exp.c b/libkern/exp.c
new file mode 100644
index 00000000..3098194b
--- /dev/null
+++ b/libkern/exp.c
@@ -0,0 +1,39 @@
+/*
+	exp returns the exponential function of its
+	floating-point argument.
+
+	The coefficients are #1069 from Hart and Cheney. (22.35D)
+*/
+
+#include <lib9.h>
+
+#define	p0  .2080384346694663001443843411e7
+#define	p1  .3028697169744036299076048876e5
+#define	p2  .6061485330061080841615584556e2
+#define	q0  .6002720360238832528230907598e7
+#define	q1  .3277251518082914423057964422e6
+#define	q2  .1749287689093076403844945335e4
+#define	log2e  1.4426950408889634073599247
+#define	sqrt2  1.4142135623730950488016887
+#define	maxf  10000
+
+double
+exp(double arg)
+{
+	double fract, temp1, temp2, xsq;
+	int ent;
+
+	if(arg == 0)
+		return 1;
+	if(arg < -maxf)
+		return 0;
+	if(arg > maxf)
+		return Inf(1);
+	arg *= log2e;
+	ent = floor(arg);
+	fract = (arg-ent) - 0.5;
+	xsq = fract*fract;
+	temp1 = ((p2*xsq+p1)*xsq+p0)*fract;
+	temp2 = ((xsq+q2)*xsq+q1)*xsq + q0;
+	return ldexp(sqrt2*(temp2+temp1)/(temp2-temp1), ent);
+}
diff --git a/libkern/fcallfmt.c b/libkern/fcallfmt.c
new file mode 100644
index 00000000..e8659b2e
--- /dev/null
+++ b/libkern/fcallfmt.c
@@ -0,0 +1,240 @@
+/*
+ * Copyright © 2001, 2002 Lucent Technologies Inc.  All rights reserved.
+ * Use, distribution etc are subject to the terms of the
+ * Lucent Technologies Inc. Plan 9 Open Source Licence Agreement
+ * available at http://plan9.bell-labs.com/plan9dist/license.html
+ */
+
+#include <lib9.h>
+#include <fcall.h>
+
+static uint dumpsome(char*, char*, char*, long);
+static void fdirconv(char*, char*, Dir*);
+static char *qidtype(char*, uchar);
+
+#define	QIDFMT	"(%.16llux %lud %s)"
+
+int
+fcallfmt(Fmt *fmt)
+{
+	Fcall *f;
+	int fid, type, tag, i;
+	char buf[512], tmp[200];
+	char *p, *e;
+	Dir *d;
+	Qid *q;
+
+	e = buf+sizeof(buf);
+	f = va_arg(fmt->args, Fcall*);
+	type = f->type;
+	fid = f->fid;
+	tag = f->tag;
+	switch(type){
+	case Tversion:	/* 100 */
+		seprint(buf, e, "Tversion tag %ud msize %ud version '%s'", tag, f->msize, f->version);
+		break;
+	case Rversion:
+		seprint(buf, e, "Rversion tag %ud msize %ud version '%s'", tag, f->msize, f->version);
+		break;
+	case Tauth:	/* 102 */
+		seprint(buf, e, "Tauth tag %ud afid %d uname %s aname %s", tag,
+			f->afid, f->uname, f->aname);
+		break;
+	case Rauth:
+		seprint(buf, e, "Rauth tag %ud qid " QIDFMT, tag,
+			f->aqid.path, f->aqid.vers, qidtype(tmp, f->aqid.type));
+		break;
+	case Tattach:	/* 104 */
+		seprint(buf, e, "Tattach tag %ud fid %d afid %d uname %s aname %s", tag,
+			fid, f->afid, f->uname, f->aname);
+		break;
+	case Rattach:
+		seprint(buf, e, "Rattach tag %ud qid " QIDFMT, tag,
+			f->qid.path, f->qid.vers, qidtype(tmp, f->qid.type));
+		break;
+	case Rerror:	/* 107; 106 (Terror) illegal */
+		seprint(buf, e, "Rerror tag %ud ename %s", tag, f->ename);
+		break;
+	case Tflush:	/* 108 */
+		seprint(buf, e, "Tflush tag %ud oldtag %ud", tag, f->oldtag);
+		break;
+	case Rflush:
+		seprint(buf, e, "Rflush tag %ud", tag);
+		break;
+	case Twalk:	/* 110 */
+		p = seprint(buf, e, "Twalk tag %ud fid %d newfid %d nwname %d ", tag, fid, f->newfid, f->nwname);
+		if(f->nwname <= MAXWELEM)
+			for(i=0; i<f->nwname; i++)
+				p = seprint(p, e, "%d:%s ", i, f->wname[i]);
+		break;
+	case Rwalk:
+		p = seprint(buf, e, "Rwalk tag %ud nwqid %ud ", tag, f->nwqid);
+		if(f->nwqid <= MAXWELEM)
+			for(i=0; i<f->nwqid; i++){
+				q = &f->wqid[i];
+				p = seprint(p, e, "%d:" QIDFMT " ", i,
+					q->path, q->vers, qidtype(tmp, q->type));
+			}
+		break;
+	case Topen:	/* 112 */
+		seprint(buf, e, "Topen tag %ud fid %ud mode %d", tag, fid, f->mode);
+		break;
+	case Ropen:
+		seprint(buf, e, "Ropen tag %ud qid " QIDFMT " iounit %ud ", tag,
+			f->qid.path, f->qid.vers, qidtype(tmp, f->qid.type), f->iounit);
+		break;
+	case Tcreate:	/* 114 */
+		seprint(buf, e, "Tcreate tag %ud fid %ud name %s perm %M mode %d", tag, fid, f->name, (ulong)f->perm, f->mode);
+		break;
+	case Rcreate:
+		seprint(buf, e, "Rcreate tag %ud qid " QIDFMT " iounit %ud ", tag,
+			f->qid.path, f->qid.vers, qidtype(tmp, f->qid.type), f->iounit);
+		break;
+	case Tread:	/* 116 */
+		seprint(buf, e, "Tread tag %ud fid %d offset %lld count %ud",
+			tag, fid, f->offset, f->count);
+		break;
+	case Rread:
+		p = seprint(buf, e, "Rread tag %ud count %ud ", tag, f->count);
+			dumpsome(p, e, f->data, f->count);
+		break;
+	case Twrite:	/* 118 */
+		p = seprint(buf, e, "Twrite tag %ud fid %d offset %lld count %ud ",
+			tag, fid, f->offset, f->count);
+		dumpsome(p, e, f->data, f->count);
+		break;
+	case Rwrite:
+		seprint(buf, e, "Rwrite tag %ud count %ud", tag, f->count);
+		break;
+	case Tclunk:	/* 120 */
+		seprint(buf, e, "Tclunk tag %ud fid %ud", tag, fid);
+		break;
+	case Rclunk:
+		seprint(buf, e, "Rclunk tag %ud", tag);
+		break;
+	case Tremove:	/* 122 */
+		seprint(buf, e, "Tremove tag %ud fid %ud", tag, fid);
+		break;
+	case Rremove:
+		seprint(buf, e, "Rremove tag %ud", tag);
+		break;
+	case Tstat:	/* 124 */
+		seprint(buf, e, "Tstat tag %ud fid %ud", tag, fid);
+		break;
+	case Rstat:
+		p = seprint(buf, e, "Rstat tag %ud ", tag);
+		if(f->nstat > sizeof tmp)
+			seprint(p, e, " stat(%d bytes)", f->nstat);
+		else{
+			d = (Dir*)tmp;
+			convM2D(f->stat, f->nstat, d, (char*)(d+1));
+			seprint(p, e, " stat ");
+			fdirconv(p+6, e, d);
+		}
+		break;
+	case Twstat:	/* 126 */
+		p = seprint(buf, e, "Twstat tag %ud fid %ud", tag, fid);
+		if(f->nstat > sizeof tmp)
+			seprint(p, e, " stat(%d bytes)", f->nstat);
+		else{
+			d = (Dir*)tmp;
+			convM2D(f->stat, f->nstat, d, (char*)(d+1));
+			seprint(p, e, " stat ");
+			fdirconv(p+6, e, d);
+		}
+		break;
+	case Rwstat:
+		seprint(buf, e, "Rwstat tag %ud", tag);
+		break;
+	default:
+		seprint(buf, e,  "unknown type %d", type);
+	}
+	return fmtstrcpy(fmt, buf);
+}
+
+static char*
+qidtype(char *s, uchar t)
+{
+	char *p;
+
+	p = s;
+	if(t & QTDIR)
+		*p++ = 'd';
+	if(t & QTAPPEND)
+		*p++ = 'a';
+	if(t & QTEXCL)
+		*p++ = 'l';
+	if(t & QTAUTH)
+		*p++ = 'A';
+	*p = '\0';
+	return s;
+}
+
+int
+dirfmt(Fmt *fmt)
+{
+	char buf[160];
+
+	fdirconv(buf, buf+sizeof buf, va_arg(fmt->args, Dir*));
+	return fmtstrcpy(fmt, buf);
+}
+
+static void
+fdirconv(char *buf, char *e, Dir *d)
+{
+	char tmp[16];
+
+	seprint(buf, e, "'%s' '%s' '%s' '%s' "
+		"q " QIDFMT " m %#luo "
+		"at %ld mt %ld l %lld "
+		"t %d d %d",
+			d->name, d->uid, d->gid, d->muid,
+			d->qid.path, d->qid.vers, qidtype(tmp, d->qid.type), d->mode,
+			d->atime, d->mtime, d->length,
+			d->type, d->dev);
+}
+
+/*
+ * dump out count (or DUMPL, if count is bigger) bytes from
+ * buf to ans, as a string if they are all printable,
+ * else as a series of hex bytes
+ */
+#define DUMPL 64
+
+static uint
+dumpsome(char *ans, char *e, char *buf, long count)
+{
+	int i, printable;
+	char *p;
+
+	if(buf == nil){
+		seprint(ans, e, "<no data>");
+		return strlen(ans);
+	}
+	printable = 1;
+	if(count > DUMPL)
+		count = DUMPL;
+	for(i=0; i<count && printable; i++)
+		if((buf[i]<32 && buf[i] !='\n' && buf[i] !='\t') || (uchar)buf[i]>127)
+			printable = 0;
+	p = ans;
+	*p++ = '\'';
+	if(printable){
+		if(count > e-p-2)
+			count = e-p-2;
+		memmove(p, buf, count);
+		p += count;
+	}else{
+		if(2*count > e-p-2)
+			count = (e-p-2)/2;
+		for(i=0; i<count; i++){
+			if(i>0 && i%4==0)
+				*p++ = ' ';
+			sprint(p, "%2.2ux", buf[i]);
+			p += 2;
+		}
+	}
+	*p++ = '\'';
+	*p = 0;
+	return p - ans;
+}
diff --git a/libkern/floor.c b/libkern/floor.c
new file mode 100644
index 00000000..86e6c816
--- /dev/null
+++ b/libkern/floor.c
@@ -0,0 +1,26 @@
+#include <lib9.h>
+/*
+ * floor and ceil-- greatest integer <= arg
+ * (resp least >=)
+ */
+
+double
+floor(double d)
+{
+	double fract;
+
+	if(d < 0) {
+		fract = modf(-d, &d);
+		if(fract != 0.0)
+			d += 1;
+		d = -d;
+	} else
+		modf(d, &d);
+	return d;
+}
+
+double
+ceil(double d)
+{
+	return -floor(-d);
+}
diff --git a/libkern/fmt.c b/libkern/fmt.c
new file mode 100644
index 00000000..11aac065
--- /dev/null
+++ b/libkern/fmt.c
@@ -0,0 +1,188 @@
+/*
+ * The authors of this software are Rob Pike and Ken Thompson.
+ *              Copyright (c) 2002 by Lucent Technologies.
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose without fee is hereby granted, provided that this entire notice
+ * is included in all copies of any software which is or includes a copy
+ * or modification of this software and in all copies of the supporting
+ * documentation for such software.
+ * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR IMPLIED
+ * WARRANTY.  IN PARTICULAR, NEITHER THE AUTHORS NOR LUCENT TECHNOLOGIES MAKE ANY
+ * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE MERCHANTABILITY
+ * OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR PURPOSE.
+ */
+#include "lib9.h"
+#include "fmtdef.h"
+
+enum
+{
+	Maxfmt = 64
+};
+
+typedef struct Convfmt Convfmt;
+struct Convfmt
+{
+	int	c;
+	volatile	Fmts	fmt;	/* for spin lock in fmtfmt; avoids race due to write order */
+};
+
+struct
+{
+	/* lock by calling _fmtlock, _fmtunlock */
+	int	nfmt;
+	Convfmt	fmt[Maxfmt];
+} fmtalloc;
+
+static Convfmt knownfmt[] = {
+	' ',	_flagfmt,
+	'#',	_flagfmt,
+	'%',	_percentfmt,
+	'+',	_flagfmt,
+	',',	_flagfmt,
+	'-',	_flagfmt,
+	'C',	_runefmt,
+	'S',	_runesfmt,
+	'X',	_ifmt,
+	'b',	_ifmt,
+	'c',	_charfmt,
+	'd',	_ifmt,
+	'h',	_flagfmt,
+	'l',	_flagfmt,
+	'n',	_countfmt,
+	'o',	_ifmt,
+	'p',	_ifmt,
+	'r',	errfmt,
+	's',	_strfmt,
+	'u',	_flagfmt,
+	'x',	_ifmt,
+	0,	nil,
+};
+
+int	(*doquote)(int);
+
+static Fmts
+fmtfmt(int c)
+{
+	Convfmt *p, *ep;
+
+	ep = &fmtalloc.fmt[fmtalloc.nfmt];
+	for(p=fmtalloc.fmt; p<ep; p++)
+		if(p->c == c)
+			return p->fmt;
+
+	/* is this a predefined format char? */
+	for(p=knownfmt; p->c; p++)
+		if(p->c == c){
+			/* no need to lock; fmtinstall is idempotent */
+			fmtinstall(p->c, p->fmt);
+			while(p->fmt == nil)	/* loop until value is updated */
+				;
+			return p->fmt;
+		}
+
+	return _badfmt;
+}
+
+int
+fmtinstall(int c, Fmts f)
+{
+	Convfmt *p, *ep;
+
+	if(c<=0 || c>=65536)
+		return -1;
+	if(!f)
+		f = _badfmt;
+
+	_fmtlock();
+
+	ep = &fmtalloc.fmt[fmtalloc.nfmt];
+	for(p=fmtalloc.fmt; p<ep; p++)
+		if(p->c == c)
+			break;
+
+	if(p == &fmtalloc.fmt[Maxfmt]){
+		_fmtunlock();
+		return -1;
+	}
+
+	p->fmt = f;
+	if(p == ep){	/* installing a new format character */
+		fmtalloc.nfmt++;
+		p->c = c;
+	}
+
+	_fmtunlock();
+	return 0;
+}
+
+void*
+_fmtdispatch(Fmt *f, void *fmt, int isrunes)
+{
+	Rune rune, r;
+	int i, n;
+
+	f->flags = 0;
+	f->width = f->prec = 0;
+
+	for(;;){
+		if(isrunes){
+			r = *(Rune*)fmt;
+			fmt = (Rune*)fmt + 1;
+		}else{
+			fmt = (char*)fmt + chartorune(&rune, fmt);
+			r = rune;
+		}
+		f->r = r;
+		switch(r){
+		case '\0':
+			return nil;
+		case '.':
+			f->flags |= FmtWidth|FmtPrec;
+			continue;
+		case '0':
+			if(!(f->flags & FmtWidth)){
+				f->flags |= FmtZero;
+				continue;
+			}
+			/* fall through */
+		case '1': case '2': case '3': case '4':
+		case '5': case '6': case '7': case '8': case '9':
+			i = 0;
+			while(r >= '0' && r <= '9'){
+				i = i * 10 + r - '0';
+				if(isrunes){
+					r = *(Rune*)fmt;
+					fmt = (Rune*)fmt + 1;
+				}else{
+					r = *(char*)fmt;
+					fmt = (char*)fmt + 1;
+				}
+			}
+			if(isrunes)
+				fmt = (Rune*)fmt - 1;
+			else
+				fmt = (char*)fmt - 1;
+		numflag:
+			if(f->flags & FmtWidth){
+				f->flags |= FmtPrec;
+				f->prec = i;
+			}else{
+				f->flags |= FmtWidth;
+				f->width = i;
+			}
+			continue;
+		case '*':
+			i = va_arg(f->args, int);
+			if(i < 0){
+				i = -i;
+				f->flags |= FmtLeft;
+			}
+			goto numflag;
+		}
+		n = (*fmtfmt(r))(f);
+		if(n < 0)
+			return nil;
+		if(n == 0)
+			return fmt;
+	}
+}
diff --git a/libkern/fmtdef.h b/libkern/fmtdef.h
new file mode 100644
index 00000000..d8f3680f
--- /dev/null
+++ b/libkern/fmtdef.h
@@ -0,0 +1,85 @@
+/*
+ * dofmt -- format to a buffer
+ * the number of characters formatted is returned,
+ * or -1 if there was an error.
+ * if the buffer is ever filled, flush is called.
+ * it should reset the buffer and return whether formatting should continue.
+ */
+
+typedef int (*Fmts)(Fmt*);
+
+typedef struct Quoteinfo Quoteinfo;
+struct Quoteinfo
+{
+	int	quoted;		/* if set, string must be quoted */
+	int	nrunesin;	/* number of input runes that can be accepted */
+	int	nbytesin;	/* number of input bytes that can be accepted */
+	int	nrunesout;	/* number of runes that will be generated */
+	int	nbytesout;	/* number of bytes that will be generated */
+};
+
+void	*_fmtflush(Fmt*, void*, int);
+void	*_fmtdispatch(Fmt*, void*, int);
+int	_floatfmt(Fmt*, double);
+int	_fmtpad(Fmt*, int);
+int	_rfmtpad(Fmt*, int);
+int	_fmtFdFlush(Fmt*);
+
+int	_efgfmt(Fmt*);
+int	_charfmt(Fmt*);
+int	_countfmt(Fmt*);
+int	_flagfmt(Fmt*);
+int	_percentfmt(Fmt*);
+int	_ifmt(Fmt*);
+int	_runefmt(Fmt*);
+int	_runesfmt(Fmt*);
+int	_strfmt(Fmt*);
+int	_badfmt(Fmt*);
+int	_fmtcpy(Fmt*, void*, int, int);
+int	_fmtrcpy(Fmt*, void*, int n);
+
+void	_fmtlock(void);
+void	_fmtunlock(void);
+
+#define FMTCHAR(f, t, s, c)\
+	do{\
+	if(t + 1 > (char*)s){\
+		t = _fmtflush(f, t, 1);\
+		if(t != nil)\
+			s = f->stop;\
+		else\
+			return -1;\
+	}\
+	*t++ = c;\
+	}while(0)
+
+#define FMTRCHAR(f, t, s, c)\
+	do{\
+	if(t + 1 > (Rune*)s){\
+		t = _fmtflush(f, t, sizeof(Rune));\
+		if(t != nil)\
+			s = f->stop;\
+		else\
+			return -1;\
+	}\
+	*t++ = c;\
+	}while(0)
+
+#define FMTRUNE(f, t, s, r)\
+	do{\
+	Rune _rune;\
+	int _runelen;\
+	if(t + UTFmax > (char*)s && t + (_runelen = runelen(r)) > (char*)s){\
+		t = _fmtflush(f, t, _runelen);\
+		if(t != nil)\
+			s = f->stop;\
+		else\
+			return -1;\
+	}\
+	if(r < Runeself)\
+		*t++ = r;\
+	else{\
+		_rune = r;\
+		t += runetochar(t, &_rune);\
+	}\
+	}while(0)
diff --git a/libkern/fmtprint.c b/libkern/fmtprint.c
new file mode 100644
index 00000000..c93cb315
--- /dev/null
+++ b/libkern/fmtprint.c
@@ -0,0 +1,44 @@
+/*
+ * The authors of this software are Rob Pike and Ken Thompson.
+ *              Copyright (c) 2002 by Lucent Technologies.
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose without fee is hereby granted, provided that this entire notice
+ * is included in all copies of any software which is or includes a copy
+ * or modification of this software and in all copies of the supporting
+ * documentation for such software.
+ * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR IMPLIED
+ * WARRANTY.  IN PARTICULAR, NEITHER THE AUTHORS NOR LUCENT TECHNOLOGIES MAKE ANY
+ * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE MERCHANTABILITY
+ * OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR PURPOSE.
+ */
+#include "lib9.h"
+#include "fmtdef.h"
+
+
+/*
+ * format a string into the output buffer
+ * designed for formats which themselves call fmt,
+ * but ignore any width flags
+ */
+int
+fmtprint(Fmt *f, char *fmt, ...)
+{
+	va_list va;
+	int n;
+
+	f->flags = 0;
+	f->width = 0;
+	f->prec = 0;
+	va = f->args;
+	va_start(f->args, fmt);
+	n = dofmt(f, fmt);
+	va_end(f->args);
+	f->flags = 0;
+	f->width = 0;
+	f->prec = 0;
+	f->args = va;
+	if(n >= 0)
+		return 0;
+	return n;
+}
+
diff --git a/libkern/fmtquote.c b/libkern/fmtquote.c
new file mode 100644
index 00000000..3e6338bf
--- /dev/null
+++ b/libkern/fmtquote.c
@@ -0,0 +1,259 @@
+/*
+ * The authors of this software are Rob Pike and Ken Thompson.
+ *              Copyright (c) 2002 by Lucent Technologies.
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose without fee is hereby granted, provided that this entire notice
+ * is included in all copies of any software which is or includes a copy
+ * or modification of this software and in all copies of the supporting
+ * documentation for such software.
+ * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR IMPLIED
+ * WARRANTY.  IN PARTICULAR, NEITHER THE AUTHORS NOR LUCENT TECHNOLOGIES MAKE ANY
+ * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE MERCHANTABILITY
+ * OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR PURPOSE.
+ */
+#include "lib9.h"
+#include "fmtdef.h"
+
+/*
+ * How many bytes of output UTF will be produced by quoting (if necessary) this string?
+ * How many runes? How much of the input will be consumed?
+ * The parameter q is filled in by _quotesetup.
+ * The string may be UTF or Runes (s or r).
+ * Return count does not include NUL.
+ * Terminate the scan at the first of:
+ *	NUL in input
+ *	count exceeded in input
+ *	count exceeded on output
+ * *ninp is set to number of input bytes accepted.
+ * nin may be <0 initially, to avoid checking input by count.
+ */
+void
+_quotesetup(char *s, Rune *r, int nin, int nout, Quoteinfo *q, int sharp, int runesout)
+{
+	int w;
+	Rune c;
+
+	q->quoted = 0;
+	q->nbytesout = 0;
+	q->nrunesout = 0;
+	q->nbytesin = 0;
+	q->nrunesin = 0;
+	if(sharp || nin==0 || (s && *s=='\0') || (r && *r=='\0')){
+		if(nout < 2)
+			return;
+		q->quoted = 1;
+		q->nbytesout = 2;
+		q->nrunesout = 2;
+	}
+	for(; nin!=0; nin-=w){
+		if(s)
+			w = chartorune(&c, s);
+		else{
+			c = *r;
+			w = runelen(c);
+		}
+
+		if(c == '\0')
+			break;
+		if(runesout){
+			if(q->nrunesout+1 > nout)
+				break;
+		}else{
+			if(q->nbytesout+w > nout)
+				break;
+		}
+
+		if((c <= L' ') || (c == L'\'') || (doquote!=nil && doquote(c))){
+			if(!q->quoted){
+				if(runesout){
+					if(1+q->nrunesout+1+1 > nout)	/* no room for quotes */
+						break;
+				}else{
+					if(1+q->nbytesout+w+1 > nout)	/* no room for quotes */
+						break;
+				}
+				q->nrunesout += 2;	/* include quotes */
+				q->nbytesout += 2;	/* include quotes */
+				q->quoted = 1;
+			}
+			if(c == '\'')	{
+				if(runesout){
+					if(1+q->nrunesout+1 > nout)	/* no room for quotes */
+						break;
+				}else{
+					if(1+q->nbytesout+w > nout)	/* no room for quotes */
+						break;
+				}
+				q->nbytesout++;
+				q->nrunesout++;	/* quotes reproduce as two characters */
+			}
+		}
+
+		/* advance input */
+		if(s)
+			s += w;
+		else
+			r++;
+		q->nbytesin += w;
+		q->nrunesin++;
+
+		/* advance output */
+		q->nbytesout += w;
+		q->nrunesout++;
+	}
+}
+
+static int
+qstrfmt(char *sin, Rune *rin, Quoteinfo *q, Fmt *f)
+{
+	Rune r, *rm, *rme;
+	char *t, *s, *m, *me;
+	Rune *rt, *rs;
+	ulong fl;
+	int nc, w;
+
+	m = sin;
+	me = m + q->nbytesin;
+	rm = rin;
+	rme = rm + q->nrunesin;
+
+	w = f->width;
+	fl = f->flags;
+	if(f->runes){
+		if(!(fl & FmtLeft) && _rfmtpad(f, w - q->nrunesout) < 0)
+			return -1;
+	}else{
+		if(!(fl & FmtLeft) && _fmtpad(f, w - q->nbytesout) < 0)
+			return -1;
+	}
+	t = f->to;
+	s = f->stop;
+	rt = f->to;
+	rs = f->stop;
+	if(f->runes)
+		FMTRCHAR(f, rt, rs, '\'');
+	else
+		FMTRUNE(f, t, s, '\'');
+	for(nc = q->nrunesin; nc > 0; nc--){
+		if(sin){
+			r = *(uchar*)m;
+			if(r < Runeself)
+				m++;
+			else if((me - m) >= UTFmax || fullrune(m, me-m))
+				m += chartorune(&r, m);
+			else
+				break;
+		}else{
+			if(rm >= rme)
+				break;
+			r = *rm++;
+		}
+		if(f->runes){
+			FMTRCHAR(f, rt, rs, r);
+			if(r == '\'')
+				FMTRCHAR(f, rt, rs, r);
+		}else{
+			FMTRUNE(f, t, s, r);
+			if(r == '\'')
+				FMTRUNE(f, t, s, r);
+		}
+	}
+
+	if(f->runes){
+		FMTRCHAR(f, rt, rs, '\'');
+		USED(rs);
+		f->nfmt += rt - (Rune *)f->to;
+		f->to = rt;
+		if(fl & FmtLeft && _rfmtpad(f, w - q->nrunesout) < 0)
+			return -1;
+	}else{
+		FMTRUNE(f, t, s, '\'');
+		USED(s);
+		f->nfmt += t - (char *)f->to;
+		f->to = t;
+		if(fl & FmtLeft && _fmtpad(f, w - q->nbytesout) < 0)
+			return -1;
+	}
+	return 0;
+}
+
+int
+_quotestrfmt(int runesin, Fmt *f)
+{
+	int outlen;
+	Rune *r;
+	char *s;
+	Quoteinfo q;
+
+	f->flags &= ~FmtPrec;	/* ignored for %q %Q, so disable for %s %S in easy case */
+	if(runesin){
+		r = va_arg(f->args, Rune *);
+		s = nil;
+	}else{
+		s = va_arg(f->args, char *);
+		r = nil;
+	}
+	if(!s && !r)
+		return _fmtcpy(f, "<nil>", 5, 5);
+
+	if(f->flush)
+		outlen = 0x7FFFFFFF;	/* if we can flush, no output limit */
+	else if(f->runes)
+		outlen = (Rune*)f->stop - (Rune*)f->to;
+	else
+		outlen = (char*)f->stop - (char*)f->to;
+
+	_quotesetup(s, r, -1, outlen, &q, f->flags&FmtSharp, f->runes);
+//print("bytes in %d bytes out %d runes in %d runesout %d\n", q.nbytesin, q.nbytesout, q.nrunesin, q.nrunesout);
+
+	if(runesin){
+		if(!q.quoted)
+			return _fmtrcpy(f, r, q.nrunesin);
+		return qstrfmt(nil, r, &q, f);
+	}
+
+	if(!q.quoted)
+		return _fmtcpy(f, s, q.nrunesin, q.nbytesin);
+	return qstrfmt(s, nil, &q, f);
+}
+
+int
+quotestrfmt(Fmt *f)
+{
+	return _quotestrfmt(0, f);
+}
+
+int
+quoterunestrfmt(Fmt *f)
+{
+	return _quotestrfmt(1, f);
+}
+
+void
+quotefmtinstall(void)
+{
+	fmtinstall('q', quotestrfmt);
+	fmtinstall('Q', quoterunestrfmt);
+}
+
+int
+_needsquotes(char *s, int *quotelenp)
+{
+	Quoteinfo q;
+
+	_quotesetup(s, nil, -1, 0x7FFFFFFF, &q, 0, 0);
+	*quotelenp = q.nbytesout;
+
+	return q.quoted;
+}
+
+int
+_runeneedsquotes(Rune *r, int *quotelenp)
+{
+	Quoteinfo q;
+
+	_quotesetup(nil, r, -1, 0x7FFFFFFF, &q, 0, 0);
+	*quotelenp = q.nrunesout;
+
+	return q.quoted;
+}
diff --git a/libkern/fmtstr.c b/libkern/fmtstr.c
new file mode 100644
index 00000000..9f94d134
--- /dev/null
+++ b/libkern/fmtstr.c
@@ -0,0 +1,23 @@
+/*
+ * The authors of this software are Rob Pike and Ken Thompson.
+ *              Copyright (c) 2002 by Lucent Technologies.
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose without fee is hereby granted, provided that this entire notice
+ * is included in all copies of any software which is or includes a copy
+ * or modification of this software and in all copies of the supporting
+ * documentation for such software.
+ * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR IMPLIED
+ * WARRANTY.  IN PARTICULAR, NEITHER THE AUTHORS NOR LUCENT TECHNOLOGIES MAKE ANY
+ * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE MERCHANTABILITY
+ * OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR PURPOSE.
+ */
+#include "lib9.h"
+
+char*
+fmtstrflush(Fmt *f)
+{
+	if(f->start == nil)
+		return nil;
+	*(char*)f->to = '\0';
+	return f->start;
+}
diff --git a/libkern/fmtvprint.c b/libkern/fmtvprint.c
new file mode 100644
index 00000000..6682cd33
--- /dev/null
+++ b/libkern/fmtvprint.c
@@ -0,0 +1,43 @@
+/*
+ * The authors of this software are Rob Pike and Ken Thompson.
+ *              Copyright (c) 2002 by Lucent Technologies.
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose without fee is hereby granted, provided that this entire notice
+ * is included in all copies of any software which is or includes a copy
+ * or modification of this software and in all copies of the supporting
+ * documentation for such software.
+ * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR IMPLIED
+ * WARRANTY.  IN PARTICULAR, NEITHER THE AUTHORS NOR LUCENT TECHNOLOGIES MAKE ANY
+ * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE MERCHANTABILITY
+ * OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR PURPOSE.
+ */
+#include "lib9.h"
+#include "fmtdef.h"
+
+
+/*
+ * format a string into the output buffer
+ * designed for formats which themselves call fmt,
+ * but ignore any width flags
+ */
+int
+fmtvprint(Fmt *f, char *fmt, va_list args)
+{
+	va_list va;
+	int n;
+
+	f->flags = 0;
+	f->width = 0;
+	f->prec = 0;
+	va = f->args;
+	f->args = args;
+	n = dofmt(f, fmt);
+	f->flags = 0;
+	f->width = 0;
+	f->prec = 0;
+	f->args = va;
+	if(n >= 0)
+		return 0;
+	return n;
+}
+
diff --git a/libkern/frexp-386.c b/libkern/frexp-386.c
new file mode 100644
index 00000000..8973a14d
--- /dev/null
+++ b/libkern/frexp-386.c
@@ -0,0 +1,79 @@
+#include <lib9.h>
+
+#define	MASK	0x7ffL
+#define	SHIFT	20
+#define	BIAS	1022L
+
+typedef	union
+{
+	double	d;
+	struct
+	{
+		long	ls;
+		long	ms;
+	};
+} Cheat;
+
+double
+frexp(double d, int *ep)
+{
+	Cheat x;
+
+	if(d == 0) {
+		*ep = 0;
+		return 0;
+	}
+	x.d = d;
+	*ep = ((x.ms >> SHIFT) & MASK) - BIAS;
+	x.ms &= ~(MASK << SHIFT);
+	x.ms |= BIAS << SHIFT;
+	return x.d;
+}
+
+double
+ldexp(double d, int e)
+{
+	Cheat x;
+
+	if(d == 0)
+		return 0;
+	x.d = d;
+	e += (x.ms >> SHIFT) & MASK;
+	if(e <= 0)
+		return 0;	/* underflow */
+	if(e >= MASK){		/* overflow */
+		if(d < 0)
+			return Inf(-1);
+		return Inf(1);
+	}
+	x.ms &= ~(MASK << SHIFT);
+	x.ms |= (long)e << SHIFT;
+	return x.d;
+}
+
+double
+modf(double d, double *ip)
+{
+	Cheat x;
+	int e;
+
+	if(d < 1) {
+		if(d < 0) {
+			x.d = modf(-d, ip);
+			*ip = -*ip;
+			return -x.d;
+		}
+		*ip = 0;
+		return d;
+	}
+	x.d = d;
+	e = ((x.ms >> SHIFT) & MASK) - BIAS;
+	if(e <= SHIFT+1) {
+		x.ms &= ~(0x1fffffL >> e);
+		x.ls = 0;
+	} else
+	if(e <= SHIFT+33)
+		x.ls &= ~(0x7fffffffL >> (e-SHIFT-2));
+	*ip = x.d;
+	return d - x.d;
+}
diff --git a/libkern/frexp-68000.c b/libkern/frexp-68000.c
new file mode 100644
index 00000000..ac22e037
--- /dev/null
+++ b/libkern/frexp-68000.c
@@ -0,0 +1,80 @@
+#include <u.h>
+#include <libc.h>
+
+#define	MASK	0x7ffL
+#define	SHIFT	20
+#define	BIAS	1022L
+
+typedef	union
+{
+	double	d;
+	struct
+	{
+		long	ms;
+		long	ls;
+	};
+} Cheat;
+
+double
+frexp(double d, int *ep)
+{
+	Cheat x;
+
+	if(d == 0) {
+		*ep = 0;
+		return 0;
+	}
+	x.d = d;
+	*ep = ((x.ms >> SHIFT) & MASK) - BIAS;
+	x.ms &= ~(MASK << SHIFT);
+	x.ms |= BIAS << SHIFT;
+	return x.d;
+}
+
+double
+ldexp(double d, int e)
+{
+	Cheat x;
+
+	if(d == 0)
+		return 0;
+	x.d = d;
+	e += (x.ms >> SHIFT) & MASK;
+	if(e <= 0)
+		return 0;	/* underflow */
+	if(e >= MASK){		/* overflow */
+		if(d < 0)
+			return Inf(-1);
+		return Inf(1);
+	}
+	x.ms &= ~(MASK << SHIFT);
+	x.ms |= (long)e << SHIFT;
+	return x.d;
+}
+
+double
+modf(double d, double *ip)
+{
+	Cheat x;
+	int e;
+
+	if(d < 1) {
+		if(d < 0) {
+			x.d = modf(-d, ip);
+			*ip = -*ip;
+			return -x.d;
+		}
+		*ip = 0;
+		return d;
+	}
+	x.d = d;
+	e = ((x.ms >> SHIFT) & MASK) - BIAS;
+	if(e <= SHIFT+1) {
+		x.ms &= ~(0x1fffffL >> e);
+		x.ls = 0;
+	} else
+	if(e <= SHIFT+33)
+		x.ls &= ~(0x7fffffffL >> (e-SHIFT-2));
+	*ip = x.d;
+	return d - x.d;
+}
diff --git a/libkern/frexp-arm.c b/libkern/frexp-arm.c
new file mode 100644
index 00000000..0acf6ec1
--- /dev/null
+++ b/libkern/frexp-arm.c
@@ -0,0 +1,79 @@
+#include <lib9.h>
+
+#define	MASK	0x7ffL
+#define	SHIFT	20
+#define	BIAS	1022L
+
+typedef	union
+{
+	double	d;
+	struct	/* this is BIGENDIAN */
+	{
+		long	ms;
+		long	ls;
+	};
+} Cheat;
+
+double
+frexp(double d, int *ep)
+{
+	Cheat x;
+
+	if(d == 0) {
+		*ep = 0;
+		return 0;
+	}
+	x.d = d;
+	*ep = ((x.ms >> SHIFT) & MASK) - BIAS;
+	x.ms &= ~(MASK << SHIFT);
+	x.ms |= BIAS << SHIFT;
+	return x.d;
+}
+
+double
+ldexp(double d, int e)
+{
+	Cheat x;
+
+	if(d == 0)
+		return 0;
+	x.d = d;
+	e += (x.ms >> SHIFT) & MASK;
+	if(e <= 0)
+		return 0;	/* underflow */
+	if(e >= MASK){		/* overflow */
+		if(d < 0)
+			return Inf(-1);
+		return Inf(1);
+	}
+	x.ms &= ~(MASK << SHIFT);
+	x.ms |= (long)e << SHIFT;
+	return x.d;
+}
+
+double
+modf(double d, double *ip)
+{
+	Cheat x;
+	int e;
+
+	if(d < 1) {
+		if(d < 0) {
+			x.d = modf(-d, ip);
+			*ip = -*ip;
+			return -x.d;
+		}
+		*ip = 0;
+		return d;
+	}
+	x.d = d;
+	e = ((x.ms >> SHIFT) & MASK) - BIAS;
+	if(e <= SHIFT+1) {
+		x.ms &= ~(0x1fffffL >> e);
+		x.ls = 0;
+	} else
+	if(e <= SHIFT+33)
+		x.ls &= ~(0x7fffffffL >> (e-SHIFT-2));
+	*ip = x.d;
+	return d - x.d;
+}
diff --git a/libkern/frexp-mips.c b/libkern/frexp-mips.c
new file mode 100644
index 00000000..ac22e037
--- /dev/null
+++ b/libkern/frexp-mips.c
@@ -0,0 +1,80 @@
+#include <u.h>
+#include <libc.h>
+
+#define	MASK	0x7ffL
+#define	SHIFT	20
+#define	BIAS	1022L
+
+typedef	union
+{
+	double	d;
+	struct
+	{
+		long	ms;
+		long	ls;
+	};
+} Cheat;
+
+double
+frexp(double d, int *ep)
+{
+	Cheat x;
+
+	if(d == 0) {
+		*ep = 0;
+		return 0;
+	}
+	x.d = d;
+	*ep = ((x.ms >> SHIFT) & MASK) - BIAS;
+	x.ms &= ~(MASK << SHIFT);
+	x.ms |= BIAS << SHIFT;
+	return x.d;
+}
+
+double
+ldexp(double d, int e)
+{
+	Cheat x;
+
+	if(d == 0)
+		return 0;
+	x.d = d;
+	e += (x.ms >> SHIFT) & MASK;
+	if(e <= 0)
+		return 0;	/* underflow */
+	if(e >= MASK){		/* overflow */
+		if(d < 0)
+			return Inf(-1);
+		return Inf(1);
+	}
+	x.ms &= ~(MASK << SHIFT);
+	x.ms |= (long)e << SHIFT;
+	return x.d;
+}
+
+double
+modf(double d, double *ip)
+{
+	Cheat x;
+	int e;
+
+	if(d < 1) {
+		if(d < 0) {
+			x.d = modf(-d, ip);
+			*ip = -*ip;
+			return -x.d;
+		}
+		*ip = 0;
+		return d;
+	}
+	x.d = d;
+	e = ((x.ms >> SHIFT) & MASK) - BIAS;
+	if(e <= SHIFT+1) {
+		x.ms &= ~(0x1fffffL >> e);
+		x.ls = 0;
+	} else
+	if(e <= SHIFT+33)
+		x.ls &= ~(0x7fffffffL >> (e-SHIFT-2));
+	*ip = x.d;
+	return d - x.d;
+}
diff --git a/libkern/frexp-power.c b/libkern/frexp-power.c
new file mode 100644
index 00000000..b13bbead
--- /dev/null
+++ b/libkern/frexp-power.c
@@ -0,0 +1,79 @@
+#include <lib9.h>
+
+#define	MASK	0x7ffL
+#define	SHIFT	20
+#define	BIAS	1022L
+
+typedef	union
+{
+	double	d;
+	struct
+	{
+		long	ms;
+		long	ls;
+	};
+} Cheat;
+
+double
+frexp(double d, int *ep)
+{
+	Cheat x;
+
+	if(d == 0) {
+		*ep = 0;
+		return 0;
+	}
+	x.d = d;
+	*ep = ((x.ms >> SHIFT) & MASK) - BIAS;
+	x.ms &= ~(MASK << SHIFT);
+	x.ms |= BIAS << SHIFT;
+	return x.d;
+}
+
+double
+ldexp(double d, int e)
+{
+	Cheat x;
+
+	if(d == 0)
+		return 0;
+	x.d = d;
+	e += (x.ms >> SHIFT) & MASK;
+	if(e <= 0)
+		return 0;	/* underflow */
+	if(e >= MASK){		/* overflow */
+		if(d < 0)
+			return Inf(-1);
+		return Inf(1);
+	}
+	x.ms &= ~(MASK << SHIFT);
+	x.ms |= (long)e << SHIFT;
+	return x.d;
+}
+
+double
+modf(double d, double *ip)
+{
+	Cheat x;
+	int e;
+
+	if(d < 1) {
+		if(d < 0) {
+			x.d = modf(-d, ip);
+			*ip = -*ip;
+			return -x.d;
+		}
+		*ip = 0;
+		return d;
+	}
+	x.d = d;
+	e = ((x.ms >> SHIFT) & MASK) - BIAS;
+	if(e <= SHIFT+1) {
+		x.ms &= ~(0x1fffffL >> e);
+		x.ls = 0;
+	} else
+	if(e <= SHIFT+33)
+		x.ls &= ~(0x7fffffffL >> (e-SHIFT-2));
+	*ip = x.d;
+	return d - x.d;
+}
diff --git a/libkern/frexp-sparc.c b/libkern/frexp-sparc.c
new file mode 100644
index 00000000..b13bbead
--- /dev/null
+++ b/libkern/frexp-sparc.c
@@ -0,0 +1,79 @@
+#include <lib9.h>
+
+#define	MASK	0x7ffL
+#define	SHIFT	20
+#define	BIAS	1022L
+
+typedef	union
+{
+	double	d;
+	struct
+	{
+		long	ms;
+		long	ls;
+	};
+} Cheat;
+
+double
+frexp(double d, int *ep)
+{
+	Cheat x;
+
+	if(d == 0) {
+		*ep = 0;
+		return 0;
+	}
+	x.d = d;
+	*ep = ((x.ms >> SHIFT) & MASK) - BIAS;
+	x.ms &= ~(MASK << SHIFT);
+	x.ms |= BIAS << SHIFT;
+	return x.d;
+}
+
+double
+ldexp(double d, int e)
+{
+	Cheat x;
+
+	if(d == 0)
+		return 0;
+	x.d = d;
+	e += (x.ms >> SHIFT) & MASK;
+	if(e <= 0)
+		return 0;	/* underflow */
+	if(e >= MASK){		/* overflow */
+		if(d < 0)
+			return Inf(-1);
+		return Inf(1);
+	}
+	x.ms &= ~(MASK << SHIFT);
+	x.ms |= (long)e << SHIFT;
+	return x.d;
+}
+
+double
+modf(double d, double *ip)
+{
+	Cheat x;
+	int e;
+
+	if(d < 1) {
+		if(d < 0) {
+			x.d = modf(-d, ip);
+			*ip = -*ip;
+			return -x.d;
+		}
+		*ip = 0;
+		return d;
+	}
+	x.d = d;
+	e = ((x.ms >> SHIFT) & MASK) - BIAS;
+	if(e <= SHIFT+1) {
+		x.ms &= ~(0x1fffffL >> e);
+		x.ls = 0;
+	} else
+	if(e <= SHIFT+33)
+		x.ls &= ~(0x7fffffffL >> (e-SHIFT-2));
+	*ip = x.d;
+	return d - x.d;
+}
diff --git a/libkern/frexp-thumb.c b/libkern/frexp-thumb.c
new file mode 100644
index 00000000..0acf6ec1
--- /dev/null
+++ b/libkern/frexp-thumb.c
@@ -0,0 +1,79 @@
+#include <lib9.h>
+
+#define	MASK	0x7ffL
+#define	SHIFT	20
+#define	BIAS	1022L
+
+typedef	union
+{
+	double	d;
+	struct	/* this is BIGENDIAN */
+	{
+		long	ms;
+		long	ls;
+	};
+} Cheat;
+
+double
+frexp(double d, int *ep)
+{
+	Cheat x;
+
+	if(d == 0) {
+		*ep = 0;
+		return 0;
+	}
+	x.d = d;
+	*ep = ((x.ms >> SHIFT) & MASK) - BIAS;
+	x.ms &= ~(MASK << SHIFT);
+	x.ms |= BIAS << SHIFT;
+	return x.d;
+}
+
+double
+ldexp(double d, int e)
+{
+	Cheat x;
+
+	if(d == 0)
+		return 0;
+	x.d = d;
+	e += (x.ms >> SHIFT) & MASK;
+	if(e <= 0)
+		return 0;	/* underflow */
+	if(e >= MASK){		/* overflow */
+		if(d < 0)
+			return Inf(-1);
+		return Inf(1);
+	}
+	x.ms &= ~(MASK << SHIFT);
+	x.ms |= (long)e << SHIFT;
+	return x.d;
+}
+
+double
+modf(double d, double *ip)
+{
+	Cheat x;
+	int e;
+
+	if(d < 1) {
+		if(d < 0) {
+			x.d = modf(-d, ip);
+			*ip = -*ip;
+			return -x.d;
+		}
+		*ip = 0;
+		return d;
+	}
+	x.d = d;
+	e = ((x.ms >> SHIFT) & MASK) - BIAS;
+	if(e <= SHIFT+1) {
+		x.ms &= ~(0x1fffffL >> e);
+		x.ls = 0;
+	} else
+	if(e <= SHIFT+33)
+		x.ls &= ~(0x7fffffffL >> (e-SHIFT-2));
+	*ip = x.d;
+	return d - x.d;
+}
diff --git a/libkern/getfcr-386.s b/libkern/getfcr-386.s
new file mode 100644
index 00000000..c3f3d464
--- /dev/null
+++ b/libkern/getfcr-386.s
@@ -0,0 +1,27 @@
+TEXT	setfcr(SB), $0
+	MOVL	p+0(FP),AX
+	XORB	$0x3f,AX
+	PUSHW	AX
+	WAIT
+	FLDCW	0(SP)
+	POPW	AX
+	RET
+
+TEXT	getfcr(SB), $0
+	PUSHW	AX
+	WAIT
+	FSTCW	0(SP)
+	POPW	AX
+	XORB	$0x3f,AX
+	RET
+
+TEXT	getfsr(SB), $0
+	WAIT
+	FSTSW	AX
+	RET
+
+TEXT	setfsr(SB), $0
+	WAIT
+	FCLEX
+	RET
+
diff --git a/libkern/getfcr-68000.s b/libkern/getfcr-68000.s
new file mode 100644
index 00000000..7f67d038
--- /dev/null
+++ b/libkern/getfcr-68000.s
@@ -0,0 +1,19 @@
+TEXT	getfsr(SB), $0
+	MOVL	$0, R0
+	MOVL	FPSR, R0
+	RTS
+
+TEXT	setfsr(SB), $0
+	MOVL	new+0(FP), R1
+	MOVL	R1, FPSR
+	RTS
+
+TEXT	getfcr(SB), $0
+	MOVL	$0, R0
+	MOVL	FPCR, R0
+	RTS
+
+TEXT	setfcr(SB), $0
+	MOVL	new+0(FP), R1
+	MOVL	R1, FPCR
+	RTS
diff --git a/libkern/getfcr-arm.s b/libkern/getfcr-arm.s
new file mode 100644
index 00000000..dc9a207b
--- /dev/null
+++ b/libkern/getfcr-arm.s
@@ -0,0 +1,12 @@
+TEXT	setfcr(SB), $0
+	RET
+
+TEXT	getfcr(SB), $0
+	RET
+
+TEXT	getfsr(SB), $0
+	RET
+
+TEXT	setfsr(SB), $0
+	RET
+
diff --git a/libkern/getfcr-mips.s b/libkern/getfcr-mips.s
new file mode 100644
index 00000000..9e84cbcc
--- /dev/null
+++ b/libkern/getfcr-mips.s
@@ -0,0 +1,15 @@
+TEXT	getfsr(SB), $0
+	MOVW	FCR31, R1
+	RET
+
+TEXT	setfsr(SB), $0
+	MOVW	R1, FCR31
+	RET
+
+TEXT	getfcr(SB), $0
+	MOVW	FCR31, R1
+	RET
+
+TEXT	setfcr(SB), $0
+	MOVW	R1, FCR31
+	RET
diff --git a/libkern/getfcr-power.s b/libkern/getfcr-power.s
new file mode 100644
index 00000000..b61d52e6
--- /dev/null
+++ b/libkern/getfcr-power.s
@@ -0,0 +1,28 @@
+TEXT	getfcr(SB), $8
+	MOVFL	FPSCR, F3
+	FMOVD	F3, f-8(SP)
+	MOVW	-4(SP), R3
+	RETURN
+
+TEXT	getfsr(SB), $8
+	MOVFL	FPSCR, F3
+	FMOVD	F3, f-8(SP)
+	MOVW	-4(SP), R3
+	RETURN
+
+TEXT	setfcr(SB), $8
+	SYNC
+	MOVW	R3, -4(SP)
+	FMOVD	-8(SP), F3
+	MOVFL	F3, FPSCR
+	ISYNC
+	RETURN
+
+TEXT	setfsr(SB), $8
+	SYNC
+	MOVW	R3, -4(SP)
+	FMOVD	-8(SP), F3
+	MOVFL	F3, FPSCR
+	ISYNC
+	RETURN
+
diff --git a/libkern/getfcr-sparc.s b/libkern/getfcr-sparc.s
new file mode 100644
index 00000000..0059ada5
--- /dev/null
+++ b/libkern/getfcr-sparc.s
@@ -0,0 +1,27 @@
+TEXT	getfsr(SB), $0
+	SUB	$4, R1
+	MOVW	FSR, (R1)
+	MOVW	(R1), R7
+	ADD	$4, R1
+	RETURN
+
+TEXT	setfsr(SB), $0
+	SUB	$4, R1
+	MOVW	R7, (R1)
+	MOVW	(R1), FSR
+	ADD	$4, R1
+	RETURN
+
+TEXT	setfcr(SB), $0
+	SUB	$4, R1
+	MOVW	R7, (R1)
+	MOVW	(R1), FSR
+	ADD	$4, R1
+	RETURN
+
+TEXT	getfcr(SB), $0
+	SUB	$4, R1
+	MOVW	FSR, (R1)
+	MOVW	(R1), R7
+	ADD	$4, R1
+	RETURN
diff --git a/libkern/getfcr-thumb.s b/libkern/getfcr-thumb.s
new file mode 100644
index 00000000..dc9a207b
--- /dev/null
+++ b/libkern/getfcr-thumb.s
@@ -0,0 +1,12 @@
+TEXT	setfcr(SB), $0
+	RET
+
+TEXT	getfcr(SB), $0
+	RET
+
+TEXT	getfsr(SB), $0
+	RET
+
+TEXT	setfsr(SB), $0
+	RET
+
diff --git a/libkern/getfields.c b/libkern/getfields.c
new file mode 100644
index 00000000..44d7417c
--- /dev/null
+++ b/libkern/getfields.c
@@ -0,0 +1,35 @@
+#include	"lib9.h"
+int
+getfields(char *str, char **args, int max, int mflag, char *set)
+{
+	Rune r;
+	int nr, intok, narg;
+
+	if(max <= 0)
+		return 0;
+
+	narg = 0;
+	args[narg] = str;
+	if(!mflag)
+		narg++;
+	intok = 0;
+	for(;; str += nr) {
+		nr = chartorune(&r, str);
+		if(r == 0)
+			break;
+		if(utfrune(set, r)) {
+			if(narg >= max)
+				break;
+			*str = 0;
+			intok = 0;
+			args[narg] = str + nr;
+			if(!mflag)
+				narg++;
+		} else {
+			if(!intok && mflag)
+				narg++;
+			intok = 1;
+		}
+	}
+	return narg;
+}
diff --git a/libkern/log.c b/libkern/log.c
new file mode 100644
index 00000000..1387aa3d
--- /dev/null
+++ b/libkern/log.c
@@ -0,0 +1,57 @@
+/*
+	log returns the natural logarithm of its floating
+	point argument.
+
+	The coefficients are #2705 from Hart & Cheney. (19.38D)
+
+	It calls frexp.
+*/
+
+#include <lib9.h>
+
+#define	log2    0.693147180559945309e0
+#define	ln10o1  .4342944819032518276511
+#define	sqrto2  0.707106781186547524e0
+#define	p0      -.240139179559210510e2
+#define	p1      0.309572928215376501e2
+#define	p2      -.963769093377840513e1
+#define	p3      0.421087371217979714e0
+#define	q0      -.120069589779605255e2
+#define	q1      0.194809660700889731e2
+#define	q2      -.891110902798312337e1
+
+double
+log(double arg)
+{
+	double x, z, zsq, temp;
+	int exp;
+
+	if(arg <= 0)
+		return NaN();
+	x = frexp(arg, &exp);
+	while(x < 0.5) {
+		x *= 2;
+		exp--;
+	}
+	if(x < sqrto2) {
+		x *= 2;
+		exp--;
+	}
+
+	z = (x-1) / (x+1);
+	zsq = z*z;
+
+	temp = ((p3*zsq + p2)*zsq + p1)*zsq + p0;
+	temp = temp/(((zsq + q2)*zsq + q1)*zsq + q0);
+	temp = temp*z + exp*log2;
+	return temp;
+}
+
+double
+log10(double arg)
+{
+
+	if(arg <= 0)
+		return NaN();
+	return log(arg) * ln10o1;
+}
diff --git a/libkern/memccpy-power.s b/libkern/memccpy-power.s
new file mode 100644
index 00000000..120df6ed
--- /dev/null
+++ b/libkern/memccpy-power.s
@@ -0,0 +1,23 @@
+	TEXT	memccpy(SB), $0
+#define	BDNZ	BC	16,0,
+MOVW R3, s1+0(FP)
+	MOVW	n+12(FP), R7
+	MOVW	s2+4(FP), R4
+	MOVBZ	c+11(FP), R5
+	CMP	R7, $0
+	BEQ	nf
+	MOVW	R7, CTR
+	SUB	$1, R3
+	SUB	$1, R4
+l1:
+	MOVBZU	1(R4), R6
+	CMP	R6, R5
+	MOVBZU	R6, 1(R3)
+	BEQ	eq
+	BDNZ	l1
+nf:
+	MOVW	$0, R3
+	RETURN
+eq:
+	ADD	$1, R3
+	RETURN
diff --git a/libkern/memccpy.c b/libkern/memccpy.c
new file mode 100644
index 00000000..a7c9118a
--- /dev/null
+++ b/libkern/memccpy.c
@@ -0,0 +1,17 @@
+#include	<lib9.h>
+
+void*
+memccpy(void *a1, void *a2, int c, ulong n)
+{
+	uchar *s1, *s2;
+
+	s1 = a1;
+	s2 = a2;
+	c &= 0xFF;
+	while(n > 0) {
+		if((*s1++ = *s2++) == c)
+			return s1;
+		n--;
+	}
+	return 0;
+}
diff --git a/libkern/memchr.c b/libkern/memchr.c
new file mode 100644
index 00000000..5692f435
--- /dev/null
+++ b/libkern/memchr.c
@@ -0,0 +1,16 @@
+#include	<lib9.h>
+
+void*
+memchr(void *ap, int c, ulong n)
+{
+	uchar *sp;
+
+	sp = ap;
+	c &= 0xFF;
+	while(n > 0) {
+		if(*sp++ == c)
+			return sp-1;
+		n--;
+	}
+	return 0;
+}
diff --git a/libkern/memcmp-power.s b/libkern/memcmp-power.s
new file mode 100644
index 00000000..f524fa9d
--- /dev/null
+++ b/libkern/memcmp-power.s
@@ -0,0 +1,110 @@
+	TEXT	memcmp(SB), $0
+#define	BDNZ	BC	16,0,
+	MOVW R3, s1+0(FP)		/* R3 is pointer1 */
+
+/*
+ * performance:
+ *	67mb/sec aligned; 16mb/sec unaligned
+ */
+
+	MOVW	n+8(FP), R4		/* R4 is count */
+	MOVW	s2+4(FP), R5		/* R5 is pointer2 */
+
+/*
+ * let LSW do the work for 4 characters or less; aligned and unaligned
+ */
+	CMP	R4, $0
+	BLE	eq
+	CMP	R4, $4
+	BLE	out
+
+	XOR	R3, R5, R9
+	ANDCC	$3, R9
+	BNE	l4	/* pointers misaligned; use LSW loop */
+
+/*
+ * do enough bytes to align pointers
+ */
+	ANDCC	$3,R3, R9
+	BEQ	l2
+	SUBC	R9, $4, R9
+	MOVW	R9, XER
+	LSW	(R3), R10
+	ADD	R9, R3
+	LSW	(R5), R14
+	ADD	R9, R5
+	SUB	R9, R4
+	CMPU	R10, R14
+	BNE	ne
+
+/*
+ * compare 16 at a time
+ */
+l2:
+	SRAWCC	$4, R4, R9
+	BLE	l4
+	MOVW	R9, CTR
+	SUB	$4, R3
+	SUB	$4, R5
+l3:
+	MOVWU	4(R3), R10
+	MOVWU	4(R5), R12
+	MOVWU	4(R3), R11
+	MOVWU	4(R5), R13
+	CMPU	R10, R12
+	BNE	ne
+	MOVWU	4(R3), R10
+	MOVWU	4(R5), R12
+	CMPU	R11, R13
+	BNE	ne
+	MOVWU	4(R3), R11
+	MOVWU	4(R5), R13
+	CMPU	R10, R12
+	BNE	ne
+	CMPU	R11, R13
+	BNE	ne
+	BDNZ	l3
+	ADD	$4, R3
+	ADD	$4, R5
+	RLWNMCC	$0, R4, $15, R4	/* residue */
+	BEQ	eq
+
+/*
+ * do remaining words with LSW; also does unaligned case
+ */
+l4:
+	SRAWCC	$2, R4, R9
+	BLE	out
+	MOVW	R9, CTR
+l5:
+	LSW	(R3), $4, R10
+	ADD	$4, R3
+	LSW	(R5), $4, R11
+	ADD	$4, R5
+	CMPU	R10, R11
+	BNE	ne
+	BDNZ	l5
+	RLWNMCC	$0, R4, $3, R4	/* residue */
+	BEQ	eq
+
+/*
+ * do remaining bytes with final LSW
+ */
+out:
+	MOVW	R4, XER
+	LSW	(R3), R10
+	LSW	(R5), R11
+	CMPU	R10, R11
+	BNE	ne
+
+eq:
+	MOVW	$0, R3
+	RETURN
+
+ne:
+	MOVW	$1, R3
+	BGE	ret
+	MOVW	$-1,R3
+ret:
+	RETURN
+	END
diff --git a/libkern/memcmp.c b/libkern/memcmp.c
new file mode 100644
index 00000000..01be7c62
--- /dev/null
+++ b/libkern/memcmp.c
@@ -0,0 +1,22 @@
+#include	<lib9.h>
+
+int
+memcmp(void *a1, void *a2, ulong n)
+{
+	uchar *s1, *s2;
+	uint c1, c2;
+
+	s1 = a1;
+	s2 = a2;
+	while(n > 0) {
+		c1 = *s1++;
+		c2 = *s2++;
+		if(c1 != c2) {
+			if(c1 > c2)
+				return 1;
+			return -1;
+		}
+		n--;
+	}
+	return 0;
+}
diff --git a/libkern/memcpy-386.s b/libkern/memcpy-386.s
new file mode 100644
index 00000000..e85bb232
--- /dev/null
+++ b/libkern/memcpy-386.s
@@ -0,0 +1,58 @@
+	TEXT	memmove(SB), $0
+	TEXT	memcpy(SB), $0
+
+	MOVL	p1+0(FP), DI
+	MOVL	p2+4(FP), SI
+	MOVL	n+8(FP), BX
+	CMPL	BX, $0
+	JGE	ok
+	MOVL	$0, SI
+ok:
+	CLD
+/*
+ * check and set for backwards
+ */
+	CMPL	SI, DI
+	JLS	back
+/*
+ * copy whole longs
+ */
+	MOVL	BX, CX
+	SHRL	$2, CX
+	REP;	MOVSL
+/*
+ * copy the rest, by bytes
+ */
+	ANDL	$3, BX
+	MOVL	BX, CX
+	REP;	MOVSB
+
+	MOVL	p+0(FP),AX
+	RET
+/*
+ * whole thing backwards has
+ * adjusted addresses
+ */
+back:
+	ADDL	BX, DI
+	ADDL	BX, SI
+	SUBL	$4, DI
+	SUBL	$4, SI
+	STD
+/*
+ * copy whole longs
+ */
+	MOVL	BX, CX
+	SHRL	$2, CX
+	ANDL	$3, BX
+	REP;	MOVSL
+/*
+ * copy the rest, by bytes
+ */
+	ADDL	$3, DI
+	ADDL	$3, SI
+	MOVL	BX, CX
+	REP;	MOVSB
+
+	MOVL	p+0(FP),AX
+	RET
diff --git a/libkern/memcpy-arm.c b/libkern/memcpy-arm.c
new file mode 100644
index 00000000..35f0afc5
--- /dev/null
+++ b/libkern/memcpy-arm.c
@@ -0,0 +1,35 @@
+#include	<u.h>
+#include	<libc.h>
+
+void*
+memmove(void *a1, void *a2, ulong n)
+{
+	char *s1, *s2;
+
+	if((long)n < 0)
+		abort();
+	if(a1 > a2)
+		goto back;
+	s1 = a1;
+	s2 = a2;
+	while(n > 0) {
+		*s1++ = *s2++;
+		n--;
+	}
+	return a1;
+
+back:
+	s1 = (char*)a1 + n;
+	s2 = (char*)a2 + n;
+	while(n > 0) {
+		*--s1 = *--s2;
+		n--;
+	}
+	return a1;
+}
+
+void*
+memcpy(void *a1, void *a2, ulong n)
+{
+	return memmove(a1, a2, n);
+}
diff --git a/libkern/memcpy-thumb.c b/libkern/memcpy-thumb.c
new file mode 100644
index 00000000..35f0afc5
--- /dev/null
+++ b/libkern/memcpy-thumb.c
@@ -0,0 +1,35 @@
+#include	<u.h>
+#include	<libc.h>
+
+void*
+memmove(void *a1, void *a2, ulong n)
+{
+	char *s1, *s2;
+
+	if((long)n < 0)
+		abort();
+	if(a1 > a2)
+		goto back;
+	s1 = a1;
+	s2 = a2;
+	while(n > 0) {
+		*s1++ = *s2++;
+		n--;
+	}
+	return a1;
+
+back:
+	s1 = (char*)a1 + n;
+	s2 = (char*)a2 + n;
+	while(n > 0) {
+		*--s1 = *--s2;
+		n--;
+	}
+	return a1;
+}
+
+void*
+memcpy(void *a1, void *a2, ulong n)
+{
+	return memmove(a1, a2, n);
+}
diff --git a/libkern/memmove-386.s b/libkern/memmove-386.s
new file mode 100644
index 00000000..e85bb232
--- /dev/null
+++ b/libkern/memmove-386.s
@@ -0,0 +1,58 @@
+	TEXT	memmove(SB), $0
+	TEXT	memcpy(SB), $0
+
+	MOVL	p1+0(FP), DI
+	MOVL	p2+4(FP), SI
+	MOVL	n+8(FP), BX
+	CMPL	BX, $0
+	JGE	ok
+	MOVL	$0, SI
+ok:
+	CLD
+/*
+ * check and set for backwards
+ */
+	CMPL	SI, DI
+	JLS	back
+/*
+ * copy whole longs
+ */
+	MOVL	BX, CX
+	SHRL	$2, CX
+	REP;	MOVSL
+/*
+ * copy the rest, by bytes
+ */
+	ANDL	$3, BX
+	MOVL	BX, CX
+	REP;	MOVSB
+
+	MOVL	p+0(FP),AX
+	RET
+/*
+ * whole thing backwards has
+ * adjusted addresses
+ */
+back:
+	ADDL	BX, DI
+	ADDL	BX, SI
+	SUBL	$4, DI
+	SUBL	$4, SI
+	STD
+/*
+ * copy whole longs
+ */
+	MOVL	BX, CX
+	SHRL	$2, CX
+	ANDL	$3, BX
+	REP;	MOVSL
+/*
+ * copy the rest, by bytes
+ */
+	ADDL	$3, DI
+	ADDL	$3, SI
+	MOVL	BX, CX
+	REP;	MOVSB
+
+	MOVL	p+0(FP),AX
+	RET
diff --git a/libkern/memmove-68000.s b/libkern/memmove-68000.s
new file mode 100644
index 00000000..efb0853e
--- /dev/null
+++ b/libkern/memmove-68000.s
@@ -0,0 +1,124 @@
+	TEXT	memmove(SB), $0
+	BRA	move
+
+	TEXT	memcpy(SB), $0
+move:
+
+	MOVL	n+8(FP), R0	/* count */
+	BEQ	return
+	BGT	ok
+	MOVL	0, R0
+ok:
+	MOVL	s1+0(FP), A2	/* dest pointer */
+	MOVL	s2+4(FP), A1	/* source pointer */
+
+	CMPL	A2,A1
+	BHI	back
+
+/*
+ * byte-at-a-time foreward copy to
+ * get source (A1) alligned.
+ */
+f1:
+	MOVL	A1, R1
+	ANDL	$3, R1
+	BEQ	f2
+	SUBL	$1, R0
+	BLT	return
+	MOVB	(A1)+, (A2)+
+	BRA	f1
+
+/*
+ * check that dest is alligned
+ * if not, just go byte-at-a-time
+ */
+f2:
+	MOVL	A2, R1
+	ANDL	$3, R1
+	BEQ	f3
+	SUBL	$1, R0
+	BLT	return
+	BRA	f5
+/*
+ * quad-long-at-a-time forward copy
+ */
+f3:
+	SUBL	$16, R0
+	BLT	f4
+	MOVL	(A1)+, (A2)+
+	MOVL	(A1)+, (A2)+
+	MOVL	(A1)+, (A2)+
+	MOVL	(A1)+, (A2)+
+	BRA	f3
+
+/*
+ * cleanup byte-at-a-time
+ */
+f4:
+	ADDL	$15, R0
+	BLT	return
+f5:
+	MOVB	(A1)+, (A2)+
+	SUBL	$1, R0
+	BGE	f5
+	BRA	return
+
+return:
+	MOVL	s1+0(FP),R0
+	RTS
+
+/*
+ * everything the same, but
+ * copy backwards
+ */
+back:
+	ADDL	R0, A1
+	ADDL	R0, A2
+
+/*
+ * byte-at-a-time backward copy to
+ * get source (A1) alligned.
+ */
+b1:
+	MOVL	A1, R1
+	ANDL	$3, R1
+	BEQ	b2
+	SUBL	$1, R0
+	BLT	return
+	MOVB	-(A1), -(A2)
+	BRA	b1
+
+/*
+ * check that dest is alligned
+ * if not, just go byte-at-a-time
+ */
+b2:
+	MOVL	A2, R1
+	ANDL	$3, R1
+	BEQ	b3
+	SUBL	$1, R0
+	BLT	return
+	BRA	b5
+/*
+ * quad-long-at-a-time backward copy
+ */
+b3:
+	SUBL	$16, R0
+	BLT	b4
+	MOVL	-(A1), -(A2)
+	MOVL	-(A1), -(A2)
+	MOVL	-(A1), -(A2)
+	MOVL	-(A1), -(A2)
+	BRA	b3
+
+/*
+ * cleanup byte-at-a-time backward
+ */
+b4:
+	ADDL	$15, R0
+	BLT	return
+b5:
+	MOVB	-(A1), -(A2)
+	SUBL	$1, R0
+	BGE	b5
+	BRA	return
diff --git a/libkern/memmove-arm.s b/libkern/memmove-arm.s
new file mode 100644
index 00000000..8e28200a
--- /dev/null
+++ b/libkern/memmove-arm.s
@@ -0,0 +1,223 @@
+TS = 0
+TE = 1
+FROM = 2
+N = 3
+TMP = 3					/* N and TMP don't overlap */
+TMP1 = 4
+
+TEXT	memcpy(SB), $0
+TEXT memmove(SB), $-4
+_memmove:
+	MOVW	R(TS), to+0(FP)		/* need to save for return value */
+	MOVW	from+4(FP), R(FROM)
+	MOVW	n+8(FP), R(N)
+
+	ADD	R(N), R(TS), R(TE)	/* to end pointer */
+
+	CMP	R(FROM), R(TS)
+	BLS	_forward
+
+_back:
+	ADD	R(N), R(FROM)		/* from end pointer */
+	CMP	$4, R(N)		/* need at least 4 bytes to copy */
+	BLT	_b1tail
+
+_b4align:				/* align destination on 4 */
+	AND.S	$3, R(TE), R(TMP)
+	BEQ	_b4aligned
+
+	MOVBU.W	-1(R(FROM)), R(TMP)	/* pre-indexed */
+	MOVBU.W	R(TMP), -1(R(TE))	/* pre-indexed */
+	B	_b4align
+
+_b4aligned:				/* is source now aligned? */
+	AND.S	$3, R(FROM), R(TMP)
+	BNE	_bunaligned
+
+	ADD	$31, R(TS), R(TMP)	/* do 32-byte chunks if possible */
+_b32loop:
+	CMP	R(TMP), R(TE)
+	BLS	_b4tail
+
+	MOVM.DB.W (R(FROM)), [R4-R11]
+	MOVM.DB.W [R4-R11], (R(TE))
+	B	_b32loop
+
+_b4tail:				/* do remaining words if possible */
+	ADD	$3, R(TS), R(TMP)
+_b4loop:
+	CMP	R(TMP), R(TE)
+	BLS	_b1tail
+
+	MOVW.W	-4(R(FROM)), R(TMP1)	/* pre-indexed */
+	MOVW.W	R(TMP1), -4(R(TE))	/* pre-indexed */
+	B	_b4loop
+
+_b1tail:				/* remaining bytes */
+	CMP	R(TE), R(TS)
+	BEQ	_return
+
+	MOVBU.W	-1(R(FROM)), R(TMP)	/* pre-indexed */
+	MOVBU.W	R(TMP), -1(R(TE))	/* pre-indexed */
+	B	_b1tail
+
+_forward:
+	CMP	$4, R(N)		/* need at least 4 bytes to copy */
+	BLT	_f1tail
+
+_f4align:				/* align destination on 4 */
+	AND.S	$3, R(TS), R(TMP)
+	BEQ	_f4aligned
+
+	MOVBU.P	1(R(FROM)), R(TMP)	/* implicit write back */
+	MOVBU.P	R(TMP), 1(R(TS))	/* implicit write back */
+	B	_f4align
+
+_f4aligned:				/* is source now aligned? */
+	AND.S	$3, R(FROM), R(TMP)
+	BNE	_funaligned
+
+	SUB	$31, R(TE), R(TMP)	/* do 32-byte chunks if possible */
+_f32loop:
+	CMP	R(TMP), R(TS)
+	BHS	_f4tail
+
+	MOVM.IA.W (R(FROM)), [R4-R11] 
+	MOVM.IA.W [R4-R11], (R(TS))
+	B	_f32loop
+
+_f4tail:
+	SUB	$3, R(TE), R(TMP)	/* do remaining words if possible */
+_f4loop:
+	CMP	R(TMP), R(TS)
+	BHS	_f1tail
+
+	MOVW.P	4(R(FROM)), R(TMP1)	/* implicit write back */
+	MOVW.P	R4, 4(R(TS))		/* implicit write back */
+	B	_f4loop
+
+_f1tail:
+	CMP	R(TS), R(TE)
+	BEQ	_return
+
+	MOVBU.P	1(R(FROM)), R(TMP)	/* implicit write back */
+	MOVBU.P	R(TMP), 1(R(TS))	/* implicit write back */
+	B	_f1tail
+
+_return:
+	MOVW	to+0(FP), R0
+	RET
+
+RSHIFT = 4
+LSHIFT = 5
+OFFSET = 6
+
+BR0 = 7
+BW0 = 8
+BR1 = 8
+BW1 = 9
+BR2 = 9
+BW2 = 10
+BR3 = 10
+BW3 = 11
+
+_bunaligned:
+	CMP	$2, R(TMP)		/* is R(TMP) < 2 ? */
+
+	MOVW.LT	$8, R(RSHIFT)		/* (R(n)<<24)|(R(n-1)>>8) */
+	MOVW.LT	$24, R(LSHIFT)
+	MOVW.LT	$1, R(OFFSET)
+
+	MOVW.EQ	$16, R(RSHIFT)		/* (R(n)<<16)|(R(n-1)>>16) */
+	MOVW.EQ	$16, R(LSHIFT)
+	MOVW.EQ	$2, R(OFFSET)
+
+	MOVW.GT	$24, R(RSHIFT)		/* (R(n)<<8)|(R(n-1)>>24) */
+	MOVW.GT	$8, R(LSHIFT)
+	MOVW.GT	$3, R(OFFSET)
+
+	ADD	$16, R(TS), R(TMP)	/* do 16-byte chunks if possible */
+	CMP	R(TMP), R(TE)
+	BLS	_b1tail
+
+	AND	$~0x03, R(FROM)		/* align source */
+	MOVW	(R(FROM)), R(BR0)	/* prime first block register */
+
+_bu16loop:
+	CMP	R(TMP), R(TE)
+	BLS	_bu1tail
+
+	MOVW	R(BR0)<<R(LSHIFT), R(BW3)
+	MOVM.DB.W (R(FROM)), [R(BR0)-R(BR3)]
+	ORR	R(BR3)>>R(RSHIFT), R(BW3)
+
+	MOVW	R(BR3)<<R(LSHIFT), R(BW2)
+	ORR	R(BR2)>>R(RSHIFT), R(BW2)
+
+	MOVW	R(BR2)<<R(LSHIFT), R(BW1)
+	ORR	R(BR1)>>R(RSHIFT), R(BW1)
+
+	MOVW	R(BR1)<<R(LSHIFT), R(BW0)
+	ORR	R(BR0)>>R(RSHIFT), R(BW0)
+
+	MOVM.DB.W [R(BW0)-R(BW3)], (R(TE))
+	B	_bu16loop
+
+_bu1tail:
+	ADD	R(OFFSET), R(FROM)
+	B	_b1tail
+
+FW0 = 7
+FR0 = 8
+FW1 = 8
+FR1 = 9
+FW2 = 9
+FR2 = 10
+FW3 = 10
+FR3 = 11
+
+_funaligned:
+	CMP	$2, R(TMP)
+
+	MOVW.LT	$8, R(RSHIFT)		/* (R(n+1)<<24)|(R(n)>>8) */
+	MOVW.LT	$24, R(LSHIFT)
+	MOVW.LT	$3, R(OFFSET)
+
+	MOVW.EQ	$16, R(RSHIFT)		/* (R(n+1)<<16)|(R(n)>>16) */
+	MOVW.EQ	$16, R(LSHIFT)
+	MOVW.EQ	$2, R(OFFSET)
+
+	MOVW.GT	$24, R(RSHIFT)		/* (R(n+1)<<8)|(R(n)>>24) */
+	MOVW.GT	$8, R(LSHIFT)
+	MOVW.GT	$1, R(OFFSET)
+
+	SUB	$16, R(TE), R(TMP)	/* do 16-byte chunks if possible */
+	CMP	R(TMP), R(TS)
+	BHS	_f1tail
+
+	AND	$~0x03, R(FROM)		/* align source */
+	MOVW.P	4(R(FROM)), R(FR3)	/* prime last block register, implicit write back */
+
+_fu16loop:
+	CMP	R(TMP), R(TS)
+	BHS	_fu1tail
+
+	MOVW	R(FR3)>>R(RSHIFT), R(FW0)
+	MOVM.IA.W (R(FROM)), [R(FR0)-R(FR3)]
+	ORR	R(FR0)<<R(LSHIFT), R(FW0)
+
+	MOVW	R(FR0)>>R(RSHIFT), R(FW1)
+	ORR	R(FR1)<<R(LSHIFT), R(FW1)
+
+	MOVW	R(FR1)>>R(RSHIFT), R(FW2)
+	ORR	R(FR2)<<R(LSHIFT), R(FW2)
+
+	MOVW	R(FR2)>>R(RSHIFT), R(FW3)
+	ORR	R(FR3)<<R(LSHIFT), R(FW3)
+
+	MOVM.IA.W [R(FW0)-R(FW3)], (R(TS))
+	B	_fu16loop
+
+_fu1tail:
+	SUB	R(OFFSET), R(FROM)
+	B	_f1tail
diff --git a/libkern/memmove-mips.s b/libkern/memmove-mips.s
new file mode 100644
index 00000000..c14d2a82
--- /dev/null
+++ b/libkern/memmove-mips.s
@@ -0,0 +1,237 @@
+	TEXT	memmove(SB), $0
+
+	JMP	move
+
+	TEXT	memcpy(SB), $0
+move:
+	MOVW	R1, s1+0(FP)
+
+	MOVW	n+8(FP), R3		/* R3 is count */
+	MOVW	R1, R4			/* R4 is to-pointer */
+	SGT	R0, R3, R5
+	BEQ	R5, ok
+	MOVW	(R0), R0		/* abort if negative count */
+ok:
+	MOVW	s2+4(FP), R5		/* R5 is from-pointer */
+	ADDU	R3,R5, R7		/* R7 is end from-pointer */
+	ADDU	R3,R4, R6		/* R6 is end to-pointer */
+
+/*
+ * easiest test is copy backwards if
+ * destination string has higher mem address
+ */
+	SGT	$4,R3, R2
+	SGTU	R4,R5, R1
+	BNE	R1, back
+
+/*
+ * if not at least 4 chars,
+ * don't even mess around.
+ * 3 chars to guarantee any
+ * rounding up to a word
+ * boundary and 4 characters
+ * to get at least maybe one
+ * full word store.
+ */
+	BNE	R2, fout
+
+
+/*
+ * byte at a time to word align destination
+ */
+f1:
+	AND	$3,R4, R1
+	BEQ	R1, f2
+	MOVB	0(R5), R8
+	ADDU	$1, R5
+	MOVB	R8, 0(R4)
+	ADDU	$1, R4
+	JMP	f1
+
+/*
+ * test if source is now word aligned
+ */
+f2:
+	AND	$3, R5, R1
+	BNE	R1, fun2
+/*
+ * turn R3 into to-end pointer-15
+ * copy 16 at a time while theres room.
+ * R6 is smaller than R7 --
+ * there are problems if R7 is 0.
+ */
+	ADDU	$-15,R6, R3
+f3:
+	SGTU	R3,R4, R1
+	BEQ	R1, f4
+	MOVW	0(R5), R8
+	MOVW	4(R5), R9
+	MOVW	R8, 0(R4)
+	MOVW	8(R5), R8
+	MOVW	R9, 4(R4)
+	MOVW	12(R5), R9
+	ADDU	$16, R5
+	MOVW	R8, 8(R4)
+	MOVW	R9, 12(R4)
+	ADDU	$16, R4
+	JMP	f3
+
+/*
+ * turn R3 into to-end pointer-3
+ * copy 4 at a time while theres room
+ */
+f4:
+	ADDU	$-3,R6, R3
+f5:
+	SGTU	R3,R4, R1
+	BEQ	R1, fout
+	MOVW	0(R5), R8
+	ADDU	$4, R5
+	MOVW	R8, 0(R4)
+	ADDU	$4, R4
+	JMP	f5
+
+/*
+ * forward copy, unaligned
+ * turn R3 into to-end pointer-15
+ * copy 16 at a time while theres room.
+ * R6 is smaller than R7 --
+ * there are problems if R7 is 0.
+ */
+fun2:
+	ADDU	$-15,R6, R3
+fun3:
+	SGTU	R3,R4, R1
+	BEQ	R1, fun4
+	MOVWL	0(R5), R8
+	MOVWR	3(R5), R8
+	MOVWL	4(R5), R9
+	MOVWR	7(R5), R9
+	MOVW	R8, 0(R4)
+	MOVWL	8(R5), R8
+	MOVWR	11(R5), R8
+	MOVW	R9, 4(R4)
+	MOVWL	12(R5), R9
+	MOVWR	15(R5), R9
+	ADDU	$16, R5
+	MOVW	R8, 8(R4)
+	MOVW	R9, 12(R4)
+	ADDU	$16, R4
+	JMP	fun3
+
+/*
+ * turn R3 into to-end pointer-3
+ * copy 4 at a time while theres room
+ */
+fun4:
+	ADDU	$-3,R6, R3
+fun5:
+	SGTU	R3,R4, R1
+	BEQ	R1, fout
+	MOVWL	0(R5), R8
+	MOVWR	3(R5), R8
+	ADDU	$4, R5
+	MOVW	R8, 0(R4)
+	ADDU	$4, R4
+	JMP	fun5
+
+/*
+ * last loop, copy byte at a time
+ */
+fout:
+	BEQ	R7,R5, ret
+	MOVB	0(R5), R8
+	ADDU	$1, R5
+	MOVB	R8, 0(R4)
+	ADDU	$1, R4
+	JMP	fout
+
+/*
+ * whole thing repeated for backwards
+ */
+back:
+	BNE	R2, bout
+b1:
+	AND	$3,R6, R1
+	BEQ	R1, b2
+	MOVB	-1(R7), R8
+	ADDU	$-1, R7
+	MOVB	R8, -1(R6)
+	ADDU	$-1, R6
+	JMP	b1
+
+b2:
+	AND	$3, R7, R1
+	BNE	R1, bun2
+
+	ADDU	$15,R5, R3
+b3:
+	SGTU	R7,R3, R1
+	BEQ	R1, b4
+	MOVW	-4(R7), R8
+	MOVW	-8(R7), R9
+	MOVW	R8, -4(R6)
+	MOVW	-12(R7), R8
+	MOVW	R9, -8(R6)
+	MOVW	-16(R7), R9
+	ADDU	$-16, R7
+	MOVW	R8, -12(R6)
+	MOVW	R9, -16(R6)
+	ADDU	$-16, R6
+	JMP	b3
+b4:
+	ADDU	$3,R5, R3
+b5:
+	SGTU	R7,R3, R1
+	BEQ	R1, bout
+	MOVW	-4(R7), R8
+	ADDU	$-4, R7
+	MOVW	R8, -4(R6)
+	ADDU	$-4, R6
+	JMP	b5
+
+bun2:
+	ADDU	$15,R5, R3
+bun3:
+	SGTU	R7,R3, R1
+	BEQ	R1, bun4
+	MOVWL	-4(R7), R8
+	MOVWR	-1(R7), R8
+	MOVWL	-8(R7), R9
+	MOVWR	-5(R7), R9
+	MOVW	R8, -4(R6)
+	MOVWL	-12(R7), R8
+	MOVWR	-9(R7), R8
+	MOVW	R9, -8(R6)
+	MOVWL	-16(R7), R9
+	MOVWR	-13(R7), R9
+	ADDU	$-16, R7
+	MOVW	R8, -12(R6)
+	MOVW	R9, -16(R6)
+	ADDU	$-16, R6
+	JMP	bun3
+
+bun4:
+	ADDU	$3,R5, R3
+bun5:
+	SGTU	R7,R3, R1
+	BEQ	R1, bout
+	MOVWL	-4(R7), R8
+	MOVWR	-1(R7), R8
+	ADDU	$-4, R7
+	MOVW	R8, -4(R6)
+	ADDU	$-4, R6
+	JMP	bun5
+
+bout:
+	BEQ	R7,R5, ret
+	MOVB	-1(R7), R8
+	ADDU	$-1, R7
+	MOVB	R8, -1(R6)
+	ADDU	$-1, R6
+	JMP	bout
+
+ret:
+	MOVW	s1+0(FP), R1
+	RET
+	END
diff --git a/libkern/memmove-power.s b/libkern/memmove-power.s
new file mode 100644
index 00000000..1b27e5d6
--- /dev/null
+++ b/libkern/memmove-power.s
@@ -0,0 +1,170 @@
+#define	BDNZ	BC	16,0,
+	TEXT	memcpy(SB), $0
+	BR	move
+
+	TEXT	memmove(SB), $0
+move:
+
+/*
+ * performance:
+ * (tba)
+ */
+
+	MOVW	R3, s1+0(FP)
+	MOVW	n+8(FP), R9		/* R9 is count */
+	MOVW	R3, R10			/* R10 is to-pointer */
+	CMP	R9, $0
+	BEQ	ret
+	BLT	trap
+	MOVW	s2+4(FP), R11		/* R11 is from-pointer */
+
+/*
+ * if no more than 16 bytes, just use one lsw/stsw
+ */
+	CMP	R9, $16
+	BLE	fout
+
+	ADD	R9,R11, R13		/* R13 is end from-pointer */
+	ADD	R9,R10, R12		/* R12 is end to-pointer */
+
+/*
+ * easiest test is copy backwards if
+ * destination string has higher mem address
+ */
+	CMPU	R10, R11
+	BGT	back
+
+/*
+ * test if both pointers
+ * are similarly word aligned
+ */
+	XOR	R10,R11, R7
+	ANDCC	$3,R7
+	BNE	fbad
+
+/*
+ * move a few bytes to align pointers
+ */
+	ANDCC	$3,R10,R7
+	BEQ	f2
+	SUBC	R7, $4, R7
+	SUB	R7, R9
+	MOVW	R7, XER
+	LSW	(R11), R16
+	ADD	R7, R11
+	STSW	R16, (R10)
+	ADD	R7, R10
+
+/*
+ * turn R14 into doubleword count
+ * copy 16 bytes at a time while there's room.
+ */
+f2:
+	SRAWCC	$4, R9, R14
+	BLE	fout
+	MOVW	R14, CTR
+	SUB	$4, R11
+	SUB	$4, R10
+f3:
+	MOVWU	4(R11), R16
+	MOVWU	R16, 4(R10)
+	MOVWU	4(R11), R17
+	MOVWU	R17, 4(R10)
+	MOVWU	4(R11), R16
+	MOVWU	R16, 4(R10)
+	MOVWU	4(R11), R17
+	MOVWU	R17, 4(R10)
+	BDNZ	f3
+	RLWNMCC	$0, R9, $15, R9	/* residue */
+	BEQ	ret
+	ADD	$4, R11
+	ADD	$4, R10
+
+/*
+ * move up to 16 bytes through R16 .. R19; aligned and unaligned
+ */
+fout:
+	MOVW	R9, XER
+	LSW	(R11), R16
+	STSW	R16, (R10)
+	BR	ret
+
+/*
+ * loop for unaligned copy, then copy up to 15 remaining bytes
+ */
+fbad:
+	SRAWCC	$4, R9, R14
+	BLE	f6
+	MOVW	R14, CTR
+f5:
+	LSW	(R11), $16, R16
+	ADD	$16, R11
+	STSW	R16, $16, (R10)
+	ADD	$16, R10
+	BDNZ	f5
+	RLWNMCC	$0, R9, $15, R9	/* residue */
+	BEQ	ret
+f6:
+	MOVW	R9, XER
+	LSW	(R11), R16
+	STSW	R16, (R10)
+	BR	ret
+
+/*
+ * whole thing repeated for backwards
+ */
+back:
+	CMP	R9, $4
+	BLT	bout
+
+	XOR	R12,R13, R7
+	ANDCC	$3,R7
+	BNE	bout
+b1:
+	ANDCC	$3,R13, R7
+	BEQ	b2
+	MOVBZU	-1(R13), R16
+	MOVBZU	R16, -1(R12)
+	SUB	$1, R9
+	BR	b1
+b2:
+	SRAWCC	$4, R9, R14
+	BLE	b4
+	MOVW	R14, CTR
+b3:
+	MOVWU	-4(R13), R16
+	MOVWU	R16, -4(R12)
+	MOVWU	-4(R13), R17
+	MOVWU	R17, -4(R12)
+	MOVWU	-4(R13), R16
+	MOVWU	R16, -4(R12)
+	MOVWU	-4(R13), R17
+	MOVWU	R17, -4(R12)
+	BDNZ	b3
+	RLWNMCC	$0, R9, $15, R9	/* residue */
+	BEQ	ret
+b4:
+	SRAWCC	$2, R9, R14
+	BLE	bout
+	MOVW	R14, CTR
+b5:
+	MOVWU	-4(R13), R16
+	MOVWU	R16, -4(R12)
+	BDNZ	b5
+	RLWNMCC	$0, R9, $3, R9	/* residue */
+	BEQ	ret
+
+bout:
+	CMPU	R13, R11
+	BLE	ret
+	MOVBZU	-1(R13), R16
+	MOVBZU	R16, -1(R12)
+	BR	bout
+
+trap:
+/*	MOVW	$0, R0	*/
+	MOVW	R0, 0(R0)
+
+ret:
+	MOVW	s1+0(FP), R3
+	RETURN
diff --git a/libkern/memmove-sparc.s b/libkern/memmove-sparc.s
new file mode 100644
index 00000000..1295209a
--- /dev/null
+++ b/libkern/memmove-sparc.s
@@ -0,0 +1,162 @@
+	TEXT	memmove(SB), $0
+	JMP	move
+
+	TEXT	memcpy(SB), $0
+move:
+
+/*
+ * performance:
+ * (tba)
+ */
+
+	MOVW	R7, s1+0(FP)
+	MOVW	n+8(FP), R9		/* R9 is count */
+	MOVW	R7, R10			/* R10 is to-pointer */
+	SUBCC	R0,R9, R0
+	BGE	ok
+	MOVW	0(R0), R0
+
+ok:
+	MOVW	s2+4(FP), R11		/* R11 is from-pointer */
+	ADD	R9,R11, R13		/* R13 is end from-pointer */
+	ADD	R9,R10, R12		/* R12 is end to-pointer */
+
+/*
+ * easiest test is copy backwards if
+ * destination string has higher mem address
+ */
+	SUBCC	R11,R10, R0
+	BGU	back
+
+/*
+ * if not at least 8 chars,
+ * dont even mess around.
+ * 7 chars to guarantee any
+ * rounding up to a word
+ * boundary and 8 characters
+ * to get at least maybe one
+ * full word store.
+ */
+	SUBCC	$8,R9, R0
+	BL	fout
+
+/*
+ * test if both pointers
+ * are similarly word aligned
+ */
+	XOR	R10,R11, R7
+	ANDCC	$7,R7, R0
+	BNE	fout
+
+/*
+ * byte at a time to double align
+ */
+f1:
+	ANDCC	$7,R10, R0
+	BE	f2
+	MOVB	0(R11), R16
+	ADD	$1, R11
+	MOVB	R16, 0(R10)
+	ADD	$1, R10
+	JMP	f1
+
+/*
+ * turn R9 into to-end pointer-15
+ * copy 16 at a time while theres room.
+ * R12 is smaller than R13 --
+ * there are problems if R13 is 0.
+ */
+f2:
+	SUB	$15,R12, R9
+f3:
+	SUBCC	R10,R9, R0
+	BLEU	f4
+	MOVD	0(R11), R16
+	MOVD	R16, 0(R10)
+	MOVD	8(R11), R16
+	ADD	$16, R11
+	MOVD	R16, 8(R10)
+	ADD	$16, R10
+	JMP	f3
+
+/*
+ * turn R9 into to-end pointer-3
+ * copy 4 at a time while theres room
+ */
+f4:
+	SUB	$3,R12, R9
+f5:
+	SUBCC	R10,R9, R0
+	BLEU	fout
+	MOVW	0(R11), R16
+	ADD	$4, R11
+	MOVW	R16, 0(R10)
+	ADD	$4, R10
+	JMP	f5
+
+/*
+ * last loop, copy byte at a time
+ */
+fout:
+	SUBCC	R11,R13, R0
+	BLEU	ret
+	MOVB	0(R11), R16
+	ADD	$1, R11
+	MOVB	R16, 0(R10)
+	ADD	$1, R10
+	JMP	fout
+
+/*
+ * whole thing repeated for backwards
+ */
+back:
+	SUBCC	$8,R9, R0 
+	BL	bout
+
+	XOR	R12,R13, R7
+	ANDCC	$7,R7, R0
+	BNE	bout
+b1:
+	ANDCC	$7,R13, R0
+	BE	b2
+	MOVB	-1(R13), R16
+	SUB	$1, R13
+	MOVB	R16, -1(R12)
+	SUB	$1, R12
+	JMP	b1
+b2:
+	ADD	$15,R11, R9
+b3:
+	SUBCC	R9,R13, R0
+	BLEU	b4
+
+	MOVD	-8(R13), R16
+	MOVD	R16, -8(R12)
+	MOVD	-16(R13), R16
+	SUB	$16, R13
+	MOVD	R16, -16(R12);
+	SUB	$16, R12
+	JMP	b3
+b4:
+	ADD	$3,R11, R9
+b5:
+	SUBCC	R9,R13, R0
+	BLEU	bout
+	MOVW	-4(R13), R16
+	SUB	$4, R13
+	MOVW	R16, -4(R12)
+	SUB	$4, R12
+	JMP	b5
+
+bout:
+	SUBCC	R11,R13, R0
+	BLEU	ret
+	MOVB	-1(R13), R16
+	SUB	$1, R13
+	MOVB	R16, -1(R12)
+	SUB	$1, R12
+	JMP	bout
+
+ret:
+	MOVW	s1+0(FP), R7
+	RETURN
diff --git a/libkern/memmove-thumb.s b/libkern/memmove-thumb.s
new file mode 100644
index 00000000..8e28200a
--- /dev/null
+++ b/libkern/memmove-thumb.s
@@ -0,0 +1,223 @@
+TS = 0
+TE = 1
+FROM = 2
+N = 3
+TMP = 3					/* N and TMP don't overlap */
+TMP1 = 4
+
+TEXT	memcpy(SB), $0
+TEXT memmove(SB), $-4
+_memmove:
+	MOVW	R(TS), to+0(FP)		/* need to save for return value */
+	MOVW	from+4(FP), R(FROM)
+	MOVW	n+8(FP), R(N)
+
+	ADD	R(N), R(TS), R(TE)	/* to end pointer */
+
+	CMP	R(FROM), R(TS)
+	BLS	_forward
+
+_back:
+	ADD	R(N), R(FROM)		/* from end pointer */
+	CMP	$4, R(N)		/* need at least 4 bytes to copy */
+	BLT	_b1tail
+
+_b4align:				/* align destination on 4 */
+	AND.S	$3, R(TE), R(TMP)
+	BEQ	_b4aligned
+
+	MOVBU.W	-1(R(FROM)), R(TMP)	/* pre-indexed */
+	MOVBU.W	R(TMP), -1(R(TE))	/* pre-indexed */
+	B	_b4align
+
+_b4aligned:				/* is source now aligned? */
+	AND.S	$3, R(FROM), R(TMP)
+	BNE	_bunaligned
+
+	ADD	$31, R(TS), R(TMP)	/* do 32-byte chunks if possible */
+_b32loop:
+	CMP	R(TMP), R(TE)
+	BLS	_b4tail
+
+	MOVM.DB.W (R(FROM)), [R4-R11]
+	MOVM.DB.W [R4-R11], (R(TE))
+	B	_b32loop
+
+_b4tail:				/* do remaining words if possible */
+	ADD	$3, R(TS), R(TMP)
+_b4loop:
+	CMP	R(TMP), R(TE)
+	BLS	_b1tail
+
+	MOVW.W	-4(R(FROM)), R(TMP1)	/* pre-indexed */
+	MOVW.W	R(TMP1), -4(R(TE))	/* pre-indexed */
+	B	_b4loop
+
+_b1tail:				/* remaining bytes */
+	CMP	R(TE), R(TS)
+	BEQ	_return
+
+	MOVBU.W	-1(R(FROM)), R(TMP)	/* pre-indexed */
+	MOVBU.W	R(TMP), -1(R(TE))	/* pre-indexed */
+	B	_b1tail
+
+_forward:
+	CMP	$4, R(N)		/* need at least 4 bytes to copy */
+	BLT	_f1tail
+
+_f4align:				/* align destination on 4 */
+	AND.S	$3, R(TS), R(TMP)
+	BEQ	_f4aligned
+
+	MOVBU.P	1(R(FROM)), R(TMP)	/* implicit write back */
+	MOVBU.P	R(TMP), 1(R(TS))	/* implicit write back */
+	B	_f4align
+
+_f4aligned:				/* is source now aligned? */
+	AND.S	$3, R(FROM), R(TMP)
+	BNE	_funaligned
+
+	SUB	$31, R(TE), R(TMP)	/* do 32-byte chunks if possible */
+_f32loop:
+	CMP	R(TMP), R(TS)
+	BHS	_f4tail
+
+	MOVM.IA.W (R(FROM)), [R4-R11] 
+	MOVM.IA.W [R4-R11], (R(TS))
+	B	_f32loop
+
+_f4tail:
+	SUB	$3, R(TE), R(TMP)	/* do remaining words if possible */
+_f4loop:
+	CMP	R(TMP), R(TS)
+	BHS	_f1tail
+
+	MOVW.P	4(R(FROM)), R(TMP1)	/* implicit write back */
+	MOVW.P	R4, 4(R(TS))		/* implicit write back */
+	B	_f4loop
+
+_f1tail:
+	CMP	R(TS), R(TE)
+	BEQ	_return
+
+	MOVBU.P	1(R(FROM)), R(TMP)	/* implicit write back */
+	MOVBU.P	R(TMP), 1(R(TS))	/* implicit write back */
+	B	_f1tail
+
+_return:
+	MOVW	to+0(FP), R0
+	RET
+
+RSHIFT = 4
+LSHIFT = 5
+OFFSET = 6
+
+BR0 = 7
+BW0 = 8
+BR1 = 8
+BW1 = 9
+BR2 = 9
+BW2 = 10
+BR3 = 10
+BW3 = 11
+
+_bunaligned:
+	CMP	$2, R(TMP)		/* is R(TMP) < 2 ? */
+
+	MOVW.LT	$8, R(RSHIFT)		/* (R(n)<<24)|(R(n-1)>>8) */
+	MOVW.LT	$24, R(LSHIFT)
+	MOVW.LT	$1, R(OFFSET)
+
+	MOVW.EQ	$16, R(RSHIFT)		/* (R(n)<<16)|(R(n-1)>>16) */
+	MOVW.EQ	$16, R(LSHIFT)
+	MOVW.EQ	$2, R(OFFSET)
+
+	MOVW.GT	$24, R(RSHIFT)		/* (R(n)<<8)|(R(n-1)>>24) */
+	MOVW.GT	$8, R(LSHIFT)
+	MOVW.GT	$3, R(OFFSET)
+
+	ADD	$16, R(TS), R(TMP)	/* do 16-byte chunks if possible */
+	CMP	R(TMP), R(TE)
+	BLS	_b1tail
+
+	AND	$~0x03, R(FROM)		/* align source */
+	MOVW	(R(FROM)), R(BR0)	/* prime first block register */
+
+_bu16loop:
+	CMP	R(TMP), R(TE)
+	BLS	_bu1tail
+
+	MOVW	R(BR0)<<R(LSHIFT), R(BW3)
+	MOVM.DB.W (R(FROM)), [R(BR0)-R(BR3)]
+	ORR	R(BR3)>>R(RSHIFT), R(BW3)
+
+	MOVW	R(BR3)<<R(LSHIFT), R(BW2)
+	ORR	R(BR2)>>R(RSHIFT), R(BW2)
+
+	MOVW	R(BR2)<<R(LSHIFT), R(BW1)
+	ORR	R(BR1)>>R(RSHIFT), R(BW1)
+
+	MOVW	R(BR1)<<R(LSHIFT), R(BW0)
+	ORR	R(BR0)>>R(RSHIFT), R(BW0)
+
+	MOVM.DB.W [R(BW0)-R(BW3)], (R(TE))
+	B	_bu16loop
+
+_bu1tail:
+	ADD	R(OFFSET), R(FROM)
+	B	_b1tail
+
+FW0 = 7
+FR0 = 8
+FW1 = 8
+FR1 = 9
+FW2 = 9
+FR2 = 10
+FW3 = 10
+FR3 = 11
+
+_funaligned:
+	CMP	$2, R(TMP)
+
+	MOVW.LT	$8, R(RSHIFT)		/* (R(n+1)<<24)|(R(n)>>8) */
+	MOVW.LT	$24, R(LSHIFT)
+	MOVW.LT	$3, R(OFFSET)
+
+	MOVW.EQ	$16, R(RSHIFT)		/* (R(n+1)<<16)|(R(n)>>16) */
+	MOVW.EQ	$16, R(LSHIFT)
+	MOVW.EQ	$2, R(OFFSET)
+
+	MOVW.GT	$24, R(RSHIFT)		/* (R(n+1)<<8)|(R(n)>>24) */
+	MOVW.GT	$8, R(LSHIFT)
+	MOVW.GT	$1, R(OFFSET)
+
+	SUB	$16, R(TE), R(TMP)	/* do 16-byte chunks if possible */
+	CMP	R(TMP), R(TS)
+	BHS	_f1tail
+
+	AND	$~0x03, R(FROM)		/* align source */
+	MOVW.P	4(R(FROM)), R(FR3)	/* prime last block register, implicit write back */
+
+_fu16loop:
+	CMP	R(TMP), R(TS)
+	BHS	_fu1tail
+
+	MOVW	R(FR3)>>R(RSHIFT), R(FW0)
+	MOVM.IA.W (R(FROM)), [R(FR0)-R(FR3)]
+	ORR	R(FR0)<<R(LSHIFT), R(FW0)
+
+	MOVW	R(FR0)>>R(RSHIFT), R(FW1)
+	ORR	R(FR1)<<R(LSHIFT), R(FW1)
+
+	MOVW	R(FR1)>>R(RSHIFT), R(FW2)
+	ORR	R(FR2)<<R(LSHIFT), R(FW2)
+
+	MOVW	R(FR2)>>R(RSHIFT), R(FW3)
+	ORR	R(FR3)<<R(LSHIFT), R(FW3)
+
+	MOVM.IA.W [R(FW0)-R(FW3)], (R(TS))
+	B	_fu16loop
+
+_fu1tail:
+	SUB	R(OFFSET), R(FROM)
+	B	_f1tail
diff --git a/libkern/memmove.c b/libkern/memmove.c
new file mode 100644
index 00000000..a1455f51
--- /dev/null
+++ b/libkern/memmove.c
@@ -0,0 +1,43 @@
+#include	<lib9.h>
+
+/* for testing only */
+void*
+memcpy(void *a1, void *a2, ulong n)
+{
+	return memmove(a1, a2, n);
+}
+
+void*
+memmove(void *a1, void *a2, ulong n)
+{
+	int m = (int)n;
+	uchar *s, *d;
+	
+	d = a1;
+	s = a2;
+	if(d > s){
+		s += m;
+		d += m;
+		while(--m >= 0)
+			*--d = *--s;
+	}
+	else{
+		while(--m >= 0)
+			*d++ = *s++;
+	}
+	return a1;
+}
+
+/*
+void
+memset(void *a1, int c, ulong n)
+{
+	int m = (int)n;
+	uchar *d;
+
+	d = a1;
+	while(--m >= 0)
+		*d++ = c;
+}
+*/
+	
diff --git a/libkern/memset-386.s b/libkern/memset-386.s
new file mode 100644
index 00000000..dacb2f5f
--- /dev/null
+++ b/libkern/memset-386.s
@@ -0,0 +1,35 @@
+	TEXT	memset(SB),$0
+
+	CLD
+	MOVL	p+0(FP), DI
+	MOVBLZX	c+4(FP), AX
+	MOVL	n+8(FP), BX
+/*
+ * if not enough bytes, just copy
+ */
+	CMPL	BX, $9
+	JLS	c3
+/*
+ * build word in AX
+ */
+	MOVB	AL, AH
+	MOVL	AX, CX
+	SHLL	$16, CX
+	ORL	CX, AX
+/*
+ * copy whole longs
+ */
+c1:
+	MOVL	BX, CX
+	SHRL	$2, CX
+	ANDL	$3, BX
+	REP;	STOSL
+/*
+ * copy the rest, by bytes
+ */
+c3:
+	MOVL	BX, CX
+	REP;	STOSB
+ret:
+	MOVL	p+0(FP),AX
+	RET
diff --git a/libkern/memset-68000.s b/libkern/memset-68000.s
new file mode 100644
index 00000000..318f61a7
--- /dev/null
+++ b/libkern/memset-68000.s
@@ -0,0 +1,57 @@
+	TEXT	memset(SB), $0
+	MOVL	n+8(FP), R0
+	BLE	return
+	MOVL	s1+0(FP), A1
+	CLRL	R1
+	MOVB	c+7(FP), R1
+	BEQ	l1
+
+/*
+ * create 4 replicated copies
+ * of the byte in R1
+ */
+	MOVL	R1, R2
+	ASLL	$8, R2
+	ORL	R2, R1
+	MOVL	R1, R2
+	SWAP	R2
+	ORL	R2, R1
+
+/*
+ * byte-at-a-time until alligned
+ */
+l1:
+	MOVL	A1, R1
+	ANDL	$3, R1
+	BEQ	l2
+	SUBL	$1, R0
+	BLT	return
+	MOVB	R1, (A1)+
+	BRA	l1
+
+/*
+ * quad-long-at-a-time set
+ */
+l2:
+	SUBL	$16, R0
+	BLT	l3
+	MOVL	R1, (A1)+
+	MOVL	R1, (A1)+
+	MOVL	R1, (A1)+
+	MOVL	R1, (A1)+
+	BRA	l2
+
+/*
+ * cleanup byte-at-a-time
+ */
+l3:
+	ADDL	$15, R0
+	BLT	return
+l4:
+	MOVB	R1, (A1)+
+	SUBL	$1, R0
+	BGE	l4
+
+return:
+	MOVL	s1+0(FP),R0
+	RTS
diff --git a/libkern/memset-arm.s b/libkern/memset-arm.s
new file mode 100644
index 00000000..92355c11
--- /dev/null
+++ b/libkern/memset-arm.s
@@ -0,0 +1,66 @@
+TO = 1
+TOE = 2
+N = 3
+TMP = 3					/* N and TMP don't overlap */
+
+TEXT memset(SB), $0
+	MOVW	R0, R(TO)
+	MOVW	data+4(FP), R(4)
+	MOVW	n+8(FP), R(N)
+
+	ADD	R(N), R(TO), R(TOE)	/* to end pointer */
+
+	CMP	$4, R(N)		/* need at least 4 bytes to copy */
+	BLT	_1tail
+
+	AND	$0xFF, R(4)		/* it's a byte */
+	SLL	$8, R(4), R(TMP)	/* replicate to a word */
+	ORR	R(TMP), R(4)
+	SLL	$16, R(4), R(TMP)
+	ORR	R(TMP), R(4)
+
+_4align:				/* align on 4 */
+	AND.S	$3, R(TO), R(TMP)
+	BEQ	_4aligned
+
+	MOVBU.P	R(4), 1(R(TO))		/* implicit write back */
+	B	_4align
+
+_4aligned:
+	SUB	$31, R(TOE), R(TMP)	/* do 32-byte chunks if possible */
+	CMP	R(TMP), R(TO)
+	BHS	_4tail
+
+	MOVW	R4, R5			/* replicate */
+	MOVW	R4, R6
+	MOVW	R4, R7
+	MOVW	R4, R8
+	MOVW	R4, R9
+	MOVW	R4, R10
+	MOVW	R4, R11
+
+_f32loop:
+	CMP	R(TMP), R(TO)
+	BHS	_4tail
+
+	MOVM.IA.W [R4-R11], (R(TO))
+	B	_f32loop
+
+_4tail:
+	SUB	$3, R(TOE), R(TMP)	/* do remaining words if possible */
+_4loop:
+	CMP	R(TMP), R(TO)
+	BHS	_1tail
+
+	MOVW.P	R(4), 4(R(TO))		/* implicit write back */
+	B	_4loop
+
+_1tail:
+	CMP	R(TO), R(TOE)
+	BEQ	_return
+
+	MOVBU.P	R(4), 1(R(TO))		/* implicit write back */
+	B	_1tail
+
+_return:
+	RET
diff --git a/libkern/memset-mips.s b/libkern/memset-mips.s
new file mode 100644
index 00000000..8c9467f4
--- /dev/null
+++ b/libkern/memset-mips.s
@@ -0,0 +1,88 @@
+	TEXT	memset(SB),$12
+MOVW R1, 0(FP)
+
+/*
+ * performance:
+ *	about 1us/call and 28mb/sec
+ */
+
+	MOVW	n+8(FP), R3		/* R3 is count */
+	MOVW	p+0(FP), R4		/* R4 is pointer */
+	MOVW	c+4(FP), R5		/* R5 is char */
+	ADDU	R3,R4, R6		/* R6 is end pointer */
+
+/*
+ * if not at least 4 chars,
+ * dont even mess around.
+ * 3 chars to guarantee any
+ * rounding up to a word
+ * boundary and 4 characters
+ * to get at least maybe one
+ * full word store.
+ */
+	SGT	$4,R3, R1
+	BNE	R1, out
+
+/*
+ * turn R5 into a word of characters
+ */
+	AND	$0xff, R5
+	SLL	$8,R5, R1
+	OR	R1, R5
+	SLL	$16,R5, R1
+	OR	R1, R5
+
+/*
+ * store one byte at a time until pointer
+ * is alligned on a word boundary
+ */
+l1:
+	AND	$3,R4, R1
+	BEQ	R1, l2
+	MOVB	R5, 0(R4)
+	ADDU	$1, R4
+	JMP	l1
+
+/*
+ * turn R3 into end pointer-15
+ * store 16 at a time while theres room
+ */
+l2:
+	ADDU	$-15,R6, R3
+l3:
+	SGTU	R3,R4, R1
+	BEQ	R1, l4
+	MOVW	R5, 0(R4)
+	MOVW	R5, 4(R4)
+	ADDU	$16, R4
+	MOVW	R5, -8(R4)
+	MOVW	R5, -4(R4)
+	JMP	l3
+
+/*
+ * turn R3 into end pointer-3
+ * store 4 at a time while theres room
+ */
+l4:
+	ADDU	$-3,R6, R3
+l5:
+	SGTU	R3,R4, R1
+	BEQ	R1, out
+	MOVW	R5, 0(R4)
+	ADDU	$4, R4
+	JMP	l5
+
+/*
+ * last loop, store byte at a time
+ */
+out:
+	SGTU	R6,R4 ,R1
+	BEQ	R1, ret
+	MOVB	R5, 0(R4)
+	ADDU	$1, R4
+	JMP	out
+
+ret:
+	MOVW	s1+0(FP), R1
+	RET
+	END
diff --git a/libkern/memset-power.s b/libkern/memset-power.s
new file mode 100644
index 00000000..fa6e8d92
--- /dev/null
+++ b/libkern/memset-power.s
@@ -0,0 +1,73 @@
+	TEXT	memset(SB),$0
+#define	BDNZ	BC	16,0,
+	MOVW R3, p+0(FP)		/* R3 is pointer */
+
+/*
+ * performance:
+ *	about 100mbytes/sec (8k blocks) on a 603/105 without L2 cache
+ *	drops to 40mbytes/sec (10k blocks) and 28mbytes/sec with 32k blocks
+ */
+
+	MOVW	n+8(FP), R4		/* R4 is count */
+	CMP	R4, $0
+	BLE	ret
+	MOVW	c+4(FP), R5		/* R5 is char */
+
+/*
+ * create 16 copies of c in R5 .. R8
+ */
+	RLWNM	$0, R5, $0xff, R5
+	RLWMI	$8, R5, $0xff00, R5
+	RLWMI	$16, R5, $0xffff0000, R5
+	MOVW	R5, R6
+	MOVW	R5, R7
+	MOVW	R5, R8
+
+/*
+ * let STSW do the work for 16 characters or less; aligned and unaligned
+ */
+	CMP	R4, $16
+	BLE	out
+
+/*
+ * store enough bytes to align pointer
+ */
+	ANDCC	$7,R3, R9
+	BEQ	l2
+	SUBC	R9, $8, R9
+	MOVW	R9, XER
+	STSW	R5, (R3)
+	ADD	R9, R3
+	SUB	R9, R4
+
+/*
+ * store 16 at a time while there's room
+ * STSW was used here originally, but it's `completion serialised'
+ */
+l2:
+	SRAWCC	$4, R4, R9
+	BLE	out
+	MOVW	R9, CTR
+l3:
+	MOVW	R5, 0(R3)
+	ADD	$8, R3, R10
+	MOVW	R6, 4(R3)
+	MOVW	R7, 0(R10)
+	ADD	$8, R10, R3
+	MOVW	R8, 4(R10)
+	BDNZ	l3
+	RLWNMCC	$0, R4, $15, R4	/* residue */
+	BEQ	ret
+
+/*
+ * store up to 16 bytes from R5 .. R8; aligned and unaligned
+ */
+
+out:
+	MOVW	R4, XER
+	STSW	R5, (R3)
+
+ret:
+	MOVW	0(FP), R3
+	RETURN
+	END
diff --git a/libkern/memset-sparc.s b/libkern/memset-sparc.s
new file mode 100644
index 00000000..8c7b26c8
--- /dev/null
+++ b/libkern/memset-sparc.s
@@ -0,0 +1,88 @@
+	TEXT	memset(SB),$0
+
+/*
+ * performance:
+ *	(tba)
+ */
+
+MOVW	R7, 0(FP)
+	MOVW	n+8(FP), R9		/* R9 is count */
+	MOVW	p+0(FP), R10		/* R10 is pointer */
+	MOVW	c+4(FP), R11		/* R11 is char */
+	ADD	R9,R10, R12		/* R12 is end pointer */
+
+/*
+ * if not at least 4 chars,
+ * dont even mess around.
+ * 3 chars to guarantee any
+ * rounding up to a word
+ * boundary and 4 characters
+ * to get at least maybe one
+ * full word store.
+ */
+	SUBCC	$4,R9, R0
+	BL	out
+
+/*
+ * turn R11 into a word of characters
+ */
+	AND	$0xff, R11
+	SLL	$8,R11, R7
+	OR	R7, R11
+	SLL	$16,R11, R7
+	OR	R7, R11
+
+/*
+ * store one byte at a time until pointer
+ * is alligned on a word boundary
+ */
+l1:
+	ANDCC	$3,R10, R0
+	BE	l2
+	MOVB	R11, 0(R10)
+	ADD	$1, R10
+	JMP	l1
+
+/*
+ * turn R9 into end pointer-15
+ * store 16 at a time while theres room
+ */
+l2:
+	ADD	$-15,R12, R9
+	SUBCC	R10,R9, R0
+	BLEU	l4
+l3:
+	MOVW	R11, 0(R10)
+	MOVW	R11, 4(R10)
+	ADD	$16, R10
+	SUBCC	R10,R9, R0
+	MOVW	R11, -8(R10)
+	MOVW	R11, -4(R10)
+	BGU	l3
+
+/*
+ * turn R9 into end pointer-3
+ * store 4 at a time while theres room
+ */
+l4:
+	ADD	$-3,R12, R9
+l5:
+	SUBCC	R10,R9, R0
+	BLEU	out
+	MOVW	R11, 0(R10)
+	ADD	$4, R10
+	JMP	l5
+
+/*
+ * last loop, store byte at a time
+ */
+out:
+	SUBCC	R10,R12, R0
+	BLEU	ret
+	MOVB	R11, 0(R10)
+	ADD	$1, R10
+	JMP	out
+
+ret:
+	MOVW	s1+0(FP), R7
+	RETURN
diff --git a/libkern/memset-thumb.s b/libkern/memset-thumb.s
new file mode 100644
index 00000000..92355c11
--- /dev/null
+++ b/libkern/memset-thumb.s
@@ -0,0 +1,66 @@
+TO = 1
+TOE = 2
+N = 3
+TMP = 3					/* N and TMP don't overlap */
+
+TEXT memset(SB), $0
+	MOVW	R0, R(TO)
+	MOVW	data+4(FP), R(4)
+	MOVW	n+8(FP), R(N)
+
+	ADD	R(N), R(TO), R(TOE)	/* to end pointer */
+
+	CMP	$4, R(N)		/* need at least 4 bytes to copy */
+	BLT	_1tail
+
+	AND	$0xFF, R(4)		/* it's a byte */
+	SLL	$8, R(4), R(TMP)	/* replicate to a word */
+	ORR	R(TMP), R(4)
+	SLL	$16, R(4), R(TMP)
+	ORR	R(TMP), R(4)
+
+_4align:				/* align on 4 */
+	AND.S	$3, R(TO), R(TMP)
+	BEQ	_4aligned
+
+	MOVBU.P	R(4), 1(R(TO))		/* implicit write back */
+	B	_4align
+
+_4aligned:
+	SUB	$31, R(TOE), R(TMP)	/* do 32-byte chunks if possible */
+	CMP	R(TMP), R(TO)
+	BHS	_4tail
+
+	MOVW	R4, R5			/* replicate */
+	MOVW	R4, R6
+	MOVW	R4, R7
+	MOVW	R4, R8
+	MOVW	R4, R9
+	MOVW	R4, R10
+	MOVW	R4, R11
+
+_f32loop:
+	CMP	R(TMP), R(TO)
+	BHS	_4tail
+
+	MOVM.IA.W [R4-R11], (R(TO))
+	B	_f32loop
+
+_4tail:
+	SUB	$3, R(TOE), R(TMP)	/* do remaining words if possible */
+_4loop:
+	CMP	R(TMP), R(TO)
+	BHS	_1tail
+
+	MOVW.P	R(4), 4(R(TO))		/* implicit write back */
+	B	_4loop
+
+_1tail:
+	CMP	R(TO), R(TOE)
+	BEQ	_return
+
+	MOVBU.P	R(4), 1(R(TO))		/* implicit write back */
+	B	_1tail
+
+_return:
+	RET
diff --git a/libkern/memset.c b/libkern/memset.c
new file mode 100644
index 00000000..270fca25
--- /dev/null
+++ b/libkern/memset.c
@@ -0,0 +1,15 @@
+#include	<lib9.h>
+
+void*
+memset(void *ap, int c, ulong n)
+{
+	char *p;
+	int m = (int)n;
+
+	p = ap;
+	while(m > 0) {
+		*p++ = c;
+		m--;
+	}
+	return ap;
+}
diff --git a/libkern/mkfile b/libkern/mkfile
new file mode 100644
index 00000000..9bb9d65e
--- /dev/null
+++ b/libkern/mkfile
@@ -0,0 +1,81 @@
+<../mkconfig
+
+LIB=libkern.a
+
+COMMONFILES=\
+	abort.$O\
+	abs.$O\
+	atol.$O\
+	charstod.$O\
+	cistrcmp.$O\
+	cistrncmp.$O\
+	cistrstr.$O\
+	cleanname.$O\
+	convD2M.$O\
+	convM2D.$O\
+	convM2S.$O\
+	convS2M.$O\
+	dofmt.$O\
+	exp.$O\
+	fcallfmt.$O\
+	floor.$O\
+	fmt.$O\
+	fmtprint.$O\
+	fmtquote.$O\
+	fmtstr.$O\
+	fmtvprint.$O\
+	getfields.$O\
+	log.$O\
+	memccpy.$O\
+	memchr.$O\
+	memcmp.$O\
+#	netmkaddr.$O\
+	pow.$O\
+	pow10.$O\
+	qsort.$O\
+	rune.$O\
+	runestrlen.$O\
+	sin.$O\
+	seprint.$O\
+	smprint.$O\
+	snprint.$O\
+	sqrt.$O\
+	strcat.$O\
+	strcmp.$O\
+	strcpy.$O\
+	strdup.$O\
+	strecpy.$O\
+	strlen.$O\
+	strncmp.$O\
+	strncpy.$O\
+	strrchr.$O\
+	strstr.$O\
+	strtod.$O\
+	strtol.$O\
+	strtoll.$O\
+	strtoul.$O\
+	tokenize.$O\
+	toupper.$O\
+	u16.$O\
+	u32.$O\
+	u64.$O\
+	utfecpy.$O\
+	utflen.$O\
+	utfnlen.$O\
+	utfrrune.$O\
+	utfrune.$O\
+	vseprint.$O\
+	vsmprint.$O\
+	vsnprint.$O\
+
+<mkfile-$OBJTYPE		#sets $TARGFILES
+
+OFILES=	$COMMONFILES $TARGFILES
+
+<$ROOT/mkfiles/mksyslib-$SHELLTYPE
+
+convD2M.$O: $ROOT/include/fcall.h
+convM2D.$O: $ROOT/include/fcall.h
+convM2S.$O: $ROOT/include/fcall.h
+convS2M.$O: $ROOT/include/fcall.h
+fcallfmt.$O: $ROOT/include/fcall.h
diff --git a/libkern/mkfile-386 b/libkern/mkfile-386
new file mode 100644
index 00000000..00c0fcd5
--- /dev/null
+++ b/libkern/mkfile-386
@@ -0,0 +1,12 @@
+#
+#	386-specific files
+#
+TARGFILES=\
+	frexp-386.$O\
+	getfcr-386.$O\
+	memmove-386.$O\
+	memset-386.$O\
+	nan-386.$O\
+	strchr-386.$O\
+	vlop-386.$O\
+	vlrt-386.$O\
diff --git a/libkern/mkfile-68000 b/libkern/mkfile-68000
new file mode 100644
index 00000000..dd5da424
--- /dev/null
+++ b/libkern/mkfile-68000
@@ -0,0 +1,12 @@
+#
+#	68000-specific files
+#
+TARGFILES=\
+	frexp-68000.$O\
+	getfcr-68000.$O\
+	memmove-68000.$O\
+	memset-68000.$O\
+	nan-68000.$O\
+	strchr-68000.$O\
+	muldiv-68000.$O\
+	vlrt-68000.$O\
diff --git a/libkern/mkfile-arm b/libkern/mkfile-arm
new file mode 100644
index 00000000..01af1112
--- /dev/null
+++ b/libkern/mkfile-arm
@@ -0,0 +1,13 @@
+#
+#	arm-specific files
+#
+TARGFILES=\
+	frexp-arm.$O\
+	getfcr-arm.$O\
+	memmove-arm.$O\
+	memset-arm.$O\
+	nan-arm.$O\
+	strchr-arm.$O\
+	vlop-arm.$O\
+	vlrt-arm.$O\
+	div-arm.$O\
diff --git a/libkern/mkfile-mips b/libkern/mkfile-mips
new file mode 100644
index 00000000..088f915e
--- /dev/null
+++ b/libkern/mkfile-mips
@@ -0,0 +1,13 @@
+#
+#	mips-specific files
+#
+TARGFILES=\
+	frexp-mips.$O\
+	getfcr-mips.$O\
+	memmove-mips.$O\
+	memset-mips.$O\
+	nan-mips.$O\
+	strchr-mips.$O\
+	vlop-mips.$O\
+	vlrt-mips.$O\
+
diff --git a/libkern/mkfile-power b/libkern/mkfile-power
new file mode 100644
index 00000000..aa616ca7
--- /dev/null
+++ b/libkern/mkfile-power
@@ -0,0 +1,13 @@
+#
+#	power-specific files
+#
+TARGFILES=\
+	frexp-power.$O\
+	getfcr-power.$O\
+	memmove-power.$O\
+	memset-power.$O\
+	nan-power.$O\
+	strchr-power.$O\
+	vlop-power.$O\
+	vlrt-power.$O\
+
diff --git a/libkern/mkfile-sparc b/libkern/mkfile-sparc
new file mode 100644
index 00000000..465f88e7
--- /dev/null
+++ b/libkern/mkfile-sparc
@@ -0,0 +1,13 @@
+#
+#	sparc-specific files
+#
+TARGFILES=\
+	frexp-sparc.$O\
+	getfcr-sparc.$O\
+	memmove-sparc.$O\
+	memset-sparc.$O\
+	nan-sparc.$O\
+	strchr-sparc.$O\
+	vlop-sparc.$O\
+	vlrt-sparc.$O\
+
diff --git a/libkern/mkfile-spim b/libkern/mkfile-spim
new file mode 100644
index 00000000..fe9a79f5
--- /dev/null
+++ b/libkern/mkfile-spim
@@ -0,0 +1,8 @@
+#
+#	spim-specific files
+#
+TARGFILES=\
+	getfcr-spim.$O\
+	memmove-spim.$O\
+	memset-spim.$O\
+	strchr-spim.$O\
diff --git a/libkern/mkfile-thumb b/libkern/mkfile-thumb
new file mode 100644
index 00000000..4ccd1653
--- /dev/null
+++ b/libkern/mkfile-thumb
@@ -0,0 +1,12 @@
+#
+#	thumb-specific files
+#
+TARGFILES=\
+	frexp-thumb.$O\
+	getfcr-thumb.$O\
+	memmove-thumb.$O\
+	memset-thumb.$O\
+	nan-thumb.$O\
+	strchr-thumb.$O\
+	vlop-thumb.$O\
+	vlrt-thumb.$O\
diff --git a/libkern/muldiv-68000.s b/libkern/muldiv-68000.s
new file mode 100644
index 00000000..10ef6f46
--- /dev/null
+++ b/libkern/muldiv-68000.s
@@ -0,0 +1,172 @@
+/*
+ * calls _divul with
+ * absolute value arguments
+ */
+TEXT	_divsl(SB), $0
+	MOVL	R0, TOS
+
+	MOVL	b+4(FP), R0
+	BPL	y1
+	NEGL	R0
+	MOVL	R0, TOS
+
+	MOVL	a+0(FP), R0
+	BPL	y3
+	NEGL	R0
+	MOVL	R0, TOS
+
+	/* neg/neg */
+	JSR	_divul(SB)
+	MOVL	TOS, R0
+	MOVL	R0, a+0(FP)
+	MOVL	TOS, R0
+	NEGL	R0
+	MOVL	R0, b+4(FP)
+	MOVL	TOS, R0
+	RTS
+
+y1:	MOVL	R0, TOS
+
+	MOVL	a+0(FP), R0
+	BPL	y2
+	NEGL	R0
+	MOVL	R0, TOS
+
+	/* neg/pos */
+	JSR	_divul(SB)
+	MOVL	TOS, R0
+	NEGL	R0
+	MOVL	R0, a+0(FP)
+	MOVL	TOS, R0
+	NEGL	R0
+	MOVL	R0, b+4(FP)
+	MOVL	TOS, R0
+	RTS
+
+y2:	MOVL	R0, TOS
+
+	/* pos/pos */
+	JSR	_divul(SB)
+	MOVL	TOS, R0
+	MOVL	R0, a+0(FP)
+	MOVL	TOS, R0
+	MOVL	R0, b+4(FP)
+	MOVL	TOS, R0
+	RTS
+
+y3:	MOVL	R0, TOS
+
+	/* pos/neg */
+	JSR	_divul(SB)
+	MOVL	TOS, R0
+	NEGL	R0
+	MOVL	R0, a+0(FP)
+	MOVL	TOS, R0
+	MOVL	R0, b+4(FP)
+	MOVL	TOS, R0
+	RTS
+
+/*
+ *	for(i=1;; i++) {
+ *		if(den & (1<<31))
+ *			break;
+ *		den <<= 1;
+ *	}
+ *
+ *	for(; i; i--) {
+ *		quo <<= 1;
+ *		if(num >= den) {
+ *			num -= den;
+ *			quo |= 1;
+ *		}
+ *		den >>= 1;
+ *	}
+ */
+TEXT	_divul(SB), $0
+	MOVL	R0, TOS	/* i */
+	MOVL	R1, TOS	/* num */
+	MOVL	R2, TOS	/* den */
+	MOVL	R3, TOS	/* quo */
+
+	MOVL	$0, R0
+	MOVL	$0, R3
+	MOVL	a+0(FP), R1
+	MOVL	b+4(FP), R2
+	BEQ	xout
+	BMI	x1
+
+	ADDL	$1, R0
+	LSLL	$1, R2
+	BPL	-2(PC)
+
+x1:	LSLL	$1, R3
+	CMPL	R1, R2
+	BCS	3(PC)
+	SUBL	R2, R1
+	ORL	$1, R3
+	LSRL	$1, R2
+	DBMI	R0, x1
+
+	MOVL	R3, a+0(FP)
+	MOVL	R1, b+4(FP)
+
+xout:
+	MOVL	TOS, R3
+	MOVL	TOS, R2
+	MOVL	TOS, R1
+	MOVL	TOS, R0
+	RTS
+
+/*
+ *	x = 0;
+ *	for(i=0; i<32; i++) {
+ *		if(a & 1)
+ *			x += b;
+ *		a >>= 1;
+ *		b <<= 1;
+ *	}
+ *	a = x;
+ */
+TEXT	_mull(SB), $0
+	MOVL	R0, TOS	/* i */
+	MOVL	R1, TOS	/* a */
+	MOVL	R2, TOS	/* b */
+	MOVL	R3, TOS	/* x */
+
+	MOVL	a+0(FP), R1
+	MOVL	b+4(FP), R2
+	MOVL	$32, R0
+	CLRL	R3
+
+z1:	ROTRL	$1, R1
+	BCC	2(PC)
+	ADDL	R2, R3
+	LSLL	$1, R2
+	DBEQ	R0, z1
+
+	MOVL	R3, b+4(FP)
+	MOVL	TOS, R3
+	MOVL	TOS, R2
+	MOVL	TOS, R1
+	MOVL	TOS, R0
+	RTS
+
+TEXT	_ccr(SB), $0
+	PEA	(A0)
+	SUBL	A0, A0
+
+	BCC	2(PC)
+	LEA	1(A0), A0
+
+	BVC	2(PC)
+	LEA	2(A0), A0
+
+	BNE	2(PC)
+	LEA	4(A0), A0
+
+	BPL	2(PC)
+	LEA	8(A0), A0
+
+	MOVW	A0, a+0(FP)
+	MOVL	TOS, A0
+	RTS
diff --git a/libkern/nan-386.c b/libkern/nan-386.c
new file mode 100644
index 00000000..2152925a
--- /dev/null
+++ b/libkern/nan-386.c
@@ -0,0 +1,69 @@
+#include <lib9.h>
+
+#define	NANEXP	(2047<<20)
+#define	NANMASK	(2047<<20)
+#define	NANSIGN	(1<<31)
+
+double
+NaN(void)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.x[1] = NANEXP;
+	a.x[0] = 1;
+	return a.d;
+}
+
+int
+isNaN(double d)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.d = d;
+	if((a.x[1] & NANMASK) != NANEXP)
+		return 0;
+	return !isInf(d, 0);
+}
+
+double
+Inf(int sign)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.x[1] = NANEXP;
+	a.x[0] = 0;
+	if(sign < 0)
+		a.x[1] |= NANSIGN;
+	return a.d;
+}
+
+int
+isInf(double d, int sign)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.d = d;
+	if(a.x[0] != 0)
+		return 0;
+	if(a.x[1] == NANEXP)
+		return sign >= 0;
+	if(a.x[1] == (NANEXP|NANSIGN))
+		return sign <= 0;
+	return 0;
+}
diff --git a/libkern/nan-68000.c b/libkern/nan-68000.c
new file mode 100644
index 00000000..e1b3db36
--- /dev/null
+++ b/libkern/nan-68000.c
@@ -0,0 +1,70 @@
+#include <u.h>
+#include <libc.h>
+
+#define	NANEXP	(2047<<20)
+#define	NANMASK	(2047<<20)
+#define	NANSIGN	(1<<31)
+
+double
+NaN(void)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.x[0] = NANEXP;
+	a.x[1] = 1;
+	return a.d;
+}
+
+int
+isNaN(double d)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.d = d;
+	if((a.x[0] & NANMASK) != NANEXP)
+		return 0;
+	return !isInf(d, 0);
+}
+
+double
+Inf(int sign)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.x[0] = NANEXP;
+	a.x[1] = 0;
+	if(sign < 0)
+		a.x[0] |= NANSIGN;
+	return a.d;
+}
+
+int
+isInf(double d, int sign)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.d = d;
+	if(a.x[1] != 0)
+		return 0;
+	if(a.x[0] == NANEXP)
+		return sign >= 0;
+	if(a.x[0] == (NANEXP|NANSIGN))
+		return sign <= 0;
+	return 0;
+}
diff --git a/libkern/nan-arm.c b/libkern/nan-arm.c
new file mode 100644
index 00000000..f1cfa0b0
--- /dev/null
+++ b/libkern/nan-arm.c
@@ -0,0 +1,69 @@
+#include <lib9.h>
+
+#define	NANEXP	(2047<<20)
+#define	NANMASK	(2047<<20)
+#define	NANSIGN	(1<<31)
+
+double
+NaN(void)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.x[0] = NANEXP;
+	a.x[1] = 1;
+	return a.d;
+}
+
+int
+isNaN(double d)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.d = d;
+	if((a.x[0] & NANMASK) != NANEXP)
+		return 0;
+	return !isInf(d, 0);
+}
+
+double
+Inf(int sign)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.x[0] = NANEXP;
+	a.x[1] = 0;
+	if(sign < 0)
+		a.x[0] |= NANSIGN;
+	return a.d;
+}
+
+int
+isInf(double d, int sign)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.d = d;
+	if(a.x[1] != 0)
+		return 0;
+	if(a.x[0] == NANEXP)
+		return sign >= 0;
+	if(a.x[0] == (NANEXP|NANSIGN))
+		return sign <= 0;
+	return 0;
+}
diff --git a/libkern/nan-mips.c b/libkern/nan-mips.c
new file mode 100644
index 00000000..e1b3db36
--- /dev/null
+++ b/libkern/nan-mips.c
@@ -0,0 +1,70 @@
+#include <u.h>
+#include <libc.h>
+
+#define	NANEXP	(2047<<20)
+#define	NANMASK	(2047<<20)
+#define	NANSIGN	(1<<31)
+
+double
+NaN(void)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.x[0] = NANEXP;
+	a.x[1] = 1;
+	return a.d;
+}
+
+int
+isNaN(double d)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.d = d;
+	if((a.x[0] & NANMASK) != NANEXP)
+		return 0;
+	return !isInf(d, 0);
+}
+
+double
+Inf(int sign)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.x[0] = NANEXP;
+	a.x[1] = 0;
+	if(sign < 0)
+		a.x[0] |= NANSIGN;
+	return a.d;
+}
+
+int
+isInf(double d, int sign)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.d = d;
+	if(a.x[1] != 0)
+		return 0;
+	if(a.x[0] == NANEXP)
+		return sign >= 0;
+	if(a.x[0] == (NANEXP|NANSIGN))
+		return sign <= 0;
+	return 0;
+}
diff --git a/libkern/nan-power.c b/libkern/nan-power.c
new file mode 100644
index 00000000..f1cfa0b0
--- /dev/null
+++ b/libkern/nan-power.c
@@ -0,0 +1,69 @@
+#include <lib9.h>
+
+#define	NANEXP	(2047<<20)
+#define	NANMASK	(2047<<20)
+#define	NANSIGN	(1<<31)
+
+double
+NaN(void)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.x[0] = NANEXP;
+	a.x[1] = 1;
+	return a.d;
+}
+
+int
+isNaN(double d)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.d = d;
+	if((a.x[0] & NANMASK) != NANEXP)
+		return 0;
+	return !isInf(d, 0);
+}
+
+double
+Inf(int sign)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.x[0] = NANEXP;
+	a.x[1] = 0;
+	if(sign < 0)
+		a.x[0] |= NANSIGN;
+	return a.d;
+}
+
+int
+isInf(double d, int sign)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.d = d;
+	if(a.x[1] != 0)
+		return 0;
+	if(a.x[0] == NANEXP)
+		return sign >= 0;
+	if(a.x[0] == (NANEXP|NANSIGN))
+		return sign <= 0;
+	return 0;
+}
diff --git a/libkern/nan-sparc.c b/libkern/nan-sparc.c
new file mode 100644
index 00000000..f1cfa0b0
--- /dev/null
+++ b/libkern/nan-sparc.c
@@ -0,0 +1,69 @@
+#include <lib9.h>
+
+#define	NANEXP	(2047<<20)
+#define	NANMASK	(2047<<20)
+#define	NANSIGN	(1<<31)
+
+double
+NaN(void)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.x[0] = NANEXP;
+	a.x[1] = 1;
+	return a.d;
+}
+
+int
+isNaN(double d)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.d = d;
+	if((a.x[0] & NANMASK) != NANEXP)
+		return 0;
+	return !isInf(d, 0);
+}
+
+double
+Inf(int sign)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.x[0] = NANEXP;
+	a.x[1] = 0;
+	if(sign < 0)
+		a.x[0] |= NANSIGN;
+	return a.d;
+}
+
+int
+isInf(double d, int sign)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.d = d;
+	if(a.x[1] != 0)
+		return 0;
+	if(a.x[0] == NANEXP)
+		return sign >= 0;
+	if(a.x[0] == (NANEXP|NANSIGN))
+		return sign <= 0;
+	return 0;
+}
diff --git a/libkern/nan-thumb.c b/libkern/nan-thumb.c
new file mode 100644
index 00000000..f1cfa0b0
--- /dev/null
+++ b/libkern/nan-thumb.c
@@ -0,0 +1,69 @@
+#include <lib9.h>
+
+#define	NANEXP	(2047<<20)
+#define	NANMASK	(2047<<20)
+#define	NANSIGN	(1<<31)
+
+double
+NaN(void)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.x[0] = NANEXP;
+	a.x[1] = 1;
+	return a.d;
+}
+
+int
+isNaN(double d)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.d = d;
+	if((a.x[0] & NANMASK) != NANEXP)
+		return 0;
+	return !isInf(d, 0);
+}
+
+double
+Inf(int sign)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.x[0] = NANEXP;
+	a.x[1] = 0;
+	if(sign < 0)
+		a.x[0] |= NANSIGN;
+	return a.d;
+}
+
+int
+isInf(double d, int sign)
+{
+	union
+	{
+		double	d;
+		long	x[2];
+	} a;
+
+	a.d = d;
+	if(a.x[1] != 0)
+		return 0;
+	if(a.x[0] == NANEXP)
+		return sign >= 0;
+	if(a.x[0] == (NANEXP|NANSIGN))
+		return sign <= 0;
+	return 0;
+}
diff --git a/libkern/netmkaddr.c b/libkern/netmkaddr.c
new file mode 100644
index 00000000..1e3b6294
--- /dev/null
+++ b/libkern/netmkaddr.c
@@ -0,0 +1,50 @@
+#include <lib9.h>
+
+/*
+ *  make an address, add the defaults
+ */
+char *
+netmkaddr(char *linear, char *defnet, char *defsrv)
+{
+	static char addr[4*(NAMELEN+1)];
+	char *cp;
+
+	/*
+	 *  dump network name
+	 */
+	cp = strchr(linear, '!');
+	if(cp == 0){
+		if(defnet==0){
+			if(defsrv)
+				sprint(addr, "net!%.*s!%.*s", 2*NAMELEN, linear,
+					NAMELEN, defsrv);
+			else
+				sprint(addr, "net!%.*s", 2*NAMELEN, linear);
+		} else {
+			if(defsrv)
+				sprint(addr, "%.*s!%.*s!%.*s", NAMELEN, defnet,
+					2*NAMELEN, linear, NAMELEN, defsrv);
+			else
+				sprint(addr, "%.*s!%.*s", NAMELEN, defnet,
+					2*NAMELEN, linear);
+		}
+		return addr;
+	}
+
+	/*
+	 *  if there is already a service, use it
+	 */
+	cp = strchr(cp+1, '!');
+	if(cp)
+		return linear;
+
+	/*
+	 *  add default service
+	 */
+	if(defsrv == 0)
+		return linear;
+	sprint(addr, "%.*s!%.*s", 3*NAMELEN, linear,
+		NAMELEN, defsrv);
+
+	return addr;
+}
diff --git a/libkern/pow.c b/libkern/pow.c
new file mode 100644
index 00000000..2dd39143
--- /dev/null
+++ b/libkern/pow.c
@@ -0,0 +1,68 @@
+#include <lib9.h>
+
+double
+pow(double x, double y) /* return x ^ y (exponentiation) */
+{
+	double xy, y1, ye;
+	long i;
+	int ex, ey, flip;
+
+	if(y == 0.0)
+		return 1.0;
+
+	flip = 0;
+	if(y < 0.){
+		y = -y;
+		flip = 1;
+	}
+	y1 = modf(y, &ye);
+	if(y1 != 0.0){
+		if(x <= 0.)
+			goto zreturn;
+		if(y1 > 0.5) {
+			y1 -= 1.;
+			ye += 1.;
+		}
+		xy = exp(y1 * log(x));
+	}else
+		xy = 1.0;
+	if(ye > 0x7FFFFFFF){	/* should be ~0UL but compiler can't convert double to ulong */
+		if(x <= 0.){
+ zreturn:
+			if(x==0. && !flip)
+				return 0.;
+			return NaN();
+		}
+		if(flip){
+			if(y == .5)
+				return 1./sqrt(x);
+			y = -y;
+		}else if(y == .5)
+			return sqrt(x);
+		return exp(y * log(x));
+	}
+	x = frexp(x, &ex);
+	ey = 0;
+	i = ye;
+	if(i)
+		for(;;){
+			if(i & 1){
+				xy *= x;
+				ey += ex;
+			}
+			i >>= 1;
+			if(i == 0)
+				break;
+			x *= x;
+			ex <<= 1;
+			if(x < .5){
+				x += x;
+				ex -= 1;
+			}
+		}
+	if(flip){
+		xy = 1. / xy;
+		ey = -ey;
+	}
+	return ldexp(xy, ey);
+}
diff --git a/libkern/pow10.c b/libkern/pow10.c
new file mode 100644
index 00000000..933f5d13
--- /dev/null
+++ b/libkern/pow10.c
@@ -0,0 +1,40 @@
+#include	<lib9.h>
+
+/*
+ * this table might overflow 127-bit exponent representations.
+ * in that case, truncate it after 1.0e38.
+ * it is important to get all one can from this
+ * routine since it is used in atof to scale numbers.
+ * the presumption is that C converts fp numbers better
+ * than multipication of lower powers of 10.
+ */
+
+static
+double	tab[] =
+{
+	1.0e0, 1.0e1, 1.0e2, 1.0e3, 1.0e4, 1.0e5, 1.0e6, 1.0e7, 1.0e8, 1.0e9,
+	1.0e10,1.0e11,1.0e12,1.0e13,1.0e14,1.0e15,1.0e16,1.0e17,1.0e18,1.0e19,
+	1.0e20,1.0e21,1.0e22,1.0e23,1.0e24,1.0e25,1.0e26,1.0e27,1.0e28,1.0e29,
+	1.0e30,1.0e31,1.0e32,1.0e33,1.0e34,1.0e35,1.0e36,1.0e37,1.0e38,1.0e39,
+	1.0e40,1.0e41,1.0e42,1.0e43,1.0e44,1.0e45,1.0e46,1.0e47,1.0e48,1.0e49,
+	1.0e50,1.0e51,1.0e52,1.0e53,1.0e54,1.0e55,1.0e56,1.0e57,1.0e58,1.0e59,
+	1.0e60,1.0e61,1.0e62,1.0e63,1.0e64,1.0e65,1.0e66,1.0e67,1.0e68,1.0e69,
+};
+
+double
+pow10(int n)
+{
+	int m;
+
+	if(n < 0) {
+		n = -n;
+		if(n < sizeof(tab)/sizeof(tab[0]))
+			return 1/tab[n];
+		m = n/2;
+		return pow10(-m) * pow10(m-n);
+	}
+	if(n < sizeof(tab)/sizeof(tab[0]))
+		return tab[n];
+	m = n/2;
+	return pow10(m) * pow10(n-m);
+}
diff --git a/libkern/qsort.c b/libkern/qsort.c
new file mode 100644
index 00000000..3388891c
--- /dev/null
+++ b/libkern/qsort.c
@@ -0,0 +1,122 @@
+/*
+ * qsort -- simple quicksort
+ */
+
+typedef
+struct
+{
+	int	(*cmp)(void*, void*);
+	void	(*swap)(char*, char*, long);
+	long	es;
+} Sort;
+
+static	void
+swapb(char *i, char *j, long es)
+{
+	char c;
+
+	do {
+		c = *i;
+		*i++ = *j;
+		*j++ = c;
+		es--;
+	} while(es != 0);
+
+}
+
+static	void
+swapi(char *ii, char *ij, long es)
+{
+	long *i, *j, c;
+
+	i = (long*)ii;
+	j = (long*)ij;
+	do {
+		c = *i;
+		*i++ = *j;
+		*j++ = c;
+		es -= sizeof(long);
+	} while(es != 0);
+}
+
+static	char*
+pivot(char *a, long n, Sort *p)
+{
+	long j;
+	char *pi, *pj, *pk;
+
+	j = n/6 * p->es;
+	pi = a + j;	/* 1/6 */
+	j += j;
+	pj = pi + j;	/* 1/2 */
+	pk = pj + j;	/* 5/6 */
+	if(p->cmp(pi, pj) < 0) {
+		if(p->cmp(pi, pk) < 0) {
+			if(p->cmp(pj, pk) < 0)
+				return pj;
+			return pk;
+		}
+		return pi;
+	}
+	if(p->cmp(pj, pk) < 0) {
+		if(p->cmp(pi, pk) < 0)
+			return pi;
+		return pk;
+	}
+	return pj;
+}
+
+static	void
+qsorts(char *a, long n, Sort *p)
+{
+	long j, es;
+	char *pi, *pj, *pn;
+
+	es = p->es;
+	while(n > 1) {
+		if(n > 10) {
+			pi = pivot(a, n, p);
+		} else
+			pi = a + (n>>1)*es;
+
+		p->swap(a, pi, es);
+		pi = a;
+		pn = a + n*es;
+		pj = pn;
+		for(;;) {
+			do
+				pi += es;
+			while(pi < pn && p->cmp(pi, a) < 0);
+			do
+				pj -= es;
+			while(pj > a && p->cmp(pj, a) > 0);
+			if(pj < pi)
+				break;
+			p->swap(pi, pj, es);
+		}
+		p->swap(a, pj, es);
+		j = (pj - a) / es;
+
+		n = n-j-1;
+		if(j >= n) {
+			qsorts(a, j, p);
+			a += (j+1)*es;
+		} else {
+			qsorts(a + (j+1)*es, n, p);
+			n = j;
+		}
+	}
+}
+
+void
+qsort(void *va, long n, long es, int (*cmp)(void*, void*))
+{
+	Sort s;
+
+	s.cmp = cmp;
+	s.es = es;
+	s.swap = swapi;
+	if(((long)va | es) % sizeof(long))
+		s.swap = swapb;
+	qsorts((char*)va, n, &s);
+}
diff --git a/libkern/rune.c b/libkern/rune.c
new file mode 100644
index 00000000..d3cb28e0
--- /dev/null
+++ b/libkern/rune.c
@@ -0,0 +1,162 @@
+#include	"lib9.h"
+
+enum
+{
+	Bit1	= 7,
+	Bitx	= 6,
+	Bit2	= 5,
+	Bit3	= 4,
+	Bit4	= 3,
+
+	T1	= ((1<<(Bit1+1))-1) ^ 0xFF,	/* 0000 0000 */
+	Tx	= ((1<<(Bitx+1))-1) ^ 0xFF,	/* 1000 0000 */
+	T2	= ((1<<(Bit2+1))-1) ^ 0xFF,	/* 1100 0000 */
+	T3	= ((1<<(Bit3+1))-1) ^ 0xFF,	/* 1110 0000 */
+	T4	= ((1<<(Bit4+1))-1) ^ 0xFF,	/* 1111 0000 */
+
+	Rune1	= (1<<(Bit1+0*Bitx))-1,		/* 0000 0000 0111 1111 */
+	Rune2	= (1<<(Bit2+1*Bitx))-1,		/* 0000 0111 1111 1111 */
+	Rune3	= (1<<(Bit3+2*Bitx))-1,		/* 1111 1111 1111 1111 */
+
+	Maskx	= (1<<Bitx)-1,			/* 0011 1111 */
+	Testx	= Maskx ^ 0xFF,			/* 1100 0000 */
+
+	Bad	= Runeerror,
+};
+
+int
+chartorune(Rune *rune, char *str)
+{
+	int c, c1, c2;
+	long l;
+
+	/*
+	 * one character sequence
+	 *	00000-0007F => T1
+	 */
+	c = *(uchar*)str;
+	if(c < Tx) {
+		*rune = c;
+		return 1;
+	}
+
+	/*
+	 * two character sequence
+	 *	0080-07FF => T2 Tx
+	 */
+	c1 = *(uchar*)(str+1) ^ Tx;
+	if(c1 & Testx)
+		goto bad;
+	if(c < T3) {
+		if(c < T2)
+			goto bad;
+		l = ((c << Bitx) | c1) & Rune2;
+		if(l <= Rune1)
+			goto bad;
+		*rune = l;
+		return 2;
+	}
+
+	/*
+	 * three character sequence
+	 *	0800-FFFF => T3 Tx Tx
+	 */
+	c2 = *(uchar*)(str+2) ^ Tx;
+	if(c2 & Testx)
+		goto bad;
+	if(c < T4) {
+		l = ((((c << Bitx) | c1) << Bitx) | c2) & Rune3;
+		if(l <= Rune2)
+			goto bad;
+		*rune = l;
+		return 3;
+	}
+
+	/*
+	 * bad decoding
+	 */
+bad:
+	*rune = Bad;
+	return 1;
+}
+
+int
+runetochar(char *str, Rune *rune)
+{
+	long c;
+
+	/*
+	 * one character sequence
+	 *	00000-0007F => 00-7F
+	 */
+	c = *rune;
+	if(c <= Rune1) {
+		str[0] = c;
+		return 1;
+	}
+
+	/*
+	 * two character sequence
+	 *	0080-07FF => T2 Tx
+	 */
+	if(c <= Rune2) {
+		str[0] = T2 | (c >> 1*Bitx);
+		str[1] = Tx | (c & Maskx);
+		return 2;
+	}
+
+	/*
+	 * three character sequence
+	 *	0800-FFFF => T3 Tx Tx
+	 */
+	str[0] = T3 |  (c >> 2*Bitx);
+	str[1] = Tx | ((c >> 1*Bitx) & Maskx);
+	str[2] = Tx |  (c & Maskx);
+	return 3;
+}
+
+int
+runelen(long c)
+{
+	if(c <= Rune1)
+		return 1;
+	if(c <= Rune2)
+		return 2;
+	return 3;
+}
+
+int
+runenlen(Rune *r, int l)
+{
+	int n;
+	long c;
+
+	n = 0;
+	while(l--) {
+		c = *r++;
+		if(c <= Rune1)
+			n += 1;
+		else
+		if(c <= Rune2)
+			n += 2;
+		else
+			n += 3;
+	}
+	return n;
+}
+
+int
+fullrune(char *str, int n)
+{
+	int c;
+
+	if(n > 0) {
+		c = *(uchar*)str;
+		if(c < Tx)
+			return 1;
+		if(n > 1)
+			if(c < T3 || n > 2)
+				return 1;
+	}
+	return 0;
+}
diff --git a/libkern/runestrlen.c b/libkern/runestrlen.c
new file mode 100644
index 00000000..e1cd84ac
--- /dev/null
+++ b/libkern/runestrlen.c
@@ -0,0 +1,13 @@
+#include "lib9.h"
+
+
+long
+runestrlen(Rune *s)
+{
+	int i;
+
+	i = 0;
+	while(*s++)
+		i++;
+	return i;
+}
diff --git a/libkern/seprint.c b/libkern/seprint.c
new file mode 100644
index 00000000..7dadad62
--- /dev/null
+++ b/libkern/seprint.c
@@ -0,0 +1,26 @@
+/*
+ * The authors of this software are Rob Pike and Ken Thompson.
+ *              Copyright (c) 2002 by Lucent Technologies.
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose without fee is hereby granted, provided that this entire notice
+ * is included in all copies of any software which is or includes a copy
+ * or modification of this software and in all copies of the supporting
+ * documentation for such software.
+ * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR IMPLIED
+ * WARRANTY.  IN PARTICULAR, NEITHER THE AUTHORS NOR LUCENT TECHNOLOGIES MAKE ANY
+ * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE MERCHANTABILITY
+ * OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR PURPOSE.
+ */
+#include "lib9.h"
+
+char*
+seprint(char *buf, char *e, char *fmt, ...)
+{
+	char *p;
+	va_list args;
+
+	va_start(args, fmt);
+	p = vseprint(buf, e, fmt, args);
+	va_end(args);
+	return p;
+}
diff --git a/libkern/sin.c b/libkern/sin.c
new file mode 100644
index 00000000..70f90a9d
--- /dev/null
+++ b/libkern/sin.c
@@ -0,0 +1,67 @@
+/*
+	C program for floating point sin/cos.
+	Calls modf.
+	There are no error exits.
+	Coefficients are #3370 from Hart & Cheney (18.80D).
+*/
+
+#include <lib9.h>
+
+#define p0      .1357884097877375669092680e8
+#define p1     -.4942908100902844161158627e7
+#define p2      .4401030535375266501944918e6
+#define p3     -.1384727249982452873054457e5
+#define p4      .1459688406665768722226959e3
+#define q0      .8644558652922534429915149e7
+#define q1      .4081792252343299749395779e6
+#define q2      .9463096101538208180571257e4
+#define q3      .1326534908786136358911494e3
+
+static
+double
+sinus(double arg, int quad)
+{
+	double e, f, ysq, x, y, temp1, temp2;
+	int k;
+
+	x = arg;
+	if(x < 0) {
+		x = -x;
+		quad += 2;
+	}
+	x *= 1/PIO2;	/* underflow? */
+	if(x > 32764) {
+		y = modf(x, &e);
+		e += quad;
+		modf(0.25*e, &f);
+		quad = e - 4*f;
+	} else {
+		k = x;
+		y = x - k;
+		quad += k;
+		quad &= 3;
+	}
+	if(quad & 1)
+		y = 1-y;
+	if(quad > 1)
+		y = -y;
+
+	ysq = y*y;
+	temp1 = ((((p4*ysq+p3)*ysq+p2)*ysq+p1)*ysq+p0)*y;
+	temp2 = ((((ysq+q3)*ysq+q2)*ysq+q1)*ysq+q0);
+	return temp1/temp2;
+}
+
+double
+cos(double arg)
+{
+	if(arg < 0)
+		arg = -arg;
+	return sinus(arg, 1);
+}
+
+double
+sin(double arg)
+{
+	return sinus(arg, 0);
+}
diff --git a/libkern/smprint.c b/libkern/smprint.c
new file mode 100644
index 00000000..ecb55492
--- /dev/null
+++ b/libkern/smprint.c
@@ -0,0 +1,26 @@
+/*
+ * The authors of this software are Rob Pike and Ken Thompson.
+ *              Copyright (c) 2002 by Lucent Technologies.
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose without fee is hereby granted, provided that this entire notice
+ * is included in all copies of any software which is or includes a copy
+ * or modification of this software and in all copies of the supporting
+ * documentation for such software.
+ * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR IMPLIED
+ * WARRANTY.  IN PARTICULAR, NEITHER THE AUTHORS NOR LUCENT TECHNOLOGIES MAKE ANY
+ * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE MERCHANTABILITY
+ * OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR PURPOSE.
+ */
+#include "lib9.h"
+
+char*
+smprint(char *fmt, ...)
+{
+	va_list args;
+	char *p;
+
+	va_start(args, fmt);
+	p = vsmprint(fmt, args);
+	va_end(args);
+	return p;
+}
diff --git a/libkern/snprint.c b/libkern/snprint.c
new file mode 100644
index 00000000..6e65c70c
--- /dev/null
+++ b/libkern/snprint.c
@@ -0,0 +1,27 @@
+/*
+ * The authors of this software are Rob Pike and Ken Thompson.
+ *              Copyright (c) 2002 by Lucent Technologies.
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose without fee is hereby granted, provided that this entire notice
+ * is included in all copies of any software which is or includes a copy
+ * or modification of this software and in all copies of the supporting
+ * documentation for such software.
+ * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR IMPLIED
+ * WARRANTY.  IN PARTICULAR, NEITHER THE AUTHORS NOR LUCENT TECHNOLOGIES MAKE ANY
+ * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE MERCHANTABILITY
+ * OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR PURPOSE.
+ */
+#include "lib9.h"
+
+int
+snprint(char *buf, int len, char *fmt, ...)
+{
+	int n;
+	va_list args;
+
+	va_start(args, fmt);
+	n = vsnprint(buf, len, fmt, args);
+	va_end(args);
+	return n;
+}
+
diff --git a/libkern/sqrt.c b/libkern/sqrt.c
new file mode 100644
index 00000000..0d235126
--- /dev/null
+++ b/libkern/sqrt.c
@@ -0,0 +1,53 @@
+/*
+	sqrt returns the square root of its floating
+	point argument. Newton's method.
+
+	calls frexp
+*/
+
+#include <lib9.h>
+
+double
+sqrt(double arg)
+{
+	double x, temp;
+	int exp, i;
+
+	if(arg <= 0) {
+		if(arg < 0)
+			return NaN();
+		return 0;
+	}
+	if(isInf(arg, 1))
+		return arg;
+	x = frexp(arg, &exp);
+	while(x < 0.5) {
+		x *= 2;
+		exp--;
+	}
+	/*
+	 * NOTE
+	 * this wont work on 1's comp
+	 */
+	if(exp & 1) {
+		x *= 2;
+		exp--;
+	}
+	temp = 0.5 * (1.0+x);
+
+	while(exp > 60) {
+		temp *= (1L<<30);
+		exp -= 60;
+	}
+	while(exp < -60) {
+		temp /= (1L<<30);
+		exp += 60;
+	}
+	if(exp >= 0)
+		temp *= 1L << (exp/2);
+	else
+		temp /= 1L << (-exp/2);
+	for(i=0; i<=4; i++)
+		temp = 0.5*(temp + arg/temp);
+	return temp;
+}
diff --git a/libkern/strcat.c b/libkern/strcat.c
new file mode 100644
index 00000000..9b1b3046
--- /dev/null
+++ b/libkern/strcat.c
@@ -0,0 +1,9 @@
+#include <lib9.h>
+
+char*
+strcat(char *s1, char *s2)
+{
+
+	strcpy(strchr(s1, 0), s2);
+	return s1;
+}
diff --git a/libkern/strchr-386.s b/libkern/strchr-386.s
new file mode 100644
index 00000000..e437caf2
--- /dev/null
+++ b/libkern/strchr-386.s
@@ -0,0 +1,29 @@
+	TEXT	strchr(SB),$0
+/*
+ * look for null
+ */
+	MOVL	p+0(FP), DI
+	MOVL	$-1, CX
+	MOVL	$0, AX
+	CLD
+
+	REPN;	SCASB
+
+/*
+ * look for real char
+ */
+	MOVL	DI, CX
+	MOVL	p+0(FP), DI
+	SUBL	DI, CX
+	MOVBLZX	c+4(FP), AX
+
+	REPN;	SCASB
+
+	JEQ	found
+	MOVL	$0, AX
+	RET
+
+found:
+	MOVL	DI, AX
+	SUBL	$1, AX
+	RET
diff --git a/libkern/strchr-68000.s b/libkern/strchr-68000.s
new file mode 100644
index 00000000..a5d4be2c
--- /dev/null
+++ b/libkern/strchr-68000.s
@@ -0,0 +1,27 @@
+	TEXT	strchr(SB), $0
+
+	MOVL	s+0(FP), A0
+	MOVB	c+7(FP), R2
+	BEQ	null
+
+l:
+	MOVB	(A0)+, R1
+	BEQ	out
+	CMPB	R1, R2
+	BNE	l
+
+	MOVL	A0, R0
+	ADDL	$-1, R0
+	RTS
+
+out:
+	CLRL	R0
+	RTS
+
+null:
+	TSTB	(A0)+
+	BNE	null
+
+	MOVL	A0, R0
+	ADDL	$-1, R0
+	RTS
diff --git a/libkern/strchr-arm.s b/libkern/strchr-arm.s
new file mode 100644
index 00000000..9c04273c
--- /dev/null
+++ b/libkern/strchr-arm.s
@@ -0,0 +1,52 @@
+TEXT strchr(SB), $-4
+	MOVBU	c+4(FP), R1
+	CMP	$0, R1
+	BEQ	_null
+
+_strchr:				/* not looking for a null, byte at a time */
+	MOVBU.P	1(R0), R2
+	CMP	R1, R2
+	BEQ	_sub1
+
+	CMP	$0, R2
+	BNE	_strchr
+
+_return0:				/* character not found in string, return 0 */
+	MOVW	$0, R0
+	RET
+
+_null:					/* looking for null, align */
+	AND.S	$3, R0, R2
+	BEQ	_aligned
+
+	MOVBU.P	1(R0), R4
+	CMP	$0, R4
+	BEQ	_sub1
+	B	_null
+
+_aligned:
+	MOVW	$0xFF, R3		/* mask */
+
+_loop:
+	MOVW.P	4(R0), R4		/* 4 at a time */
+	TST	R4, R3			/* AND.S R2, R3, Rx */
+	BEQ	_sub4
+	TST	R4>>8, R3
+	BEQ	_sub3
+	TST	R4>>16, R3
+	BEQ	_sub2
+	TST	R4>>24, R3
+	BNE	_loop
+
+_sub1:					/* compensate for pointer increment */
+	SUB	$1, R0
+	RET
+_sub2:
+	SUB	$2, R0
+	RET
+_sub3:
+	SUB	$3, R0
+	RET
+_sub4:
+	SUB	$4, R0
+	RET
diff --git a/libkern/strchr-mips.c b/libkern/strchr-mips.c
new file mode 100644
index 00000000..e69de29b
--- /dev/null
+++ b/libkern/strchr-mips.c
diff --git a/libkern/strchr-mips.s b/libkern/strchr-mips.s
new file mode 100644
index 00000000..e009ffad
--- /dev/null
+++ b/libkern/strchr-mips.s
@@ -0,0 +1,63 @@
+	TEXT	strchr(SB), $0
+MOVW R1, 0(FP)
+	MOVB	c+7(FP), R4
+	MOVW	s+0(FP), R3
+
+	BEQ	R4, l2
+
+/*
+ * char is not null
+ */
+l1:
+	MOVB	(R3), R1
+	ADDU	$1, R3
+	BEQ	R1, ret
+	BNE	R1,R4, l1
+	JMP	rm1
+
+/*
+ * char is null
+ * align to word
+ */
+l2:
+	AND	$3,R3, R1
+	BEQ	R1, l3
+	MOVB	(R3), R1
+	ADDU	$1, R3
+	BNE	R1, l2
+	JMP	rm1
+
+l3:
+	MOVW	$0xff000000, R6
+	MOVW	$0x00ff0000, R7
+
+l4:
+	MOVW	(R3), R5
+	ADDU	$4, R3
+	AND	R6,R5, R1
+	AND	R7,R5, R2
+	BEQ	R1, b0
+	AND	$0xff00,R5, R1
+	BEQ	R2, b1
+	AND	$0xff,R5, R2
+	BEQ	R1, b2
+	BNE	R2, l4
+
+rm1:
+	ADDU	$-1,R3, R1
+	JMP	ret
+
+b2:
+	ADDU	$-2,R3, R1
+	JMP	ret
+
+b1:
+	ADDU	$-3,R3, R1
+	JMP	ret
+
+b0:
+	ADDU	$-4,R3, R1
+	JMP	ret
+
+ret:
+	RET
diff --git a/libkern/strchr-power.s b/libkern/strchr-power.s
new file mode 100644
index 00000000..1cfc0bbe
--- /dev/null
+++ b/libkern/strchr-power.s
@@ -0,0 +1,16 @@
+/*
+ * BUG: it's slow
+ */
+	TEXT	strchr(SB), $0
+	MOVBZ	c+7(FP), R4
+	SUB	$1, R3
+l1:
+	MOVBZU	1(R3), R6
+	CMP	R6, R4
+	BEQ	eq
+	CMP	R6, $0
+	BNE	l1
+nf:
+	MOVW	$0, R3
+eq:
+	RETURN
diff --git a/libkern/strchr-sparc.s b/libkern/strchr-sparc.s
new file mode 100644
index 00000000..192beab1
--- /dev/null
+++ b/libkern/strchr-sparc.s
@@ -0,0 +1,73 @@
+	TEXT	strchr(SB), $0
+
+MOVW	R7, 0(FP)
+	MOVB	c+7(FP), R10
+	MOVW	s+0(FP), R9
+
+	SUBCC	R0,R10, R0
+	BE	l2
+
+/*
+ * char is not null
+ */
+l1:
+	MOVB	(R9), R7
+	ADD	$1, R9
+	SUBCC	R0,R7, R0
+	BE	ret
+	SUBCC	R7,R10, R0
+	BNE	l1
+	JMP	rm1
+
+/*
+ * char is null
+ * align to word
+ */
+l2:
+	ANDCC	$3,R9, R0
+	BE	l3
+	MOVB	(R9), R7
+	ADD	$1, R9
+	SUBCC	R0,R7, R0
+	BNE	l2
+	JMP	rm1
+
+/*
+ * develop byte masks
+ */
+l3:
+	MOVW	$0xff, R17
+	SLL	$8,R17, R16
+	SLL	$16,R17, R13
+	SLL	$24,R17, R12
+
+l4:
+	MOVW	(R9), R11
+	ADD	$4, R9
+	ANDCC	R12,R11, R0
+	BE	b0
+	ANDCC	R13,R11, R0
+	BE	b1
+	ANDCC	R16,R11, R0
+	BE	b2
+	ANDCC	R17,R11, R0
+	BNE	l4
+
+rm1:
+	SUB	$1,R9, R7
+	JMP	ret
+
+b2:
+	SUB	$2,R9, R7
+	JMP	ret
+
+b1:
+	SUB	$3,R9, R7
+	JMP	ret
+
+b0:
+	SUB	$4,R9, R7
+	JMP	ret
+
+ret:
+	RETURN
diff --git a/libkern/strchr-thumb.s b/libkern/strchr-thumb.s
new file mode 100644
index 00000000..9c04273c
--- /dev/null
+++ b/libkern/strchr-thumb.s
@@ -0,0 +1,52 @@
+TEXT strchr(SB), $-4
+	MOVBU	c+4(FP), R1
+	CMP	$0, R1
+	BEQ	_null
+
+_strchr:				/* not looking for a null, byte at a time */
+	MOVBU.P	1(R0), R2
+	CMP	R1, R2
+	BEQ	_sub1
+
+	CMP	$0, R2
+	BNE	_strchr
+
+_return0:				/* character not found in string, return 0 */
+	MOVW	$0, R0
+	RET
+
+_null:					/* looking for null, align */
+	AND.S	$3, R0, R2
+	BEQ	_aligned
+
+	MOVBU.P	1(R0), R4
+	CMP	$0, R4
+	BEQ	_sub1
+	B	_null
+
+_aligned:
+	MOVW	$0xFF, R3		/* mask */
+
+_loop:
+	MOVW.P	4(R0), R4		/* 4 at a time */
+	TST	R4, R3			/* AND.S R2, R3, Rx */
+	BEQ	_sub4
+	TST	R4>>8, R3
+	BEQ	_sub3
+	TST	R4>>16, R3
+	BEQ	_sub2
+	TST	R4>>24, R3
+	BNE	_loop
+
+_sub1:					/* compensate for pointer increment */
+	SUB	$1, R0
+	RET
+_sub2:
+	SUB	$2, R0
+	RET
+_sub3:
+	SUB	$3, R0
+	RET
+_sub4:
+	SUB	$4, R0
+	RET
diff --git a/libkern/strchr.c b/libkern/strchr.c
new file mode 100644
index 00000000..1d4d425c
--- /dev/null
+++ b/libkern/strchr.c
@@ -0,0 +1,18 @@
+#include <lib9.h>
+
+char*
+strchr(char *s, int c)
+{
+	char c1;
+
+	if(c == 0) {
+		while(*s++)
+			;
+		return s-1;
+	}
+
+	while(c1 = *s++)
+		if(c1 == c)
+			return s-1;
+	return 0;
+}
diff --git a/libkern/strcmp-power.s b/libkern/strcmp-power.s
new file mode 100644
index 00000000..0aef5b29
--- /dev/null
+++ b/libkern/strcmp-power.s
@@ -0,0 +1,21 @@
+TEXT	strcmp(SB), $0
+
+	MOVW	s2+4(FP), R4
+
+	SUB	$1, R3
+	SUB	$1, R4
+l1:
+	MOVBZU	1(R3), R5
+	MOVBZU	1(R4), R6
+	CMP	R5, R6
+	BNE	ne
+	CMP	R5, $0
+	BNE	l1
+	MOVW	$0, R3
+	RETURN
+ne:
+	MOVW	$1, R3
+	BGT	ret
+	MOVW	$-1, R3
+ret:
+	RETURN
diff --git a/libkern/strcmp.c b/libkern/strcmp.c
new file mode 100644
index 00000000..f2a7e581
--- /dev/null
+++ b/libkern/strcmp.c
@@ -0,0 +1,19 @@
+#include <lib9.h>
+
+int
+strcmp(char *s1, char *s2)
+{
+	unsigned c1, c2;
+
+	for(;;) {
+		c1 = *s1++;
+		c2 = *s2++;
+		if(c1 != c2) {
+			if(c1 > c2)
+				return 1;
+			return -1;
+		}
+		if(c1 == 0)
+			return 0;
+	}
+}
diff --git a/libkern/strcpy.c b/libkern/strcpy.c
new file mode 100644
index 00000000..473a5eac
--- /dev/null
+++ b/libkern/strcpy.c
@@ -0,0 +1,15 @@
+#include <lib9.h>
+#define	N	10000
+
+char*
+strcpy(char *s1, char *s2)
+{
+	char *os1;
+
+	os1 = s1;
+	while(!memccpy(s1, s2, 0, N)) {
+		s1 += N;
+		s2 += N;
+	}
+	return os1;
+}
diff --git a/libkern/strdup.c b/libkern/strdup.c
new file mode 100644
index 00000000..2d6eaef5
--- /dev/null
+++ b/libkern/strdup.c
@@ -0,0 +1,12 @@
+#include <lib9.h>
+
+char*
+strdup(char *s) 
+{  
+	char *os;
+
+	os = malloc(strlen(s) + 1);
+	if(os == 0)
+		return 0;
+	return strcpy(os, s);
+}
diff --git a/libkern/strecpy.c b/libkern/strecpy.c
new file mode 100644
index 00000000..7d2f2277
--- /dev/null
+++ b/libkern/strecpy.c
@@ -0,0 +1,16 @@
+#include <lib9.h>
+
+char*
+strecpy(char *to, char *e, char *from)
+{
+	if(to >= e)
+		return to;
+	to = memccpy(to, from, '\0', e - to);
+	if(to == nil){
+		to = e - 1;
+		*to = '\0';
+	}else{
+		to--;
+	}
+	return to;
+}
diff --git a/libkern/strlen.c b/libkern/strlen.c
new file mode 100644
index 00000000..5cac06f2
--- /dev/null
+++ b/libkern/strlen.c
@@ -0,0 +1,8 @@
+#include <lib9.h>
+
+long
+strlen(char *s)
+{
+
+	return strchr(s, 0) - s;
+}
diff --git a/libkern/strncmp-power.s b/libkern/strncmp-power.s
new file mode 100644
index 00000000..c55962fa
--- /dev/null
+++ b/libkern/strncmp-power.s
@@ -0,0 +1,29 @@
+TEXT	strncmp(SB), $0
+#define	BDNZ	BC	16,0,
+
+	MOVW	s2+4(FP), R4
+	MOVW	n+8(FP), R7
+
+	CMP	R7, $0
+	MOVW	R7, CTR
+	BLE	eq
+
+	SUB	$1, R3
+	SUB	$1, R4
+l1:
+	MOVBZU	1(R3), R5
+	MOVBZU	1(R4), R6
+	CMP	R5, R6
+	BNE	ne
+	CMP	R5, $0
+	BEQ	eq
+	BDNZ	l1
+eq:
+	MOVW	$0, R3
+	RETURN
+ne:
+	MOVW	$1, R3
+	BGT	ret
+	MOVW	$-1, R3
+ret:
+	RETURN
diff --git a/libkern/strncmp.c b/libkern/strncmp.c
new file mode 100644
index 00000000..ef926489
--- /dev/null
+++ b/libkern/strncmp.c
@@ -0,0 +1,21 @@
+#include <lib9.h>
+
+int
+strncmp(char *s1, char *s2, long n)
+{
+	unsigned c1, c2;
+
+	while(n > 0) {
+		c1 = *s1++;
+		c2 = *s2++;
+		n--;
+		if(c1 != c2) {
+			if(c1 > c2)
+				return 1;
+			return -1;
+		}
+		if(c1 == 0)
+			break;
+	}
+	return 0;
+}
diff --git a/libkern/strncpy.c b/libkern/strncpy.c
new file mode 100644
index 00000000..44f85d6f
--- /dev/null
+++ b/libkern/strncpy.c
@@ -0,0 +1,17 @@
+#include <lib9.h>
+
+char*
+strncpy(char *s1, char *s2, long n)
+{
+	int i;
+	char *os1;
+
+	os1 = s1;
+	for(i = 0; i < n; i++)
+		if((*s1++ = *s2++) == 0) {
+			while(++i < n)
+				*s1++ = 0;
+			return os1;
+		}
+	return os1;
+}
diff --git a/libkern/strrchr.c b/libkern/strrchr.c
new file mode 100644
index 00000000..fc28ede4
--- /dev/null
+++ b/libkern/strrchr.c
@@ -0,0 +1,14 @@
+#include <lib9.h>
+
+char*
+strrchr(char *s, int c)
+{
+	char *r;
+
+	if(c == 0)
+		return strchr(s, 0);
+	r = 0;
+	while(s = strchr(s, c))
+		r = s++;
+	return r;
+}
diff --git a/libkern/strstr.c b/libkern/strstr.c
new file mode 100644
index 00000000..b9d8a35e
--- /dev/null
+++ b/libkern/strstr.c
@@ -0,0 +1,21 @@
+#include <lib9.h>
+
+/*
+ * Return pointer to first occurrence of s2 in s1,
+ * 0 if none
+ */
+char*
+strstr(char *s1, char *s2)
+{
+	char *p;
+	int f, n;
+
+	f = s2[0];
+	if(f == 0)
+		return s1;
+	n = strlen(s2);
+	for(p=strchr(s1, f); p; p=strchr(p+1, f))
+		if(strncmp(p, s2, n) == 0)
+			return p;
+	return 0;
+}
diff --git a/libkern/strtod.c b/libkern/strtod.c
new file mode 100644
index 00000000..c591e44c
--- /dev/null
+++ b/libkern/strtod.c
@@ -0,0 +1,31 @@
+#include <lib9.h>
+
+static int
+strtodf(void *vp)
+{
+	return *(*((char**)vp))++;
+}
+
+double
+strtod(char *s, char **end)
+{
+	double d;
+	char *ss;
+	int c;
+
+	ss = s;
+	d = charstod(strtodf, &s);
+	/*
+	 * Fix cases like 2.3e+ , which charstod will consume
+	 */
+	if(end){
+		*end = --s;
+		while(s > ss){
+			c = *--s;
+			if(c!='-' && c!='+' && c!='e' && c!='E')
+				break;
+			(*end)--;
+		}
+	}
+	return d;
+}
diff --git a/libkern/strtol.c b/libkern/strtol.c
new file mode 100644
index 00000000..a519f02d
--- /dev/null
+++ b/libkern/strtol.c
@@ -0,0 +1,94 @@
+#include <lib9.h>
+
+#define LONG_MAX	2147483647L
+#define LONG_MIN	-2147483648L
+
+long
+strtol(char *nptr, char **endptr, int base)
+{
+	char *p;
+	long n, nn;
+	int c, ovfl, v, neg, ndig;
+
+	p = nptr;
+	neg = 0;
+	n = 0;
+	ndig = 0;
+	ovfl = 0;
+
+	/*
+	 * White space
+	 */
+	for(;;p++){
+		switch(*p){
+		case ' ':
+		case '\t':
+		case '\n':
+		case '\f':
+		case '\r':
+		case '\v':
+			continue;
+		}
+		break;
+	}
+
+	/*
+	 * Sign
+	 */
+	if(*p=='-' || *p=='+')
+		if(*p++ == '-')
+			neg = 1;
+
+	/*
+	 * Base
+	 */
+	if(base==0){
+		if(*p != '0')
+			base = 10;
+		else{
+			base = 8;
+			if(p[1]=='x' || p[1]=='X'){
+				p += 2;
+				base = 16;
+			}
+		}
+	}else if(base==16 && *p=='0'){
+		if(p[1]=='x' || p[1]=='X')
+			p += 2;
+	}else if(base<0 || 36<base)
+		goto Return;
+
+	/*
+	 * Non-empty sequence of digits
+	 */
+	for(;; p++,ndig++){
+		c = *p;
+		v = base;
+		if('0'<=c && c<='9')
+			v = c - '0';
+		else if('a'<=c && c<='z')
+			v = c - 'a' + 10;
+		else if('A'<=c && c<='Z')
+			v = c - 'A' + 10;
+		if(v >= base)
+			break;
+		nn = n*base + v;
+		if(nn < n)
+			ovfl = 1;
+		n = nn;
+	}
+
+    Return:
+	if(ndig == 0)
+		p = nptr;
+	if(endptr)
+		*endptr = p;
+	if(ovfl){
+		if(neg)
+			return LONG_MIN;
+		return LONG_MAX;
+	}
+	if(neg)
+		return -n;
+	return n;
+}
diff --git a/libkern/strtoll.c b/libkern/strtoll.c
new file mode 100644
index 00000000..ff98e1c4
--- /dev/null
+++ b/libkern/strtoll.c
@@ -0,0 +1,81 @@
+#include "lib9.h"
+
+vlong
+strtoll(const char *nptr, char **endptr, int base)
+{
+	const char *p;
+	vlong n;
+	int c, v, neg, ndig;
+
+	p = nptr;
+	neg = 0;
+	n = 0;
+	ndig = 0;
+
+	/*
+	 * White space
+	 */
+	for(;;p++){
+		switch(*p){
+		case ' ':
+		case '\t':
+		case '\n':
+		case '\f':
+		case '\r':
+		case '\v':
+			continue;
+		}
+		break;
+	}
+
+	/*
+	 * Sign
+	 */
+	if(*p=='-' || *p=='+')
+		if(*p++ == '-')
+			neg = 1;
+
+	/*
+	 * Base
+	 */
+	if(base==0){
+		if(*p != '0')
+			base = 10;
+		else{
+			base = 8;
+			if(p[1]=='x' || p[1]=='X'){
+				p += 2;
+				base = 16;
+			}
+		}
+	}else if(base==16 && *p=='0'){
+		if(p[1]=='x' || p[1]=='X')
+			p += 2;
+	}else if(base<0 || 36<base)
+		goto Return;
+
+	/*
+	 * Non-empty sequence of digits
+	 */
+	for(;; p++,ndig++){
+		c = *p;
+		v = base;
+		if('0'<=c && c<='9')
+			v = c - '0';
+		else if('a'<=c && c<='z')
+			v = c - 'a' + 10;
+		else if('A'<=c && c<='Z')
+			v = c - 'A' + 10;
+		if(v >= base)
+			break;
+		n = n*base + v;
+	}
+    Return:
+	if(ndig == 0)
+		p = nptr;
+	if(endptr)
+		*endptr = (char*) p;
+	if(neg)
+		return -n;
+	return n;
+}
diff --git a/libkern/strtoul.c b/libkern/strtoul.c
new file mode 100644
index 00000000..bdf2fc08
--- /dev/null
+++ b/libkern/strtoul.c
@@ -0,0 +1,96 @@
+#include <lib9.h>
+
+#define ULONG_MAX	4294967295UL
+
+ulong
+strtoul(char *nptr, char **endptr, int base)
+{
+	char *p;
+	ulong n, nn, m;
+	int c, ovfl, neg, v, ndig;
+
+	p = (char*)nptr;
+	neg = 0;
+	n = 0;
+	ndig = 0;
+	ovfl = 0;
+
+	/*
+	 * White space
+	 */
+	for(;;p++){
+		switch(*p){
+		case ' ':
+		case '\t':
+		case '\n':
+		case '\f':
+		case '\r':
+		case '\v':
+			continue;
+		}
+		break;
+	}
+
+	/*
+	 * Sign
+	 */
+	if(*p=='-' || *p=='+')
+		if(*p++ == '-')
+			neg = 1;
+
+	/*
+	 * Base
+	 */
+	if(base==0){
+		if(*p != '0')
+			base = 10;
+		else{
+			base = 8;
+			if(p[1]=='x' || p[1]=='X')
+				base = 16;
+		}
+	}
+ 	if(base<2 || 36<base)
+		goto Return;
+	if(base==16 && *p=='0'){
+		if(p[1]=='x' || p[1]=='X')
+			if(('0' <= p[2] && p[2] <= '9')
+			 ||('a' <= p[2] && p[2] <= 'f')
+			 ||('A' <= p[2] && p[2] <= 'F'))
+				p += 2;
+	}
+	/*
+	 * Non-empty sequence of digits
+	 */
+	n = 0;
+	m = ULONG_MAX/base;
+	for(;; p++,ndig++){
+		c = *p;
+		v = base;
+		if('0'<=c && c<='9')
+			v = c - '0';
+		else if('a'<=c && c<='z')
+			v = c - 'a' + 10;
+		else if('A'<=c && c<='Z')
+			v = c - 'A' + 10;
+		if(v >= base)
+			break;
+		if(n > m)
+			ovfl = 1;
+		nn = n*base + v;
+		if(nn < n)
+			ovfl = 1;
+		n = nn;
+	}
+
+    Return:
+	if(ndig == 0)
+		p = nptr;
+	if(endptr)
+		*endptr = p;
+	if(ovfl)
+		return ULONG_MAX;
+	if(neg)
+		return -n;
+	return n;
+}
diff --git a/libkern/tokenize.c b/libkern/tokenize.c
new file mode 100644
index 00000000..1b222c52
--- /dev/null
+++ b/libkern/tokenize.c
@@ -0,0 +1,58 @@
+#include "lib9.h"
+
+static char qsep[] = " \t\r\n";
+
+static char*
+qtoken(char *s)
+{
+	int quoting;
+	char *t;
+
+	quoting = 0;
+	t = s;	/* s is output string, t is input string */
+	while(*t!='\0' && (quoting || utfrune(qsep, *t)==nil)){
+		if(*t != '\''){
+			*s++ = *t++;
+			continue;
+		}
+		/* *t is a quote */
+		if(!quoting){
+			quoting = 1;
+			t++;
+			continue;
+		}
+		/* quoting and we're on a quote */
+		if(t[1] != '\''){
+			/* end of quoted section; absorb closing quote */
+			t++;
+			quoting = 0;
+			continue;
+		}
+		/* doubled quote; fold one quote into two */
+		t++;
+		*s++ = *t++;
+	}
+	if(*s != '\0'){
+		*s = '\0';
+		if(t == s)
+			t++;
+	}
+	return t;
+}
+
+int
+tokenize(char *s, char **args, int maxargs)
+{
+	int nargs;
+
+	for(nargs=0; nargs<maxargs; nargs++){
+		while(*s!='\0' && utfrune(qsep, *s)!=nil)
+			s++;
+		if(*s == '\0')
+			break;
+		args[nargs] = s;
+		s = qtoken(s);
+	}
+
+	return nargs;
+}
diff --git a/libkern/toupper.c b/libkern/toupper.c
new file mode 100644
index 00000000..4c0b02ac
--- /dev/null
+++ b/libkern/toupper.c
@@ -0,0 +1,16 @@
+toupper(int c)
+{
+
+	if(c < 'a' || c > 'z')
+		return c;
+	return (c-'a'+'A');
+}
+
+tolower(int c)
+{
+
+	if(c < 'A' || c > 'Z')
+		return c;
+	return (c-'A'+'a');
+}
+
diff --git a/libkern/u16.c b/libkern/u16.c
new file mode 100644
index 00000000..d9f41e46
--- /dev/null
+++ b/libkern/u16.c
@@ -0,0 +1,52 @@
+#include <lib9.h>
+static char t16e[] = "0123456789ABCDEF";
+
+int
+dec16(uchar *out, int lim, char *in, int n)
+{
+	int c, w = 0, i = 0;
+	uchar *start = out;
+	uchar *eout = out + lim;
+
+	while(n-- > 0){
+		c = *in++;
+		if('0' <= c && c <= '9')
+			c = c - '0';
+		else if('a' <= c && c <= 'z')
+			c = c - 'a' + 10;
+		else if('A' <= c && c <= 'Z')
+			c = c - 'A' + 10;
+		else
+			continue;
+		w = (w<<4) + c;
+		i++;
+		if(i == 2){
+			if(out + 1 > eout)
+				goto exhausted;
+			*out++ = w;
+			w = 0;
+			i = 0;
+		}
+	}
+exhausted:
+	return out - start;
+}
+
+int
+enc16(char *out, int lim, uchar *in, int n)
+{
+	uint c;
+	char *eout = out + lim;
+	char *start = out;
+
+	while(n-- > 0){
+		c = *in++;
+		if(out + 2 >= eout)
+			goto exhausted;
+		*out++ = t16e[c>>4];
+		*out++ = t16e[c&0xf];
+	}
+exhausted:
+	*out = 0;
+	return out - start;
+}
diff --git a/libkern/u32.c b/libkern/u32.c
new file mode 100644
index 00000000..1eb0c6e0
--- /dev/null
+++ b/libkern/u32.c
@@ -0,0 +1,109 @@
+#include <lib9.h>
+
+int
+dec32(uchar *dest, int ndest, char *src, int nsrc)
+{
+	char *s, *tab;
+	uchar *start;
+	int i, u[8];
+
+	if(ndest+1 < (5*nsrc+7)/8)
+		return -1;
+	start = dest;
+	tab = "23456789abcdefghijkmnpqrstuvwxyz";
+	while(nsrc>=8){
+		for(i=0; i<8; i++){
+			s = strchr(tab,(int)src[i]);
+			u[i] = s ? s-tab : 0;
+		}
+		*dest++ = (u[0]<<3) | (0x7 & (u[1]>>2));
+		*dest++ = ((0x3 & u[1])<<6) | (u[2]<<1) | (0x1 & (u[3]>>4));
+		*dest++ = ((0xf & u[3])<<4) | (0xf & (u[4]>>1));
+		*dest++ = ((0x1 & u[4])<<7) | (u[5]<<2) | (0x3 & (u[6]>>3));
+		*dest++ = ((0x7 & u[6])<<5) | u[7];
+		src  += 8;
+		nsrc -= 8;
+	}
+	if(nsrc > 0){
+		if(nsrc == 1 || nsrc == 3 || nsrc == 6)
+			return -1;
+		for(i=0; i<nsrc; i++){
+			s = strchr(tab,(int)src[i]);
+			u[i] = s ? s-tab : 0;
+		}
+		*dest++ = (u[0]<<3) | (0x7 & (u[1]>>2));
+		if(nsrc == 2)
+			goto out;
+		*dest++ = ((0x3 & u[1])<<6) | (u[2]<<1) | (0x1 & (u[3]>>4));
+		if(nsrc == 4)
+			goto out;
+		*dest++ = ((0xf & u[3])<<4) | (0xf & (u[4]>>1));
+		if(nsrc == 5)
+			goto out;
+		*dest++ = ((0x1 & u[4])<<7) | (u[5]<<2) | (0x3 & (u[6]>>3));
+	}
+out:
+	return dest-start;
+}
+
+int
+enc32(char *dest, int ndest, uchar *src, int nsrc)
+{
+	char *tab, *start;
+	int j;
+
+	if(ndest <= (8*nsrc+4)/5 )
+		return -1;
+	start = dest;
+	tab = "23456789abcdefghijkmnpqrstuvwxyz";
+	while(nsrc>=5){
+		j = (0x1f & (src[0]>>3));
+		*dest++ = tab[j];
+		j = (0x1c & (src[0]<<2)) | (0x03 & (src[1]>>6));
+		*dest++ = tab[j];
+		j = (0x1f & (src[1]>>1));
+		*dest++ = tab[j];
+		j = (0x10 & (src[1]<<4)) | (0x0f & (src[2]>>4));
+		*dest++ = tab[j];
+		j = (0x1e & (src[2]<<1)) | (0x01 & (src[3]>>7));
+		*dest++ = tab[j];
+		j = (0x1f & (src[3]>>2));
+		*dest++ = tab[j];
+		j = (0x18 & (src[3]<<3)) | (0x07 & (src[4]>>5));
+		*dest++ = tab[j];
+		j = (0x1f & (src[4]));
+		*dest++ = tab[j];
+		src  += 5;
+		nsrc -= 5;
+	}
+	if(nsrc){
+		j = (0x1f & (src[0]>>3));
+		*dest++ = tab[j];
+		j = (0x1c & (src[0]<<2));
+		if(nsrc == 1)
+			goto out;
+		j |= (0x03 & (src[1]>>6));
+		*dest++ = tab[j];
+		j = (0x1f & (src[1]>>1));
+		if(nsrc == 2)
+			goto out;
+		*dest++ = tab[j];
+		j = (0x10 & (src[1]<<4));
+		if(nsrc == 3)
+			goto out;
+		j |= (0x0f & (src[2]>>4));
+		*dest++ = tab[j];
+		j = (0x1e & (src[2]<<1));
+		if(nsrc == 4)
+			goto out;
+		j |= (0x01 & (src[3]>>7));
+		*dest++ = tab[j];
+		j = (0x1f & (src[3]>>2));
+		*dest++ = tab[j];
+		j = (0x18 & (src[3]<<3));
+out:
+		*dest++ = tab[j];
+	}
+	*dest = 0;
+	return dest-start;
+}
diff --git a/libkern/u64.c b/libkern/u64.c
new file mode 100644
index 00000000..a17bdf1d
--- /dev/null
+++ b/libkern/u64.c
@@ -0,0 +1,126 @@
+#include <lib9.h>
+
+enum {
+	INVAL=	255
+};
+
+static uchar t64d[256] = {
+   INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,
+   INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,
+   INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,   62,INVAL,INVAL,INVAL,   63,
+      52,   53,   54,   55,   56,   57,   58,   59,   60,   61,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,
+   INVAL,    0,    1,    2,    3,    4,    5,    6,    7,    8,    9,   10,   11,   12,   13,   14,
+      15,   16,   17,   18,   19,   20,   21,   22,   23,   24,   25,INVAL,INVAL,INVAL,INVAL,INVAL,
+   INVAL,   26,   27,   28,   29,   30,   31,   32,   33,   34,   35,   36,   37,   38,   39,   40,
+      41,   42,   43,   44,   45,   46,   47,   48,   49,   50,   51,INVAL,INVAL,INVAL,INVAL,INVAL,
+   INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,
+   INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,
+   INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,
+   INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,
+   INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,
+   INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,
+   INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,
+   INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL,INVAL
+};
+static char t64e[] = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/";
+
+int
+dec64(uchar *out, int lim, char *in, int n)
+{
+	ulong b24;
+	uchar *start = out;
+	uchar *e = out + lim;
+	int i, c;
+
+	b24 = 0;
+	i = 0;
+	while(n-- > 0){
+ 
+		c = t64d[*(uchar*)in++];
+		if(c == INVAL)
+			continue;
+		switch(i){
+		case 0:
+			b24 = c<<18;
+			break;
+		case 1:
+			b24 |= c<<12;
+			break;
+		case 2:
+			b24 |= c<<6;
+			break;
+		case 3:
+			if(out + 3 > e)
+				goto exhausted;
+
+			b24 |= c;
+			*out++ = b24>>16;
+			*out++ = b24>>8;
+			*out++ = b24;
+			i = -1;
+			break;
+		}
+		i++;
+	}
+	switch(i){
+	case 2:
+		if(out + 1 > e)
+			goto exhausted;
+		*out++ = b24>>16;
+		break;
+	case 3:
+		if(out + 2 > e)
+			goto exhausted;
+		*out++ = b24>>16;
+		*out++ = b24>>8;
+		break;
+	}
+exhausted:
+	return out - start;
+}
+
+int
+enc64(char *out, int lim, uchar *in, int n)
+{
+	int i;
+	ulong b24;
+	char *start = out;
+	char *e = out + lim;
+
+	for(i = n/3; i > 0; i--){
+		b24 = (*in++)<<16;
+		b24 |= (*in++)<<8;
+		b24 |= *in++;
+		if(out + 4 >= e)
+			goto exhausted;
+		*out++ = t64e[(b24>>18)];
+		*out++ = t64e[(b24>>12)&0x3f];
+		*out++ = t64e[(b24>>6)&0x3f];
+		*out++ = t64e[(b24)&0x3f];
+	}
+
+	switch(n%3){
+	case 2:
+		b24 = (*in++)<<16;
+		b24 |= (*in)<<8;
+		if(out + 4 >= e)
+			goto exhausted;
+		*out++ = t64e[(b24>>18)];
+		*out++ = t64e[(b24>>12)&0x3f];
+		*out++ = t64e[(b24>>6)&0x3f];
+		*out++ = '=';
+		break;
+	case 1:
+		b24 = (*in)<<16;
+		if(out + 4 >= e)
+			goto exhausted;
+		*out++ = t64e[(b24>>18)];
+		*out++ = t64e[(b24>>12)&0x3f];
+		*out++ = '=';
+		*out++ = '=';
+		break;
+	}
+exhausted:
+	*out = 0;
+	return out - start;
+}
diff --git a/libkern/utfecpy.c b/libkern/utfecpy.c
new file mode 100644
index 00000000..8ccb841c
--- /dev/null
+++ b/libkern/utfecpy.c
@@ -0,0 +1,20 @@
+#include "lib9.h"
+
+char*
+utfecpy(char *to, char *e, char *from)
+{
+	char *end;
+
+	if(to >= e)
+		return to;
+	end = memccpy(to, from, '\0', e - to);
+	if(end == nil){
+		end = e-1;
+		while(end>to && (*--end&0xC0)==0x80)
+			;
+		*end = '\0';
+	}else{
+		end--;
+	}
+	return end;
+}
diff --git a/libkern/utflen.c b/libkern/utflen.c
new file mode 100644
index 00000000..a588bbd1
--- /dev/null
+++ b/libkern/utflen.c
@@ -0,0 +1,22 @@
+#include <lib9.h>
+
+int
+utflen(char *s)
+{
+	int c;
+	long n;
+	Rune rune;
+
+	n = 0;
+	for(;;) {
+		c = *(uchar*)s;
+		if(c < Runeself) {
+			if(c == 0)
+				return n;
+			s++;
+		} else
+			s += chartorune(&rune, s);
+		n++;
+	}
+	return 0;
+}
diff --git a/libkern/utfnlen.c b/libkern/utfnlen.c
new file mode 100644
index 00000000..02623f51
--- /dev/null
+++ b/libkern/utfnlen.c
@@ -0,0 +1,25 @@
+#include "lib9.h"
+
+int
+utfnlen(char *s, long m)
+{
+	int c;
+	long n;
+	Rune rune;
+	char *es;
+
+	es = s + m;
+	for(n = 0; s < es; n++) {
+		c = *(uchar*)s;
+		if(c < Runeself){
+			if(c == '\0')
+				break;
+			s++;
+			continue;
+		}
+		if(!fullrune(s, es-s))
+			break;
+		s += chartorune(&rune, s);
+	}
+	return n;
+}
diff --git a/libkern/utfrrune.c b/libkern/utfrrune.c
new file mode 100644
index 00000000..b4594321
--- /dev/null
+++ b/libkern/utfrrune.c
@@ -0,0 +1,30 @@
+#include <lib9.h>
+
+char*
+utfrrune(char *s, long c)
+{
+	long c1;
+	Rune r;
+	char *s1;
+
+	if(c < Runesync)		/* not part of utf sequence */
+		return strrchr(s, c);
+
+	s1 = 0;
+	for(;;) {
+		c1 = *(uchar*)s;
+		if(c1 < Runeself) {	/* one byte rune */
+			if(c1 == 0)
+				return s1;
+			if(c1 == c)
+				s1 = s;
+			s++;
+			continue;
+		}
+		c1 = chartorune(&r, s);
+		if(r == c)
+			s1 = s;
+		s += c1;
+	}
+	return 0;
+}
diff --git a/libkern/utfrune.c b/libkern/utfrune.c
new file mode 100644
index 00000000..9925af81
--- /dev/null
+++ b/libkern/utfrune.c
@@ -0,0 +1,29 @@
+#include <lib9.h>
+
+char*
+utfrune(char *s, long c)
+{
+	long c1;
+	Rune r;
+	int n;
+
+	if(c < Runesync)		/* not part of utf sequence */
+		return strchr(s, c);
+
+	for(;;) {
+		c1 = *(uchar*)s;
+		if(c1 < Runeself) {	/* one byte rune */
+			if(c1 == 0)
+				return 0;
+			if(c1 == c)
+				return s;
+			s++;
+			continue;
+		}
+		n = chartorune(&r, s);
+		if(r == c)
+			return s;
+		s += n;
+	}
+	return 0;
+}
diff --git a/libkern/vlop-386.s b/libkern/vlop-386.s
new file mode 100644
index 00000000..42e8a70d
--- /dev/null
+++ b/libkern/vlop-386.s
@@ -0,0 +1,44 @@
+TEXT	_mulv(SB), $0
+	MOVL	r+0(FP), CX
+	MOVL	a+4(FP), AX
+	MULL	b+12(FP)
+	MOVL	AX, 0(CX)
+	MOVL	DX, BX
+	MOVL	a+4(FP), AX
+	MULL	b+16(FP)
+	ADDL	AX, BX
+	MOVL	a+8(FP), AX
+	MULL	b+12(FP)
+	ADDL	AX, BX
+	MOVL	BX, 4(CX)
+	RET
+
+TEXT	_mul64by32(SB), $0
+	MOVL	r+0(FP), CX
+	MOVL	a+4(FP), AX
+	MULL	b+12(FP)
+	MOVL	AX, 0(CX)
+	MOVL	DX, BX
+	MOVL	a+8(FP), AX
+	MULL	b+12(FP)
+	ADDL	AX, BX
+	MOVL	BX, 4(CX)
+	RET
+
+TEXT	_div64by32(SB), $0
+	MOVL	r+12(FP), CX
+	MOVL	a+0(FP), AX
+	MOVL	a+4(FP), DX
+	DIVL	b+8(FP)
+	MOVL	DX, 0(CX)
+	RET
+
+TEXT	_addv(SB), $0
+	MOVL	r+0(FP), CX
+	MOVL	a+4(FP), AX
+	MOVL	a+8(FP), BX
+	ADDL	b+12(FP), AX
+	ADCL	b+16(FP), BX
+	MOVL	AX, 0(CX)
+	MOVL	BX, 4(CX)
+	RET
diff --git a/libkern/vlop-arm.s b/libkern/vlop-arm.s
new file mode 100644
index 00000000..d972ac1b
--- /dev/null
+++ b/libkern/vlop-arm.s
@@ -0,0 +1,34 @@
+#define	UMULL(Rs,Rm,Rhi,Rlo,S)	WORD	$((14<<28)|(4<<21)|(S<<20)|(Rhi<<16)|(Rlo<<12)|(Rs<<8)|(9<<4)|Rm)
+#define UMLAL(Rs,Rm,Rhi,Rlo,S)	WORD	$((14<<28)|(5<<21)|(S<<20)|(Rhi<<16)|(Rlo<<12)|(Rs<<8)|(9<<4)|Rm)
+#define	MUL(Rs,Rm,Rd,S)	WORD	$((14<<28)|(0<<21)|(S<<20)|(Rd<<16)|(Rs<<8)|(9<<4)|Rm)
+arg=0
+
+/* replaced use of R10 by R11 because the former can be the data segment base register */
+
+TEXT	_mulv(SB), $0
+	MOVW	8(FP), R9	/* l0 */
+	MOVW	4(FP), R11	/* h0 */
+	MOVW	16(FP), R4	/* l1 */
+	MOVW	12(FP), R5	/* h1 */
+	UMULL(4, 9, 7, 6, 0)
+	MUL(11, 4, 8, 0)
+	ADD	R8, R7
+	MUL(9, 5, 8, 0)
+	ADD	R8, R7
+	MOVW	R6, 4(R(arg))
+	MOVW	R7, 0(R(arg))
+	RET
+
+/* multiply, add, and right-shift, yielding a 32-bit result, while
+   using 64-bit accuracy for the multiply -- for fast fixed-point math */
+TEXT	_mularsv(SB), $0
+	MOVW	4(FP), R11	/* m1 */
+	MOVW	8(FP),	R8	/* a */
+	MOVW	12(FP), R4	/* rs */
+	MOVW	$0, R9
+	UMLAL(0, 11, 9, 8, 0)
+	MOVW	R8>>R4, R8
+	RSB	$32, R4, R4
+	ORR	R9<<R4, R8, R0
+	RET
+
diff --git a/libkern/vlop-mips.s b/libkern/vlop-mips.s
new file mode 100644
index 00000000..17f487ad
--- /dev/null
+++ b/libkern/vlop-mips.s
@@ -0,0 +1,17 @@
+TEXT	_mulv(SB), $0
+	MOVW	8(FP), R2
+	MOVW	4(FP), R3
+	MOVW	16(FP), R4
+	MOVW	12(FP), R5
+	MULU	R4, R2
+	MOVW	LO, R6
+	MOVW	HI, R7
+	MULU	R3, R4
+	MOVW	LO, R8
+	ADDU	R8, R7
+	MULU	R2, R5
+	MOVW	LO, R8
+	ADDU	R8, R7
+	MOVW	R6, 4(R1)
+	MOVW	R7, 0(R1)
+	RET
diff --git a/libkern/vlop-power.s b/libkern/vlop-power.s
new file mode 100644
index 00000000..715ef14d
--- /dev/null
+++ b/libkern/vlop-power.s
@@ -0,0 +1,14 @@
+TEXT	_mulv(SB), $0
+	MOVW	8(FP), R9
+	MOVW	4(FP), R10
+	MOVW	16(FP), R4
+	MOVW	12(FP), R5
+	MULLW	R4, R9, R6
+	MULHWU	R4, R9, R7
+	MULLW	R10, R4, R8
+	ADD	R8, R7
+	MULLW	R9, R5, R8
+	ADD	R8, R7
+	MOVW	R6, 4(R3)
+	MOVW	R7, 0(R3)
+	RETURN
diff --git a/libkern/vlop-sparc.s b/libkern/vlop-sparc.s
new file mode 100644
index 00000000..ac36b414
--- /dev/null
+++ b/libkern/vlop-sparc.s
@@ -0,0 +1,112 @@
+TEXT	_mulv(SB), $0
+	MOVW	u1+8(FP), R8
+	MOVW	u2+16(FP), R13
+
+	MOVW	R13, R16		/* save low parts for later */
+	MOVW	R8, R12
+
+	/*
+	 * unsigned 32x32 => 64 multiply
+	 */
+	CMP	R13, R8
+	BLE	mul1
+	MOVW	R12, R13
+	MOVW	R16, R8
+mul1:
+	MOVW	R13, Y
+	ANDNCC	$0xFFF, R13, R0
+	BE	mul_shortway
+	ANDCC	R0, R0, R9		/* zero partial product and clear N and V cond's */
+
+	/* long multiply */
+	MULSCC	R8, R9, R9		/* 0 */
+	MULSCC	R8, R9, R9		/* 1 */
+	MULSCC	R8, R9, R9		/* 2 */
+	MULSCC	R8, R9, R9		/* 3 */
+	MULSCC	R8, R9, R9		/* 4 */
+	MULSCC	R8, R9, R9		/* 5 */
+	MULSCC	R8, R9, R9		/* 6 */
+	MULSCC	R8, R9, R9		/* 7 */
+	MULSCC	R8, R9, R9		/* 8 */
+	MULSCC	R8, R9, R9		/* 9 */
+	MULSCC	R8, R9, R9		/* 10 */
+	MULSCC	R8, R9, R9		/* 11 */
+	MULSCC	R8, R9, R9		/* 12 */
+	MULSCC	R8, R9, R9		/* 13 */
+	MULSCC	R8, R9, R9		/* 14 */
+	MULSCC	R8, R9, R9		/* 15 */
+	MULSCC	R8, R9, R9		/* 16 */
+	MULSCC	R8, R9, R9		/* 17 */
+	MULSCC	R8, R9, R9		/* 18 */
+	MULSCC	R8, R9, R9		/* 19 */
+	MULSCC	R8, R9, R9		/* 20 */
+	MULSCC	R8, R9, R9		/* 21 */
+	MULSCC	R8, R9, R9		/* 22 */
+	MULSCC	R8, R9, R9		/* 23 */
+	MULSCC	R8, R9, R9		/* 24 */
+	MULSCC	R8, R9, R9		/* 25 */
+	MULSCC	R8, R9, R9		/* 26 */
+	MULSCC	R8, R9, R9		/* 27 */
+	MULSCC	R8, R9, R9		/* 28 */
+	MULSCC	R8, R9, R9		/* 29 */
+	MULSCC	R8, R9, R9		/* 30 */
+	MULSCC	R8, R9, R9		/* 31 */
+	MULSCC	R0, R9, R9		/* 32; shift only; r9 is high part */
+
+	/*
+	 * need to correct top word if top bit set
+	 */
+	CMP	R8, R0
+	BGE	mul_tstlow
+	ADD	R13, R9			/* adjust the high parts */
+
+mul_tstlow:
+	MOVW	Y, R13			/* get low part */
+	BA	mul_done
+
+mul_shortway:
+	ANDCC	R0, R0, R9		/* zero partial product and clear N and V cond's */
+	MULSCC	R8, R9, R9		/*  0 */
+	MULSCC	R8, R9, R9		/*  1 */
+	MULSCC	R8, R9, R9		/*  2 */
+	MULSCC	R8, R9, R9		/*  3 */
+	MULSCC	R8, R9, R9		/*  4 */
+	MULSCC	R8, R9, R9		/*  5 */
+	MULSCC	R8, R9, R9		/*  6 */
+	MULSCC	R8, R9, R9		/*  7 */
+	MULSCC	R8, R9, R9		/*  8 */
+	MULSCC	R8, R9, R9		/*  9 */
+	MULSCC	R8, R9, R9		/* 10 */
+	MULSCC	R8, R9, R9		/* 11 */
+	MULSCC	R0, R9, R9		/* 12; shift only; r9 is high part */
+
+	MOVW	Y, R8			/* make low part of partial low part & high part */
+	SLL	$12, R9, R13
+	SRL	$20, R8
+	OR	R8, R13
+
+	SRA	$20, R9			/* high part */
+
+mul_done:
+
+	/*
+	 * mul by high halves if needed
+	 */
+	MOVW	R13, 4(R7)
+	MOVW	u2+12(FP), R11
+	CMP	R11, R0
+	BE	nomul1
+	MUL	R11, R12
+	ADD	R12, R9
+
+nomul1:
+	MOVW	u1+4(FP), R11
+	CMP	R11, R0
+	BE	nomul2
+	MUL	R11, R16
+	ADD	R16, R9
+
+nomul2:
+
+	MOVW	R9, 0(R7)
+	RETURN	
diff --git a/libkern/vlop-thumb.s b/libkern/vlop-thumb.s
new file mode 100644
index 00000000..d972ac1b
--- /dev/null
+++ b/libkern/vlop-thumb.s
@@ -0,0 +1,34 @@
+#define	UMULL(Rs,Rm,Rhi,Rlo,S)	WORD	$((14<<28)|(4<<21)|(S<<20)|(Rhi<<16)|(Rlo<<12)|(Rs<<8)|(9<<4)|Rm)
+#define UMLAL(Rs,Rm,Rhi,Rlo,S)	WORD	$((14<<28)|(5<<21)|(S<<20)|(Rhi<<16)|(Rlo<<12)|(Rs<<8)|(9<<4)|Rm)
+#define	MUL(Rs,Rm,Rd,S)	WORD	$((14<<28)|(0<<21)|(S<<20)|(Rd<<16)|(Rs<<8)|(9<<4)|Rm)
+arg=0
+
+/* replaced use of R10 by R11 because the former can be the data segment base register */
+
+TEXT	_mulv(SB), $0
+	MOVW	8(FP), R9	/* l0 */
+	MOVW	4(FP), R11	/* h0 */
+	MOVW	16(FP), R4	/* l1 */
+	MOVW	12(FP), R5	/* h1 */
+	UMULL(4, 9, 7, 6, 0)
+	MUL(11, 4, 8, 0)
+	ADD	R8, R7
+	MUL(9, 5, 8, 0)
+	ADD	R8, R7
+	MOVW	R6, 4(R(arg))
+	MOVW	R7, 0(R(arg))
+	RET
+
+/* multiply, add, and right-shift, yielding a 32-bit result, while
+   using 64-bit accuracy for the multiply -- for fast fixed-point math */
+TEXT	_mularsv(SB), $0
+	MOVW	4(FP), R11	/* m1 */
+	MOVW	8(FP),	R8	/* a */
+	MOVW	12(FP), R4	/* rs */
+	MOVW	$0, R9
+	UMLAL(0, 11, 9, 8, 0)
+	MOVW	R8>>R4, R8
+	RSB	$32, R4, R4
+	ORR	R9<<R4, R8, R0
+	RET
+
diff --git a/libkern/vlrt-386.c b/libkern/vlrt-386.c
new file mode 100644
index 00000000..f0b100ef
--- /dev/null
+++ b/libkern/vlrt-386.c
@@ -0,0 +1,695 @@
+typedef	unsigned long	ulong;
+typedef	unsigned int	uint;
+typedef	unsigned short	ushort;
+typedef	unsigned char	uchar;
+typedef	signed char	schar;
+
+#define	SIGN(n)	(1UL<<(n-1))
+
+typedef	struct	Vlong	Vlong;
+struct	Vlong
+{
+	union
+	{
+		struct
+		{
+			ulong	lo;
+			ulong	hi;
+		};
+		struct
+		{
+			ushort	lols;
+			ushort	loms;
+			ushort	hils;
+			ushort	hims;
+		};
+	};
+};
+
+void	abort(void);
+
+void
+_subv(Vlong *r, Vlong a, Vlong b)
+{
+	ulong lo, hi;
+
+	lo = a.lo - b.lo;
+	hi = a.hi - b.hi;
+	if(lo > a.lo)
+		hi--;
+	r->lo = lo;
+	r->hi = hi;
+}
+
+void
+_d2v(Vlong *y, double d)
+{
+	union { double d; struct Vlong; } x;
+	ulong xhi, xlo, ylo, yhi;
+	int sh;
+
+	x.d = d;
+
+	xhi = (x.hi & 0xfffff) | 0x100000;
+	xlo = x.lo;
+	sh = 1075 - ((x.hi >> 20) & 0x7ff);
+
+	ylo = 0;
+	yhi = 0;
+	if(sh >= 0) {
+		/* v = (hi||lo) >> sh */
+		if(sh < 32) {
+			if(sh == 0) {
+				ylo = xlo;
+				yhi = xhi;
+			} else {
+				ylo = (xlo >> sh) | (xhi << (32-sh));
+				yhi = xhi >> sh;
+			}
+		} else {
+			if(sh == 32) {
+				ylo = xhi;
+			} else
+			if(sh < 64) {
+				ylo = xhi >> (sh-32);
+			}
+		}
+	} else {
+		/* v = (hi||lo) << -sh */
+		sh = -sh;
+		if(sh <= 10) {
+			ylo = xlo << sh;
+			yhi = (xhi << sh) | (xlo >> (32-sh));
+		} else {
+			/* overflow */
+			yhi = d;	/* causes something awful */
+		}
+	}
+	if(x.hi & SIGN(32)) {
+		if(ylo != 0) {
+			ylo = -ylo;
+			yhi = ~yhi;
+		} else
+			yhi = -yhi;
+	}
+
+	y->hi = yhi;
+	y->lo = ylo;
+}
+
+void
+_f2v(Vlong *y, float f)
+{
+
+	_d2v(y, f);
+}
+
+double
+_v2d(Vlong x)
+{
+	if(x.hi & SIGN(32)) {
+		if(x.lo) {
+			x.lo = -x.lo;
+			x.hi = ~x.hi;
+		} else
+			x.hi = -x.hi;
+		return -((long)x.hi*4294967296. + x.lo);
+	}
+	return (long)x.hi*4294967296. + x.lo;
+}
+
+float
+_v2f(Vlong x)
+{
+	return _v2d(x);
+}
+
+ulong	_div64by32(Vlong, ulong, ulong*);
+void	_mul64by32(Vlong*, Vlong, ulong);
+
+static void
+dodiv(Vlong num, Vlong den, Vlong *qp, Vlong *rp)
+{
+	ulong n;
+	Vlong x, q, r;
+
+	if(den.hi > num.hi || (den.hi == num.hi && den.lo > num.lo)){
+		if(qp) {
+			qp->hi = 0;
+			qp->lo = 0;
+		}
+		if(rp) {
+			rp->hi = num.hi;
+			rp->lo = num.lo;
+		}
+		return;
+	}
+
+	if(den.hi != 0){
+		q.hi = 0;
+		n = num.hi/den.hi;
+		_mul64by32(&x, den, n);
+		if(x.hi > num.hi || (x.hi == num.hi && x.lo > num.lo)){
+			n--;
+			_mul64by32(&x, den, n);
+		}
+		q.lo = n;
+		_subv(&r, num, x);
+	} else {
+		if(num.hi >= den.lo){
+			q.hi = n = num.hi/den.lo;
+			num.hi -= den.lo*n;
+		} else {
+			q.hi = 0;
+		}
+		q.lo = _div64by32(num, den.lo, &r.lo);
+		r.hi = 0;
+	}
+	if(qp) {
+		qp->lo = q.lo;
+		qp->hi = q.hi;
+	}
+	if(rp) {
+		rp->lo = r.lo;
+		rp->hi = r.hi;
+	}
+}
+
+void
+_divvu(Vlong *q, Vlong n, Vlong d)
+{
+
+	if(n.hi == 0 && d.hi == 0) {
+		q->hi = 0;
+		q->lo = n.lo / d.lo;
+		return;
+	}
+	dodiv(n, d, q, 0);
+}
+
+void
+_modvu(Vlong *r, Vlong n, Vlong d)
+{
+
+	if(n.hi == 0 && d.hi == 0) {
+		r->hi = 0;
+		r->lo = n.lo % d.lo;
+		return;
+	}
+	dodiv(n, d, 0, r);
+}
+
+static void
+vneg(Vlong *v)
+{
+
+	if(v->lo == 0) {
+		v->hi = -v->hi;
+		return;
+	}
+	v->lo = -v->lo;
+	v->hi = ~v->hi;
+}
+
+void
+_divv(Vlong *q, Vlong n, Vlong d)
+{
+	long nneg, dneg;
+
+	if(n.hi == (((long)n.lo)>>31) && d.hi == (((long)d.lo)>>31)) {
+		q->lo = (long)n.lo / (long)d.lo;
+		q->hi = ((long)q->lo) >> 31;
+		return;
+	}
+	nneg = n.hi >> 31;
+	if(nneg)
+		vneg(&n);
+	dneg = d.hi >> 31;
+	if(dneg)
+		vneg(&d);
+	dodiv(n, d, q, 0);
+	if(nneg != dneg)
+		vneg(q);
+}
+
+void
+_modv(Vlong *r, Vlong n, Vlong d)
+{
+	long nneg, dneg;
+
+	if(n.hi == (((long)n.lo)>>31) && d.hi == (((long)d.lo)>>31)) {
+		r->lo = (long)n.lo % (long)d.lo;
+		r->hi = ((long)r->lo) >> 31;
+		return;
+	}
+	nneg = n.hi >> 31;
+	if(nneg)
+		vneg(&n);
+	dneg = d.hi >> 31;
+	if(dneg)
+		vneg(&d);
+	dodiv(n, d, 0, r);
+	if(nneg)
+		vneg(r);
+}
+
+void
+_rshav(Vlong *r, Vlong a, int b)
+{
+	long t;
+
+	t = a.hi;
+	if(b >= 32) {
+		r->hi = t>>31;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->lo = t>>31;
+			return;
+		}
+		r->lo = t >> (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->hi = t;
+		r->lo = a.lo;
+		return;
+	}
+	r->hi = t >> b;
+	r->lo = (t << (32-b)) | (a.lo >> b);
+}
+
+void
+_rshlv(Vlong *r, Vlong a, int b)
+{
+	ulong t;
+
+	t = a.hi;
+	if(b >= 32) {
+		r->hi = 0;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->lo = 0;
+			return;
+		}
+		r->lo = t >> (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->hi = t;
+		r->lo = a.lo;
+		return;
+	}
+	r->hi = t >> b;
+	r->lo = (t << (32-b)) | (a.lo >> b);
+}
+
+void
+_lshv(Vlong *r, Vlong a, int b)
+{
+	ulong t;
+
+	t = a.lo;
+	if(b >= 32) {
+		r->lo = 0;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->hi = 0;
+			return;
+		}
+		r->hi = t << (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->lo = t;
+		r->hi = a.hi;
+		return;
+	}
+	r->lo = t << b;
+	r->hi = (t >> (32-b)) | (a.hi << b);
+}
+
+void
+_andv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi & b.hi;
+	r->lo = a.lo & b.lo;
+}
+
+void
+_orv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi | b.hi;
+	r->lo = a.lo | b.lo;
+}
+
+void
+_xorv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi ^ b.hi;
+	r->lo = a.lo ^ b.lo;
+}
+
+void
+_vpp(Vlong *l, Vlong *r)
+{
+
+	l->hi = r->hi;
+	l->lo = r->lo;
+	r->lo++;
+	if(r->lo == 0)
+		r->hi++;
+}
+
+void
+_vmm(Vlong *l, Vlong *r)
+{
+
+	l->hi = r->hi;
+	l->lo = r->lo;
+	if(r->lo == 0)
+		r->hi--;
+	r->lo--;
+}
+
+void
+_ppv(Vlong *l, Vlong *r)
+{
+
+	r->lo++;
+	if(r->lo == 0)
+		r->hi++;
+	l->hi = r->hi;
+	l->lo = r->lo;
+}
+
+void
+_mmv(Vlong *l, Vlong *r)
+{
+
+	if(r->lo == 0)
+		r->hi--;
+	r->lo--;
+	l->hi = r->hi;
+	l->lo = r->lo;
+}
+
+void
+_vasop(Vlong *ret, void *lv, void fn(Vlong*, Vlong, Vlong), int type, Vlong rv)
+{
+	Vlong t, u;
+
+	u.lo = 0;
+	u.hi = 0;
+	switch(type) {
+	default:
+		abort();
+		break;
+
+	case 1:	/* schar */
+		t.lo = *(schar*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(schar*)lv = u.lo;
+		break;
+
+	case 2:	/* uchar */
+		t.lo = *(uchar*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(uchar*)lv = u.lo;
+		break;
+
+	case 3:	/* short */
+		t.lo = *(short*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(short*)lv = u.lo;
+		break;
+
+	case 4:	/* ushort */
+		t.lo = *(ushort*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(ushort*)lv = u.lo;
+		break;
+
+	case 9:	/* int */
+		t.lo = *(int*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(int*)lv = u.lo;
+		break;
+
+	case 10:	/* uint */
+		t.lo = *(uint*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(uint*)lv = u.lo;
+		break;
+
+	case 5:	/* long */
+		t.lo = *(long*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(long*)lv = u.lo;
+		break;
+
+	case 6:	/* ulong */
+		t.lo = *(ulong*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(ulong*)lv = u.lo;
+		break;
+
+	case 7:	/* vlong */
+	case 8:	/* uvlong */
+		fn(&u, *(Vlong*)lv, rv);
+		*(Vlong*)lv = u;
+		break;
+	}
+	*ret = u;
+}
+
+void
+_p2v(Vlong *ret, void *p)
+{
+	long t;
+
+	t = (ulong)p;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sl2v(Vlong *ret, long sl)
+{
+	long t;
+
+	t = sl;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_ul2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_si2v(Vlong *ret, int si)
+{
+	long t;
+
+	t = si;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_ui2v(Vlong *ret, uint ui)
+{
+	long t;
+
+	t = ui;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sh2v(Vlong *ret, long sh)
+{
+	long t;
+
+	t = (sh << 16) >> 16;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_uh2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul & 0xffff;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sc2v(Vlong *ret, long uc)
+{
+	long t;
+
+	t = (uc << 24) >> 24;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_uc2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul & 0xff;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+long
+_v2sc(Vlong rv)
+{
+	long t;
+
+	t = rv.lo & 0xff;
+	return (t << 24) >> 24;
+}
+
+long
+_v2uc(Vlong rv)
+{
+
+	return rv.lo & 0xff;
+}
+
+long
+_v2sh(Vlong rv)
+{
+	long t;
+
+	t = rv.lo & 0xffff;
+	return (t << 16) >> 16;
+}
+
+long
+_v2uh(Vlong rv)
+{
+
+	return rv.lo & 0xffff;
+}
+
+long
+_v2sl(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2ul(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2si(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2ui(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+int
+_testv(Vlong rv)
+{
+	return rv.lo || rv.hi;
+}
+
+int
+_eqv(Vlong lv, Vlong rv)
+{
+	return lv.lo == rv.lo && lv.hi == rv.hi;
+}
+
+int
+_nev(Vlong lv, Vlong rv)
+{
+	return lv.lo != rv.lo || lv.hi != rv.hi;
+}
+
+int
+_ltv(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi < (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo < rv.lo);
+}
+
+int
+_lev(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi < (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo <= rv.lo);
+}
+
+int
+_gtv(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi > (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo > rv.lo);
+}
+
+int
+_gev(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi > (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo >= rv.lo);
+}
+
+int
+_lov(Vlong lv, Vlong rv)
+{
+	return lv.hi < rv.hi || 
+		(lv.hi == rv.hi && lv.lo < rv.lo);
+}
+
+int
+_lsv(Vlong lv, Vlong rv)
+{
+	return lv.hi < rv.hi || 
+		(lv.hi == rv.hi && lv.lo <= rv.lo);
+}
+
+int
+_hiv(Vlong lv, Vlong rv)
+{
+	return lv.hi > rv.hi || 
+		(lv.hi == rv.hi && lv.lo > rv.lo);
+}
+
+int
+_hsv(Vlong lv, Vlong rv)
+{
+	return lv.hi > rv.hi || 
+		(lv.hi == rv.hi && lv.lo >= rv.lo);
+}
diff --git a/libkern/vlrt-68000.c b/libkern/vlrt-68000.c
new file mode 100644
index 00000000..071aa180
--- /dev/null
+++ b/libkern/vlrt-68000.c
@@ -0,0 +1,771 @@
+typedef	unsigned long	ulong;
+typedef	unsigned int	uint;
+typedef	unsigned short	ushort;
+typedef	unsigned char	uchar;
+typedef	signed char	schar;
+
+#define	SIGN(n)	(1UL<<(n-1))
+
+typedef	struct	Vlong	Vlong;
+struct	Vlong
+{
+	union
+	{
+		struct
+		{
+			ulong	hi;
+			ulong	lo;
+		};
+		struct
+		{
+			ushort	hims;
+			ushort	hils;
+			ushort	loms;
+			ushort	lols;
+		};
+	};
+};
+
+void	abort(void);
+
+void
+_addv(Vlong *r, Vlong a, Vlong b)
+{
+	ulong lo, hi;
+
+	lo = a.lo + b.lo;
+	hi = a.hi + b.hi;
+	if(lo < a.lo)
+		hi++;
+	r->lo = lo;
+	r->hi = hi;
+}
+
+void
+_subv(Vlong *r, Vlong a, Vlong b)
+{
+	ulong lo, hi;
+
+	lo = a.lo - b.lo;
+	hi = a.hi - b.hi;
+	if(lo > a.lo)
+		hi--;
+	r->lo = lo;
+	r->hi = hi;
+}
+
+void
+_d2v(Vlong *y, double d)
+{
+	union { double d; struct Vlong; } x;
+	ulong xhi, xlo, ylo, yhi;
+	int sh;
+
+	x.d = d;
+
+	xhi = (x.hi & 0xfffff) | 0x100000;
+	xlo = x.lo;
+	sh = 1075 - ((x.hi >> 20) & 0x7ff);
+
+	ylo = 0;
+	yhi = 0;
+	if(sh >= 0) {
+		/* v = (hi||lo) >> sh */
+		if(sh < 32) {
+			if(sh == 0) {
+				ylo = xlo;
+				yhi = xhi;
+			} else {
+				ylo = (xlo >> sh) | (xhi << (32-sh));
+				yhi = xhi >> sh;
+			}
+		} else {
+			if(sh == 32) {
+				ylo = xhi;
+			} else
+			if(sh < 64) {
+				ylo = xhi >> (sh-32);
+			}
+		}
+	} else {
+		/* v = (hi||lo) << -sh */
+		sh = -sh;
+		if(sh <= 10) {
+			ylo = xlo << sh;
+			yhi = (xhi << sh) | (xlo >> (32-sh));
+		} else {
+			/* overflow */
+			yhi = d;	/* causes something awful */
+		}
+	}
+	if(x.hi & SIGN(32)) {
+		if(ylo != 0) {
+			ylo = -ylo;
+			yhi = ~yhi;
+		} else
+			yhi = -yhi;
+	}
+
+	y->hi = yhi;
+	y->lo = ylo;
+}
+
+void
+_f2v(Vlong *y, float f)
+{
+
+	_d2v(y, f);
+}
+
+double
+_v2d(Vlong x)
+{
+	if(x.hi & SIGN(32)) {
+		if(x.lo) {
+			x.lo = -x.lo;
+			x.hi = ~x.hi;
+		} else
+			x.hi = -x.hi;
+		return -((long)x.hi*4294967296. + x.lo);
+	}
+	return (long)x.hi*4294967296. + x.lo;
+}
+
+float
+_v2f(Vlong x)
+{
+	return _v2d(x);
+}
+
+static void
+dodiv(Vlong num, Vlong den, Vlong *q, Vlong *r)
+{
+	ulong numlo, numhi, denhi, denlo, quohi, quolo, t;
+	int i;
+
+	numhi = num.hi;
+	numlo = num.lo;
+	denhi = den.hi;
+	denlo = den.lo;
+
+	/*
+	 * get a divide by zero
+	 */
+	if(denlo==0 && denhi==0) {
+		numlo = numlo / denlo;
+	}
+
+	/*
+	 * set up the divisor and find the number of iterations needed
+	 */
+	if(numhi >= SIGN(32)) {
+		quohi = SIGN(32);
+		quolo = 0;
+	} else {
+		quohi = numhi;
+		quolo = numlo;
+	}
+	i = 0;
+	while(denhi < quohi || (denhi == quohi && denlo < quolo)) {
+		denhi = (denhi<<1) | (denlo>>31);
+		denlo <<= 1;
+		i++;
+	}
+
+	quohi = 0;
+	quolo = 0;
+	for(; i >= 0; i--) {
+		quohi = (quohi<<1) | (quolo>>31);
+		quolo <<= 1;
+		if(numhi > denhi || (numhi == denhi && numlo >= denlo)) {
+			t = numlo;
+			numlo -= denlo;
+			if(numlo > t)
+				numhi--;
+			numhi -= denhi;
+			quolo |= 1;
+		}
+		denlo = (denlo>>1) | (denhi<<31);
+		denhi >>= 1;
+	}
+
+	if(q) {
+		q->lo = quolo;
+		q->hi = quohi;
+	}
+	if(r) {
+		r->lo = numlo;
+		r->hi = numhi;
+	}
+}
+
+void
+_divvu(Vlong *q, Vlong n, Vlong d)
+{
+
+	if(n.hi == 0 && d.hi == 0) {
+		q->hi = 0;
+		q->lo = n.lo / d.lo;
+		return;
+	}
+	dodiv(n, d, q, 0);
+}
+
+void
+_modvu(Vlong *r, Vlong n, Vlong d)
+{
+
+	if(n.hi == 0 && d.hi == 0) {
+		r->hi = 0;
+		r->lo = n.lo % d.lo;
+		return;
+	}
+	dodiv(n, d, 0, r);
+}
+
+static void
+vneg(Vlong *v)
+{
+
+	if(v->lo == 0) {
+		v->hi = -v->hi;
+		return;
+	}
+	v->lo = -v->lo;
+	v->hi = ~v->hi;
+}
+
+void
+_divv(Vlong *q, Vlong n, Vlong d)
+{
+	long nneg, dneg;
+
+	if(n.hi == (((long)n.lo)>>31) && d.hi == (((long)d.lo)>>31)) {
+		q->lo = (long)n.lo / (long)d.lo;
+		q->hi = ((long)q->lo) >> 31;
+		return;
+	}
+	nneg = n.hi >> 31;
+	if(nneg)
+		vneg(&n);
+	dneg = d.hi >> 31;
+	if(dneg)
+		vneg(&d);
+	dodiv(n, d, q, 0);
+	if(nneg != dneg)
+		vneg(q);
+}
+
+void
+_modv(Vlong *r, Vlong n, Vlong d)
+{
+	long nneg, dneg;
+
+	if(n.hi == (((long)n.lo)>>31) && d.hi == (((long)d.lo)>>31)) {
+		r->lo = (long)n.lo % (long)d.lo;
+		r->hi = ((long)r->lo) >> 31;
+		return;
+	}
+	nneg = n.hi >> 31;
+	if(nneg)
+		vneg(&n);
+	dneg = d.hi >> 31;
+	if(dneg)
+		vneg(&d);
+	dodiv(n, d, 0, r);
+	if(nneg)
+		vneg(r);
+}
+
+void
+_rshav(Vlong *r, Vlong a, int b)
+{
+	long t;
+
+	t = a.hi;
+	if(b >= 32) {
+		r->hi = t>>31;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->lo = t>>31;
+			return;
+		}
+		r->lo = t >> (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->hi = t;
+		r->lo = a.lo;
+		return;
+	}
+	r->hi = t >> b;
+	r->lo = (t << (32-b)) | (a.lo >> b);
+}
+
+void
+_rshlv(Vlong *r, Vlong a, int b)
+{
+	ulong t;
+
+	t = a.hi;
+	if(b >= 32) {
+		r->hi = 0;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->lo = 0;
+			return;
+		}
+		r->lo = t >> (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->hi = t;
+		r->lo = a.lo;
+		return;
+	}
+	r->hi = t >> b;
+	r->lo = (t << (32-b)) | (a.lo >> b);
+}
+
+void
+_lshv(Vlong *r, Vlong a, int b)
+{
+	ulong t;
+
+	t = a.lo;
+	if(b >= 32) {
+		r->lo = 0;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->hi = 0;
+			return;
+		}
+		r->hi = t << (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->lo = t;
+		r->hi = a.hi;
+		return;
+	}
+	r->lo = t << b;
+	r->hi = (t >> (32-b)) | (a.hi << b);
+}
+
+void
+_andv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi & b.hi;
+	r->lo = a.lo & b.lo;
+}
+
+void
+_orv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi | b.hi;
+	r->lo = a.lo | b.lo;
+}
+
+void
+_xorv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi ^ b.hi;
+	r->lo = a.lo ^ b.lo;
+}
+
+void
+_negv(Vlong *r, Vlong a)
+{
+	if(a.lo == 0) {
+		r->hi = -a.hi;
+		r->lo = 0;
+		return;
+	}
+	r->hi = ~a.hi;
+	r->lo = -a.lo;
+}
+
+void
+_vpp(Vlong *l, Vlong *r)
+{
+
+	l->hi = r->hi;
+	l->lo = r->lo;
+	r->lo++;
+	if(r->lo == 0)
+		r->hi++;
+}
+
+void
+_vmm(Vlong *l, Vlong *r)
+{
+
+	l->hi = r->hi;
+	l->lo = r->lo;
+	if(r->lo == 0)
+		r->hi--;
+	r->lo--;
+}
+
+void
+_ppv(Vlong *l, Vlong *r)
+{
+
+	r->lo++;
+	if(r->lo == 0)
+		r->hi++;
+	l->hi = r->hi;
+	l->lo = r->lo;
+}
+
+void
+_mmv(Vlong *l, Vlong *r)
+{
+
+	if(r->lo == 0)
+		r->hi--;
+	r->lo--;
+	l->hi = r->hi;
+	l->lo = r->lo;
+}
+
+void
+_vasop(Vlong *ret, void *lv, void fn(Vlong*, Vlong, Vlong), int type, Vlong rv)
+{
+	Vlong t, u;
+
+	u = *ret;
+	switch(type) {
+	default:
+		abort();
+		break;
+
+	case 1:	/* schar */
+		t.lo = *(schar*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(schar*)lv = u.lo;
+		break;
+
+	case 2:	/* uchar */
+		t.lo = *(uchar*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(uchar*)lv = u.lo;
+		break;
+
+	case 3:	/* short */
+		t.lo = *(short*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(short*)lv = u.lo;
+		break;
+
+	case 4:	/* ushort */
+		t.lo = *(ushort*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(ushort*)lv = u.lo;
+		break;
+
+	case 9:	/* int */
+		t.lo = *(int*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(int*)lv = u.lo;
+		break;
+
+	case 10:	/* uint */
+		t.lo = *(uint*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(uint*)lv = u.lo;
+		break;
+
+	case 5:	/* long */
+		t.lo = *(long*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(long*)lv = u.lo;
+		break;
+
+	case 6:	/* ulong */
+		t.lo = *(ulong*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(ulong*)lv = u.lo;
+		break;
+
+	case 7:	/* vlong */
+	case 8:	/* uvlong */
+		fn(&u, *(Vlong*)lv, rv);
+		*(Vlong*)lv = u;
+		break;
+	}
+	*ret = u;
+}
+
+void
+_p2v(Vlong *ret, void *p)
+{
+	long t;
+
+	t = (ulong)p;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sl2v(Vlong *ret, long sl)
+{
+	long t;
+
+	t = sl;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_ul2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_si2v(Vlong *ret, int si)
+{
+	long t;
+
+	t = si;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_ui2v(Vlong *ret, uint ui)
+{
+	long t;
+
+	t = ui;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sh2v(Vlong *ret, long sh)
+{
+	long t;
+
+	t = (sh << 16) >> 16;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_uh2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul & 0xffff;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sc2v(Vlong *ret, long uc)
+{
+	long t;
+
+	t = (uc << 24) >> 24;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_uc2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul & 0xff;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+long
+_v2sc(Vlong rv)
+{
+	long t;
+
+	t = rv.lo & 0xff;
+	return (t << 24) >> 24;
+}
+
+long
+_v2uc(Vlong rv)
+{
+
+	return rv.lo & 0xff;
+}
+
+long
+_v2sh(Vlong rv)
+{
+	long t;
+
+	t = rv.lo & 0xffff;
+	return (t << 16) >> 16;
+}
+
+long
+_v2uh(Vlong rv)
+{
+
+	return rv.lo & 0xffff;
+}
+
+long
+_v2sl(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2ul(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2si(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2ui(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+int
+_testv(Vlong rv)
+{
+	return rv.lo || rv.hi;
+}
+
+int
+_eqv(Vlong lv, Vlong rv)
+{
+	return lv.lo == rv.lo && lv.hi == rv.hi;
+}
+
+int
+_nev(Vlong lv, Vlong rv)
+{
+	return lv.lo != rv.lo || lv.hi != rv.hi;
+}
+
+int
+_ltv(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi < (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo < rv.lo);
+}
+
+int
+_lev(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi < (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo <= rv.lo);
+}
+
+int
+_gtv(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi > (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo > rv.lo);
+}
+
+int
+_gev(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi > (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo >= rv.lo);
+}
+
+int
+_lov(Vlong lv, Vlong rv)
+{
+	return lv.hi < rv.hi || 
+		(lv.hi == rv.hi && lv.lo < rv.lo);
+}
+
+int
+_lsv(Vlong lv, Vlong rv)
+{
+	return lv.hi < rv.hi || 
+		(lv.hi == rv.hi && lv.lo <= rv.lo);
+}
+
+int
+_hiv(Vlong lv, Vlong rv)
+{
+	return lv.hi > rv.hi || 
+		(lv.hi == rv.hi && lv.lo > rv.lo);
+}
+
+int
+_hsv(Vlong lv, Vlong rv)
+{
+	return lv.hi > rv.hi || 
+		(lv.hi == rv.hi && lv.lo >= rv.lo);
+}
+
+void
+_mulv(Vlong *r, Vlong a, Vlong b)
+{
+	ulong ahi, alo, chi, clo, x;
+	int i;
+
+	ahi = a.hi;
+	alo = a.lo;
+	chi = 0;
+	clo = 0;
+
+	x = b.lo;
+	for(i=0; i<32; i++) {
+		if(x & 1) {
+			chi += ahi;
+			clo += alo;
+			if(clo < alo)
+				chi++;
+		}
+		ahi <<= 1;
+		if(alo & SIGN(32))
+			ahi += 1;
+		alo <<= 1;
+		x >>= 1;
+	}
+
+	/*
+	 * same, but
+	 *	alo is known to be 0
+	 *	can stop when x == 0
+	 */
+	for(x=b.hi; x; x>>=1) {
+		if(x & 1)
+			chi += ahi;
+		ahi <<= 1;
+	}
+
+	r->hi = chi;
+	r->lo = clo;
+}
diff --git a/libkern/vlrt-arm.c b/libkern/vlrt-arm.c
new file mode 100644
index 00000000..58cf5591
--- /dev/null
+++ b/libkern/vlrt-arm.c
@@ -0,0 +1,718 @@
+typedef	unsigned long	ulong;
+typedef	unsigned int	uint;
+typedef	unsigned short	ushort;
+typedef	unsigned char	uchar;
+typedef	signed char	schar;
+
+#define	SIGN(n)	(1UL<<(n-1))
+
+typedef	struct	Vlong	Vlong;
+struct	Vlong
+{
+	union
+	{
+		struct
+		{
+			ulong	hi;
+			ulong	lo;
+		};
+		struct
+		{
+			ushort	hims;
+			ushort	hils;
+			ushort	loms;
+			ushort	lols;
+		};
+	};
+};
+
+void	abort(void);
+
+void
+_addv(Vlong *r, Vlong a, Vlong b)
+{
+	ulong lo, hi;
+
+	lo = a.lo + b.lo;
+	hi = a.hi + b.hi;
+	if(lo < a.lo)
+		hi++;
+	r->lo = lo;
+	r->hi = hi;
+}
+
+void
+_subv(Vlong *r, Vlong a, Vlong b)
+{
+	ulong lo, hi;
+
+	lo = a.lo - b.lo;
+	hi = a.hi - b.hi;
+	if(lo > a.lo)
+		hi--;
+	r->lo = lo;
+	r->hi = hi;
+}
+
+
+void
+_d2v(Vlong *y, double d)
+{
+	union { double d; struct Vlong; } x;
+	ulong xhi, xlo, ylo, yhi;
+	int sh;
+
+	x.d = d;
+
+	xhi = (x.hi & 0xfffff) | 0x100000;
+	xlo = x.lo;
+	sh = 1075 - ((x.hi >> 20) & 0x7ff);
+
+	ylo = 0;
+	yhi = 0;
+	if(sh >= 0) {
+		/* v = (hi||lo) >> sh */
+		if(sh < 32) {
+			if(sh == 0) {
+				ylo = xlo;
+				yhi = xhi;
+			} else {
+				ylo = (xlo >> sh) | (xhi << (32-sh));
+				yhi = xhi >> sh;
+			}
+		} else {
+			if(sh == 32) {
+				ylo = xhi;
+			} else
+			if(sh < 64) {
+				ylo = xhi >> (sh-32);
+			}
+		}
+	} else {
+		/* v = (hi||lo) << -sh */
+		sh = -sh;
+		if(sh <= 10) {
+			ylo = xlo << sh;
+			yhi = (xhi << sh) | (xlo >> (32-sh));
+		} else {
+			/* overflow */
+			yhi = d;	/* causes something awful */
+		}
+	}
+	if(x.hi & SIGN(32)) {
+		if(ylo != 0) {
+			ylo = -ylo;
+			yhi = ~yhi;
+		} else
+			yhi = -yhi;
+	}
+
+	y->hi = yhi;
+	y->lo = ylo;
+}
+
+void
+_f2v(Vlong *y, float f)
+{
+	_d2v(y, f);
+}
+
+double
+_v2d(Vlong x)
+{
+	if(x.hi & SIGN(32)) {
+		if(x.lo) {
+			x.lo = -x.lo;
+			x.hi = ~x.hi;
+		} else
+			x.hi = -x.hi;
+		return -((long)x.hi*4294967296. + x.lo);
+	}
+	return (long)x.hi*4294967296. + x.lo;
+}
+
+float
+_v2f(Vlong x)
+{
+	return _v2d(x);
+}
+
+static void
+dodiv(Vlong num, Vlong den, Vlong *q, Vlong *r)
+{
+	ulong numlo, numhi, denhi, denlo, quohi, quolo, t;
+	int i;
+
+	numhi = num.hi;
+	numlo = num.lo;
+	denhi = den.hi;
+	denlo = den.lo;
+
+	/*
+	 * get a divide by zero
+	 */
+	if(denlo==0 && denhi==0) {
+		numlo = numlo / denlo;
+	}
+
+	/*
+	 * set up the divisor and find the number of iterations needed
+	 */
+	if(numhi >= SIGN(32)) {
+		quohi = SIGN(32);
+		quolo = 0;
+	} else {
+		quohi = numhi;
+		quolo = numlo;
+	}
+	i = 0;
+	while(denhi < quohi || (denhi == quohi && denlo < quolo)) {
+		denhi = (denhi<<1) | (denlo>>31);
+		denlo <<= 1;
+		i++;
+	}
+
+	quohi = 0;
+	quolo = 0;
+	for(; i >= 0; i--) {
+		quohi = (quohi<<1) | (quolo>>31);
+		quolo <<= 1;
+		if(numhi > denhi || (numhi == denhi && numlo >= denlo)) {
+			t = numlo;
+			numlo -= denlo;
+			if(numlo > t)
+				numhi--;
+			numhi -= denhi;
+			quolo |= 1;
+		}
+		denlo = (denlo>>1) | (denhi<<31);
+		denhi >>= 1;
+	}
+
+	if(q) {
+		q->lo = quolo;
+		q->hi = quohi;
+	}
+	if(r) {
+		r->lo = numlo;
+		r->hi = numhi;
+	}
+}
+
+void
+_divvu(Vlong *q, Vlong n, Vlong d)
+{
+
+	if(n.hi == 0 && d.hi == 0) {
+		q->hi = 0;
+		q->lo = n.lo / d.lo;
+		return;
+	}
+	dodiv(n, d, q, 0);
+}
+
+void
+_modvu(Vlong *r, Vlong n, Vlong d)
+{
+
+	if(n.hi == 0 && d.hi == 0) {
+		r->hi = 0;
+		r->lo = n.lo % d.lo;
+		return;
+	}
+	dodiv(n, d, 0, r);
+}
+
+static void
+vneg(Vlong *v)
+{
+
+	if(v->lo == 0) {
+		v->hi = -v->hi;
+		return;
+	}
+	v->lo = -v->lo;
+	v->hi = ~v->hi;
+}
+
+void
+_divv(Vlong *q, Vlong n, Vlong d)
+{
+	long nneg, dneg;
+
+	if(n.hi == (((long)n.lo)>>31) && d.hi == (((long)d.lo)>>31)) {
+		q->lo = (long)n.lo / (long)d.lo;
+		q->hi = ((long)q->lo) >> 31;
+		return;
+	}
+	nneg = n.hi >> 31;
+	if(nneg)
+		vneg(&n);
+	dneg = d.hi >> 31;
+	if(dneg)
+		vneg(&d);
+	dodiv(n, d, q, 0);
+	if(nneg != dneg)
+		vneg(q);
+}
+
+void
+_modv(Vlong *r, Vlong n, Vlong d)
+{
+	long nneg, dneg;
+
+	if(n.hi == (((long)n.lo)>>31) && d.hi == (((long)d.lo)>>31)) {
+		r->lo = (long)n.lo % (long)d.lo;
+		r->hi = ((long)r->lo) >> 31;
+		return;
+	}
+	nneg = n.hi >> 31;
+	if(nneg)
+		vneg(&n);
+	dneg = d.hi >> 31;
+	if(dneg)
+		vneg(&d);
+	dodiv(n, d, 0, r);
+	if(nneg)
+		vneg(r);
+}
+
+void
+_rshav(Vlong *r, Vlong a, int b)
+{
+	long t;
+
+	t = a.hi;
+	if(b >= 32) {
+		r->hi = t>>31;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->lo = t>>31;
+			return;
+		}
+		r->lo = t >> (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->hi = t;
+		r->lo = a.lo;
+		return;
+	}
+	r->hi = t >> b;
+	r->lo = (t << (32-b)) | (a.lo >> b);
+}
+
+void
+_rshlv(Vlong *r, Vlong a, int b)
+{
+	ulong t;
+
+	t = a.hi;
+	if(b >= 32) {
+		r->hi = 0;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->lo = 0;
+			return;
+		}
+		r->lo = t >> (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->hi = t;
+		r->lo = a.lo;
+		return;
+	}
+	r->hi = t >> b;
+	r->lo = (t << (32-b)) | (a.lo >> b);
+}
+
+void
+_lshv(Vlong *r, Vlong a, int b)
+{
+	ulong t;
+
+	t = a.lo;
+	if(b >= 32) {
+		r->lo = 0;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->hi = 0;
+			return;
+		}
+		r->hi = t << (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->lo = t;
+		r->hi = a.hi;
+		return;
+	}
+	r->lo = t << b;
+	r->hi = (t >> (32-b)) | (a.hi << b);
+}
+
+void
+_andv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi & b.hi;
+	r->lo = a.lo & b.lo;
+}
+
+void
+_orv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi | b.hi;
+	r->lo = a.lo | b.lo;
+}
+
+void
+_xorv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi ^ b.hi;
+	r->lo = a.lo ^ b.lo;
+}
+
+void
+_vpp(Vlong *l, Vlong *r)
+{
+
+	l->hi = r->hi;
+	l->lo = r->lo;
+	r->lo++;
+	if(r->lo == 0)
+		r->hi++;
+}
+
+void
+_vmm(Vlong *l, Vlong *r)
+{
+
+	l->hi = r->hi;
+	l->lo = r->lo;
+	if(r->lo == 0)
+		r->hi--;
+	r->lo--;
+}
+
+void
+_ppv(Vlong *l, Vlong *r)
+{
+
+	r->lo++;
+	if(r->lo == 0)
+		r->hi++;
+	l->hi = r->hi;
+	l->lo = r->lo;
+}
+
+void
+_mmv(Vlong *l, Vlong *r)
+{
+
+	if(r->lo == 0)
+		r->hi--;
+	r->lo--;
+	l->hi = r->hi;
+	l->lo = r->lo;
+}
+
+void
+_vasop(Vlong *ret, void *lv, void fn(Vlong*, Vlong, Vlong), int type, Vlong rv)
+{
+	Vlong t, u;
+
+	u = *ret;
+	switch(type) {
+	default:
+		abort();
+		break;
+
+	case 1:	/* schar */
+		t.lo = *(schar*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(schar*)lv = u.lo;
+		break;
+
+	case 2:	/* uchar */
+		t.lo = *(uchar*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(uchar*)lv = u.lo;
+		break;
+
+	case 3:	/* short */
+		t.lo = *(short*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(short*)lv = u.lo;
+		break;
+
+	case 4:	/* ushort */
+		t.lo = *(ushort*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(ushort*)lv = u.lo;
+		break;
+
+	case 9:	/* int */
+		t.lo = *(int*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(int*)lv = u.lo;
+		break;
+
+	case 10:	/* uint */
+		t.lo = *(uint*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(uint*)lv = u.lo;
+		break;
+
+	case 5:	/* long */
+		t.lo = *(long*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(long*)lv = u.lo;
+		break;
+
+	case 6:	/* ulong */
+		t.lo = *(ulong*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(ulong*)lv = u.lo;
+		break;
+
+	case 7:	/* vlong */
+	case 8:	/* uvlong */
+		fn(&u, *(Vlong*)lv, rv);
+		*(Vlong*)lv = u;
+		break;
+	}
+	*ret = u;
+}
+
+void
+_p2v(Vlong *ret, void *p)
+{
+	long t;
+
+	t = (ulong)p;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sl2v(Vlong *ret, long sl)
+{
+	long t;
+
+	t = sl;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_ul2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_si2v(Vlong *ret, int si)
+{
+	long t;
+
+	t = si;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_ui2v(Vlong *ret, uint ui)
+{
+	long t;
+
+	t = ui;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sh2v(Vlong *ret, long sh)
+{
+	long t;
+
+	t = (sh << 16) >> 16;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_uh2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul & 0xffff;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sc2v(Vlong *ret, long uc)
+{
+	long t;
+
+	t = (uc << 24) >> 24;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_uc2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul & 0xff;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+long
+_v2sc(Vlong rv)
+{
+	long t;
+
+	t = rv.lo & 0xff;
+	return (t << 24) >> 24;
+}
+
+long
+_v2uc(Vlong rv)
+{
+
+	return rv.lo & 0xff;
+}
+
+long
+_v2sh(Vlong rv)
+{
+	long t;
+
+	t = rv.lo & 0xffff;
+	return (t << 16) >> 16;
+}
+
+long
+_v2uh(Vlong rv)
+{
+
+	return rv.lo & 0xffff;
+}
+
+long
+_v2sl(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2ul(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2si(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2ui(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+int
+_testv(Vlong rv)
+{
+	return rv.lo || rv.hi;
+}
+
+int
+_eqv(Vlong lv, Vlong rv)
+{
+	return lv.lo == rv.lo && lv.hi == rv.hi;
+}
+
+int
+_nev(Vlong lv, Vlong rv)
+{
+	return lv.lo != rv.lo || lv.hi != rv.hi;
+}
+
+int
+_ltv(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi < (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo < rv.lo);
+}
+
+int
+_lev(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi < (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo <= rv.lo);
+}
+
+int
+_gtv(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi > (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo > rv.lo);
+}
+
+int
+_gev(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi > (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo >= rv.lo);
+}
+
+int
+_lov(Vlong lv, Vlong rv)
+{
+	return lv.hi < rv.hi || 
+		(lv.hi == rv.hi && lv.lo < rv.lo);
+}
+
+int
+_lsv(Vlong lv, Vlong rv)
+{
+	return lv.hi < rv.hi || 
+		(lv.hi == rv.hi && lv.lo <= rv.lo);
+}
+
+int
+_hiv(Vlong lv, Vlong rv)
+{
+	return lv.hi > rv.hi || 
+		(lv.hi == rv.hi && lv.lo > rv.lo);
+}
+
+int
+_hsv(Vlong lv, Vlong rv)
+{
+	return lv.hi > rv.hi || 
+		(lv.hi == rv.hi && lv.lo >= rv.lo);
+}
diff --git a/libkern/vlrt-mips.c b/libkern/vlrt-mips.c
new file mode 100644
index 00000000..96456277
--- /dev/null
+++ b/libkern/vlrt-mips.c
@@ -0,0 +1,719 @@
+typedef	unsigned long	ulong;
+typedef	unsigned int	uint;
+typedef	unsigned short	ushort;
+typedef	unsigned char	uchar;
+typedef	signed char	schar;
+
+#define	SIGN(n)	(1UL<<(n-1))
+
+typedef	struct	Vlong	Vlong;
+struct	Vlong
+{
+	union
+	{
+		struct
+		{
+			ulong	hi;
+			ulong	lo;
+		};
+		struct
+		{
+			ushort	hims;
+			ushort	hils;
+			ushort	loms;
+			ushort	lols;
+		};
+	};
+};
+
+void	abort(void);
+
+void
+_addv(Vlong *r, Vlong a, Vlong b)
+{
+	ulong lo, hi;
+
+	lo = a.lo + b.lo;
+	hi = a.hi + b.hi;
+	if(lo < a.lo)
+		hi++;
+	r->lo = lo;
+	r->hi = hi;
+}
+
+void
+_subv(Vlong *r, Vlong a, Vlong b)
+{
+	ulong lo, hi;
+
+	lo = a.lo - b.lo;
+	hi = a.hi - b.hi;
+	if(lo > a.lo)
+		hi--;
+	r->lo = lo;
+	r->hi = hi;
+}
+
+void
+_d2v(Vlong *y, double d)
+{
+	union { double d; struct Vlong; } x;
+	ulong xhi, xlo, ylo, yhi;
+	int sh;
+
+	x.d = d;
+
+	xhi = (x.hi & 0xfffff) | 0x100000;
+	xlo = x.lo;
+	sh = 1075 - ((x.hi >> 20) & 0x7ff);
+
+	ylo = 0;
+	yhi = 0;
+	if(sh >= 0) {
+		/* v = (hi||lo) >> sh */
+		if(sh < 32) {
+			if(sh == 0) {
+				ylo = xlo;
+				yhi = xhi;
+			} else {
+				ylo = (xlo >> sh) | (xhi << (32-sh));
+				yhi = xhi >> sh;
+			}
+		} else {
+			if(sh == 32) {
+				ylo = xhi;
+			} else
+			if(sh < 64) {
+				ylo = xhi >> (sh-32);
+			}
+		}
+	} else {
+		/* v = (hi||lo) << -sh */
+		sh = -sh;
+		if(sh <= 10) {
+			ylo = xlo << sh;
+			yhi = (xhi << sh) | (xlo >> (32-sh));
+		} else {
+			/* overflow */
+			yhi = d;	/* causes something awful */
+		}
+	}
+	if(x.hi & SIGN(32)) {
+		if(ylo != 0) {
+			ylo = -ylo;
+			yhi = ~yhi;
+		} else
+			yhi = -yhi;
+	}
+
+	y->hi = yhi;
+	y->lo = ylo;
+}
+
+void
+_f2v(Vlong *y, float f)
+{
+
+	_d2v(y, f);
+}
+
+double
+_v2d(Vlong x)
+{
+	if(x.hi & SIGN(32)) {
+		if(x.lo) {
+			x.lo = -x.lo;
+			x.hi = ~x.hi;
+		} else
+			x.hi = -x.hi;
+		return -((long)x.hi*4294967296. + x.lo);
+	}
+	return (long)x.hi*4294967296. + x.lo;
+}
+
+float
+_v2f(Vlong x)
+{
+	return _v2d(x);
+}
+
+static void
+dodiv(Vlong num, Vlong den, Vlong *qp, Vlong *rp)
+{
+	ulong numlo, numhi, denhi, denlo, quohi, quolo, t;
+	int i;
+
+	numhi = num.hi;
+	numlo = num.lo;
+	denhi = den.hi;
+	denlo = den.lo;
+
+	/*
+	 * get a divide by zero
+	 */
+	if(denlo==0 && denhi==0) {
+		numlo = numlo / denlo;
+	}
+
+	/*
+	 * set up the divisor and find the number of iterations needed
+	 */
+	if(numhi >= SIGN(32)) {
+		quohi = SIGN(32);
+		quolo = 0;
+	} else {
+		quohi = numhi;
+		quolo = numlo;
+	}
+	i = 0;
+	while(denhi < quohi || (denhi == quohi && denlo < quolo)) {
+		denhi = (denhi<<1) | (denlo>>31);
+		denlo <<= 1;
+		i++;
+	}
+
+	quohi = 0;
+	quolo = 0;
+	for(; i >= 0; i--) {
+		quohi = (quohi<<1) | (quolo>>31);
+		quolo <<= 1;
+		if(numhi > denhi || (numhi == denhi && numlo >= denlo)) {
+			t = numlo;
+			numlo -= denlo;
+			if(numlo > t)
+				numhi--;
+			numhi -= denhi;
+			quolo |= 1;
+		}
+		denlo = (denlo>>1) | (denhi<<31);
+		denhi >>= 1;
+	}
+
+	if(qp) {
+		qp->lo = quolo;
+		qp->hi = quohi;
+	}
+	if(rp) {
+		rp->lo = numlo;
+		rp->hi = numhi;
+	}
+}
+
+void
+_divvu(Vlong *q, Vlong n, Vlong d)
+{
+
+	if(n.hi == 0 && d.hi == 0) {
+		q->hi = 0;
+		q->lo = n.lo / d.lo;
+		return;
+	}
+	dodiv(n, d, q, 0);
+}
+
+void
+_modvu(Vlong *r, Vlong n, Vlong d)
+{
+
+	if(n.hi == 0 && d.hi == 0) {
+		r->hi = 0;
+		r->lo = n.lo % d.lo;
+		return;
+	}
+	dodiv(n, d, 0, r);
+}
+
+static void
+vneg(Vlong *v)
+{
+
+	if(v->lo == 0) {
+		v->hi = -v->hi;
+		return;
+	}
+	v->lo = -v->lo;
+	v->hi = ~v->hi;
+}
+
+void
+_divv(Vlong *q, Vlong n, Vlong d)
+{
+	long nneg, dneg;
+
+	if(n.hi == (((long)n.lo)>>31) && d.hi == (((long)d.lo)>>31)) {
+		q->lo = (long)n.lo / (long)d.lo;
+		q->hi = ((long)q->lo) >> 31;
+		return;
+	}
+	nneg = n.hi >> 31;
+	if(nneg)
+		vneg(&n);
+	dneg = d.hi >> 31;
+	if(dneg)
+		vneg(&d);
+	dodiv(n, d, q, 0);
+	if(nneg != dneg)
+		vneg(q);
+}
+
+void
+_modv(Vlong *r, Vlong n, Vlong d)
+{
+	long nneg, dneg;
+
+	if(n.hi == (((long)n.lo)>>31) && d.hi == (((long)d.lo)>>31)) {
+		r->lo = (long)n.lo % (long)d.lo;
+		r->hi = ((long)r->lo) >> 31;
+		return;
+	}
+	nneg = n.hi >> 31;
+	if(nneg)
+		vneg(&n);
+	dneg = d.hi >> 31;
+	if(dneg)
+		vneg(&d);
+	dodiv(n, d, 0, r);
+	if(nneg)
+		vneg(r);
+}
+
+void
+_rshav(Vlong *r, Vlong a, int b)
+{
+	long t;
+
+	t = a.hi;
+	if(b >= 32) {
+		r->hi = t>>31;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->lo = t>>31;
+			return;
+		}
+		r->lo = t >> (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->hi = t;
+		r->lo = a.lo;
+		return;
+	}
+	r->hi = t >> b;
+	r->lo = (t << (32-b)) | (a.lo >> b);
+}
+
+void
+_rshlv(Vlong *r, Vlong a, int b)
+{
+	ulong t;
+
+	t = a.hi;
+	if(b >= 32) {
+		r->hi = 0;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->lo = 0;
+			return;
+		}
+		r->lo = t >> (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->hi = t;
+		r->lo = a.lo;
+		return;
+	}
+	r->hi = t >> b;
+	r->lo = (t << (32-b)) | (a.lo >> b);
+}
+
+void
+_lshv(Vlong *r, Vlong a, int b)
+{
+	ulong t;
+
+	t = a.lo;
+	if(b >= 32) {
+		r->lo = 0;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->hi = 0;
+			return;
+		}
+		r->hi = t << (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->lo = t;
+		r->hi = a.hi;
+		return;
+	}
+	r->lo = t << b;
+	r->hi = (t >> (32-b)) | (a.hi << b);
+}
+
+void
+_andv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi & b.hi;
+	r->lo = a.lo & b.lo;
+}
+
+void
+_orv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi | b.hi;
+	r->lo = a.lo | b.lo;
+}
+
+void
+_xorv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi ^ b.hi;
+	r->lo = a.lo ^ b.lo;
+}
+
+void
+_vpp(Vlong *l, Vlong *r)
+{
+
+	l->hi = r->hi;
+	l->lo = r->lo;
+	r->lo++;
+	if(r->lo == 0)
+		r->hi++;
+}
+
+void
+_vmm(Vlong *l, Vlong *r)
+{
+
+	l->hi = r->hi;
+	l->lo = r->lo;
+	if(r->lo == 0)
+		r->hi--;
+	r->lo--;
+}
+
+void
+_ppv(Vlong *l, Vlong *r)
+{
+
+	r->lo++;
+	if(r->lo == 0)
+		r->hi++;
+	l->hi = r->hi;
+	l->lo = r->lo;
+}
+
+void
+_mmv(Vlong *l, Vlong *r)
+{
+
+	if(r->lo == 0)
+		r->hi--;
+	r->lo--;
+	l->hi = r->hi;
+	l->lo = r->lo;
+}
+
+void
+_vasop(Vlong *ret, void *lv, void fn(Vlong*, Vlong, Vlong), int type, Vlong rv)
+{
+	Vlong t, u;
+
+	u.lo = 0;
+	u.hi = 0;
+	switch(type) {
+	default:
+		abort();
+		break;
+
+	case 1:	/* schar */
+		t.lo = *(schar*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(schar*)lv = u.lo;
+		break;
+
+	case 2:	/* uchar */
+		t.lo = *(uchar*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(uchar*)lv = u.lo;
+		break;
+
+	case 3:	/* short */
+		t.lo = *(short*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(short*)lv = u.lo;
+		break;
+
+	case 4:	/* ushort */
+		t.lo = *(ushort*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(ushort*)lv = u.lo;
+		break;
+
+	case 9:	/* int */
+		t.lo = *(int*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(int*)lv = u.lo;
+		break;
+
+	case 10:	/* uint */
+		t.lo = *(uint*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(uint*)lv = u.lo;
+		break;
+
+	case 5:	/* long */
+		t.lo = *(long*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(long*)lv = u.lo;
+		break;
+
+	case 6:	/* ulong */
+		t.lo = *(ulong*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(ulong*)lv = u.lo;
+		break;
+
+	case 7:	/* vlong */
+	case 8:	/* uvlong */
+		fn(&u, *(Vlong*)lv, rv);
+		*(Vlong*)lv = u;
+		break;
+	}
+	*ret = u;
+}
+
+void
+_p2v(Vlong *ret, void *p)
+{
+	long t;
+
+	t = (ulong)p;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sl2v(Vlong *ret, long sl)
+{
+	long t;
+
+	t = sl;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_ul2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_si2v(Vlong *ret, int si)
+{
+	long t;
+
+	t = si;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_ui2v(Vlong *ret, uint ui)
+{
+	long t;
+
+	t = ui;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sh2v(Vlong *ret, long sh)
+{
+	long t;
+
+	t = (sh << 16) >> 16;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_uh2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul & 0xffff;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sc2v(Vlong *ret, long uc)
+{
+	long t;
+
+	t = (uc << 24) >> 24;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_uc2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul & 0xff;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+long
+_v2sc(Vlong rv)
+{
+	long t;
+
+	t = rv.lo & 0xff;
+	return (t << 24) >> 24;
+}
+
+long
+_v2uc(Vlong rv)
+{
+
+	return rv.lo & 0xff;
+}
+
+long
+_v2sh(Vlong rv)
+{
+	long t;
+
+	t = rv.lo & 0xffff;
+	return (t << 16) >> 16;
+}
+
+long
+_v2uh(Vlong rv)
+{
+
+	return rv.lo & 0xffff;
+}
+
+long
+_v2sl(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2ul(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2si(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2ui(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+int
+_testv(Vlong rv)
+{
+	return rv.lo || rv.hi;
+}
+
+int
+_eqv(Vlong lv, Vlong rv)
+{
+	return lv.lo == rv.lo && lv.hi == rv.hi;
+}
+
+int
+_nev(Vlong lv, Vlong rv)
+{
+	return lv.lo != rv.lo || lv.hi != rv.hi;
+}
+
+int
+_ltv(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi < (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo < rv.lo);
+}
+
+int
+_lev(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi < (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo <= rv.lo);
+}
+
+int
+_gtv(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi > (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo > rv.lo);
+}
+
+int
+_gev(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi > (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo >= rv.lo);
+}
+
+int
+_lov(Vlong lv, Vlong rv)
+{
+	return lv.hi < rv.hi || 
+		(lv.hi == rv.hi && lv.lo < rv.lo);
+}
+
+int
+_lsv(Vlong lv, Vlong rv)
+{
+	return lv.hi < rv.hi || 
+		(lv.hi == rv.hi && lv.lo <= rv.lo);
+}
+
+int
+_hiv(Vlong lv, Vlong rv)
+{
+	return lv.hi > rv.hi || 
+		(lv.hi == rv.hi && lv.lo > rv.lo);
+}
+
+int
+_hsv(Vlong lv, Vlong rv)
+{
+	return lv.hi > rv.hi || 
+		(lv.hi == rv.hi && lv.lo >= rv.lo);
+}
diff --git a/libkern/vlrt-power.c b/libkern/vlrt-power.c
new file mode 100644
index 00000000..6fbd83ca
--- /dev/null
+++ b/libkern/vlrt-power.c
@@ -0,0 +1,718 @@
+typedef	unsigned long	ulong;
+typedef	unsigned int	uint;
+typedef	unsigned short	ushort;
+typedef	unsigned char	uchar;
+typedef	signed char	schar;
+
+#define	SIGN(n)	(1UL<<(n-1))
+
+typedef	struct	Vlong	Vlong;
+struct	Vlong
+{
+	union
+	{
+		struct
+		{
+			ulong	hi;
+			ulong	lo;
+		};
+		struct
+		{
+			ushort	hims;
+			ushort	hils;
+			ushort	loms;
+			ushort	lols;
+		};
+	};
+};
+
+void	abort(void);
+
+void
+_addv(Vlong *r, Vlong a, Vlong b)
+{
+	ulong lo, hi;
+
+	lo = a.lo + b.lo;
+	hi = a.hi + b.hi;
+	if(lo < a.lo)
+		hi++;
+	r->lo = lo;
+	r->hi = hi;
+}
+
+void
+_subv(Vlong *r, Vlong a, Vlong b)
+{
+	ulong lo, hi;
+
+	lo = a.lo - b.lo;
+	hi = a.hi - b.hi;
+	if(lo > a.lo)
+		hi--;
+	r->lo = lo;
+	r->hi = hi;
+}
+
+void
+_d2v(Vlong *y, double d)
+{
+	union { double d; struct Vlong; } x;
+	ulong xhi, xlo, ylo, yhi;
+	int sh;
+
+	x.d = d;
+
+	xhi = (x.hi & 0xfffff) | 0x100000;
+	xlo = x.lo;
+	sh = 1075 - ((x.hi >> 20) & 0x7ff);
+
+	ylo = 0;
+	yhi = 0;
+	if(sh >= 0) {
+		/* v = (hi||lo) >> sh */
+		if(sh < 32) {
+			if(sh == 0) {
+				ylo = xlo;
+				yhi = xhi;
+			} else {
+				ylo = (xlo >> sh) | (xhi << (32-sh));
+				yhi = xhi >> sh;
+			}
+		} else {
+			if(sh == 32) {
+				ylo = xhi;
+			} else
+			if(sh < 64) {
+				ylo = xhi >> (sh-32);
+			}
+		}
+	} else {
+		/* v = (hi||lo) << -sh */
+		sh = -sh;
+		if(sh <= 10) {
+			ylo = xlo << sh;
+			yhi = (xhi << sh) | (xlo >> (32-sh));
+		} else {
+			/* overflow */
+			yhi = d;	/* causes something awful */
+		}
+	}
+	if(x.hi & SIGN(32)) {
+		if(ylo != 0) {
+			ylo = -ylo;
+			yhi = ~yhi;
+		} else
+			yhi = -yhi;
+	}
+
+	y->hi = yhi;
+	y->lo = ylo;
+}
+
+void
+_f2v(Vlong *y, float f)
+{
+
+	_d2v(y, f);
+}
+
+double
+_v2d(Vlong x)
+{
+	if(x.hi & SIGN(32)) {
+		if(x.lo) {
+			x.lo = -x.lo;
+			x.hi = ~x.hi;
+		} else
+			x.hi = -x.hi;
+		return -((long)x.hi*4294967296. + x.lo);
+	}
+	return (long)x.hi*4294967296. + x.lo;
+}
+
+float
+_v2f(Vlong x)
+{
+	return _v2d(x);
+}
+
+static void
+dodiv(Vlong num, Vlong den, Vlong *q, Vlong *r)
+{
+	ulong numlo, numhi, denhi, denlo, quohi, quolo, t;
+	int i;
+
+	numhi = num.hi;
+	numlo = num.lo;
+	denhi = den.hi;
+	denlo = den.lo;
+
+	/*
+	 * get a divide by zero
+	 */
+	if(denlo==0 && denhi==0) {
+		numlo = numlo / denlo;
+	}
+
+	/*
+	 * set up the divisor and find the number of iterations needed
+	 */
+	if(numhi >= SIGN(32)) {
+		quohi = SIGN(32);
+		quolo = 0;
+	} else {
+		quohi = numhi;
+		quolo = numlo;
+	}
+	i = 0;
+	while(denhi < quohi || (denhi == quohi && denlo < quolo)) {
+		denhi = (denhi<<1) | (denlo>>31);
+		denlo <<= 1;
+		i++;
+	}
+
+	quohi = 0;
+	quolo = 0;
+	for(; i >= 0; i--) {
+		quohi = (quohi<<1) | (quolo>>31);
+		quolo <<= 1;
+		if(numhi > denhi || (numhi == denhi && numlo >= denlo)) {
+			t = numlo;
+			numlo -= denlo;
+			if(numlo > t)
+				numhi--;
+			numhi -= denhi;
+			quolo |= 1;
+		}
+		denlo = (denlo>>1) | (denhi<<31);
+		denhi >>= 1;
+	}
+
+	if(q) {
+		q->lo = quolo;
+		q->hi = quohi;
+	}
+	if(r) {
+		r->lo = numlo;
+		r->hi = numhi;
+	}
+}
+
+void
+_divvu(Vlong *q, Vlong n, Vlong d)
+{
+
+	if(n.hi == 0 && d.hi == 0) {
+		q->hi = 0;
+		q->lo = n.lo / d.lo;
+		return;
+	}
+	dodiv(n, d, q, 0);
+}
+
+void
+_modvu(Vlong *r, Vlong n, Vlong d)
+{
+
+	if(n.hi == 0 && d.hi == 0) {
+		r->hi = 0;
+		r->lo = n.lo % d.lo;
+		return;
+	}
+	dodiv(n, d, 0, r);
+}
+
+static void
+vneg(Vlong *v)
+{
+
+	if(v->lo == 0) {
+		v->hi = -v->hi;
+		return;
+	}
+	v->lo = -v->lo;
+	v->hi = ~v->hi;
+}
+
+void
+_divv(Vlong *q, Vlong n, Vlong d)
+{
+	long nneg, dneg;
+
+	if(n.hi == (((long)n.lo)>>31) && d.hi == (((long)d.lo)>>31)) {
+		q->lo = (long)n.lo / (long)d.lo;
+		q->hi = ((long)q->lo) >> 31;
+		return;
+	}
+	nneg = n.hi >> 31;
+	if(nneg)
+		vneg(&n);
+	dneg = d.hi >> 31;
+	if(dneg)
+		vneg(&d);
+	dodiv(n, d, q, 0);
+	if(nneg != dneg)
+		vneg(q);
+}
+
+void
+_modv(Vlong *r, Vlong n, Vlong d)
+{
+	long nneg, dneg;
+
+	if(n.hi == (((long)n.lo)>>31) && d.hi == (((long)d.lo)>>31)) {
+		r->lo = (long)n.lo % (long)d.lo;
+		r->hi = ((long)r->lo) >> 31;
+		return;
+	}
+	nneg = n.hi >> 31;
+	if(nneg)
+		vneg(&n);
+	dneg = d.hi >> 31;
+	if(dneg)
+		vneg(&d);
+	dodiv(n, d, 0, r);
+	if(nneg)
+		vneg(r);
+}
+
+void
+_rshav(Vlong *r, Vlong a, int b)
+{
+	long t;
+
+	t = a.hi;
+	if(b >= 32) {
+		r->hi = t>>31;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->lo = t>>31;
+			return;
+		}
+		r->lo = t >> (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->hi = t;
+		r->lo = a.lo;
+		return;
+	}
+	r->hi = t >> b;
+	r->lo = (t << (32-b)) | (a.lo >> b);
+}
+
+void
+_rshlv(Vlong *r, Vlong a, int b)
+{
+	ulong t;
+
+	t = a.hi;
+	if(b >= 32) {
+		r->hi = 0;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->lo = 0;
+			return;
+		}
+		r->lo = t >> (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->hi = t;
+		r->lo = a.lo;
+		return;
+	}
+	r->hi = t >> b;
+	r->lo = (t << (32-b)) | (a.lo >> b);
+}
+
+void
+_lshv(Vlong *r, Vlong a, int b)
+{
+	ulong t;
+
+	t = a.lo;
+	if(b >= 32) {
+		r->lo = 0;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->hi = 0;
+			return;
+		}
+		r->hi = t << (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->lo = t;
+		r->hi = a.hi;
+		return;
+	}
+	r->lo = t << b;
+	r->hi = (t >> (32-b)) | (a.hi << b);
+}
+
+void
+_andv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi & b.hi;
+	r->lo = a.lo & b.lo;
+}
+
+void
+_orv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi | b.hi;
+	r->lo = a.lo | b.lo;
+}
+
+void
+_xorv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi ^ b.hi;
+	r->lo = a.lo ^ b.lo;
+}
+
+void
+_vpp(Vlong *l, Vlong *r)
+{
+
+	l->hi = r->hi;
+	l->lo = r->lo;
+	r->lo++;
+	if(r->lo == 0)
+		r->hi++;
+}
+
+void
+_vmm(Vlong *l, Vlong *r)
+{
+
+	l->hi = r->hi;
+	l->lo = r->lo;
+	if(r->lo == 0)
+		r->hi--;
+	r->lo--;
+}
+
+void
+_ppv(Vlong *l, Vlong *r)
+{
+
+	r->lo++;
+	if(r->lo == 0)
+		r->hi++;
+	l->hi = r->hi;
+	l->lo = r->lo;
+}
+
+void
+_mmv(Vlong *l, Vlong *r)
+{
+
+	if(r->lo == 0)
+		r->hi--;
+	r->lo--;
+	l->hi = r->hi;
+	l->lo = r->lo;
+}
+
+void
+_vasop(Vlong *ret, void *lv, void fn(Vlong*, Vlong, Vlong), int type, Vlong rv)
+{
+	Vlong t, u;
+
+	u = *ret;
+	switch(type) {
+	default:
+		abort();
+		break;
+
+	case 1:	/* schar */
+		t.lo = *(schar*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(schar*)lv = u.lo;
+		break;
+
+	case 2:	/* uchar */
+		t.lo = *(uchar*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(uchar*)lv = u.lo;
+		break;
+
+	case 3:	/* short */
+		t.lo = *(short*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(short*)lv = u.lo;
+		break;
+
+	case 4:	/* ushort */
+		t.lo = *(ushort*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(ushort*)lv = u.lo;
+		break;
+
+	case 9:	/* int */
+		t.lo = *(int*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(int*)lv = u.lo;
+		break;
+
+	case 10:	/* uint */
+		t.lo = *(uint*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(uint*)lv = u.lo;
+		break;
+
+	case 5:	/* long */
+		t.lo = *(long*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(long*)lv = u.lo;
+		break;
+
+	case 6:	/* ulong */
+		t.lo = *(ulong*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(ulong*)lv = u.lo;
+		break;
+
+	case 7:	/* vlong */
+	case 8:	/* uvlong */
+		fn(&u, *(Vlong*)lv, rv);
+		*(Vlong*)lv = u;
+		break;
+	}
+	*ret = u;
+}
+
+void
+_p2v(Vlong *ret, void *p)
+{
+	long t;
+
+	t = (ulong)p;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sl2v(Vlong *ret, long sl)
+{
+	long t;
+
+	t = sl;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_ul2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_si2v(Vlong *ret, int si)
+{
+	long t;
+
+	t = si;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_ui2v(Vlong *ret, uint ui)
+{
+	long t;
+
+	t = ui;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sh2v(Vlong *ret, long sh)
+{
+	long t;
+
+	t = (sh << 16) >> 16;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_uh2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul & 0xffff;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sc2v(Vlong *ret, long uc)
+{
+	long t;
+
+	t = (uc << 24) >> 24;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_uc2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul & 0xff;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+long
+_v2sc(Vlong rv)
+{
+	long t;
+
+	t = rv.lo & 0xff;
+	return (t << 24) >> 24;
+}
+
+long
+_v2uc(Vlong rv)
+{
+
+	return rv.lo & 0xff;
+}
+
+long
+_v2sh(Vlong rv)
+{
+	long t;
+
+	t = rv.lo & 0xffff;
+	return (t << 16) >> 16;
+}
+
+long
+_v2uh(Vlong rv)
+{
+
+	return rv.lo & 0xffff;
+}
+
+long
+_v2sl(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2ul(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2si(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2ui(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+int
+_testv(Vlong rv)
+{
+	return rv.lo || rv.hi;
+}
+
+int
+_eqv(Vlong lv, Vlong rv)
+{
+	return lv.lo == rv.lo && lv.hi == rv.hi;
+}
+
+int
+_nev(Vlong lv, Vlong rv)
+{
+	return lv.lo != rv.lo || lv.hi != rv.hi;
+}
+
+int
+_ltv(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi < (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo < rv.lo);
+}
+
+int
+_lev(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi < (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo <= rv.lo);
+}
+
+int
+_gtv(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi > (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo > rv.lo);
+}
+
+int
+_gev(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi > (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo >= rv.lo);
+}
+
+int
+_lov(Vlong lv, Vlong rv)
+{
+	return lv.hi < rv.hi || 
+		(lv.hi == rv.hi && lv.lo < rv.lo);
+}
+
+int
+_lsv(Vlong lv, Vlong rv)
+{
+	return lv.hi < rv.hi || 
+		(lv.hi == rv.hi && lv.lo <= rv.lo);
+}
+
+int
+_hiv(Vlong lv, Vlong rv)
+{
+	return lv.hi > rv.hi || 
+		(lv.hi == rv.hi && lv.lo > rv.lo);
+}
+
+int
+_hsv(Vlong lv, Vlong rv)
+{
+	return lv.hi > rv.hi || 
+		(lv.hi == rv.hi && lv.lo >= rv.lo);
+}
diff --git a/libkern/vlrt-sparc.c b/libkern/vlrt-sparc.c
new file mode 100644
index 00000000..6fbd83ca
--- /dev/null
+++ b/libkern/vlrt-sparc.c
@@ -0,0 +1,718 @@
+typedef	unsigned long	ulong;
+typedef	unsigned int	uint;
+typedef	unsigned short	ushort;
+typedef	unsigned char	uchar;
+typedef	signed char	schar;
+
+#define	SIGN(n)	(1UL<<(n-1))
+
+typedef	struct	Vlong	Vlong;
+struct	Vlong
+{
+	union
+	{
+		struct
+		{
+			ulong	hi;
+			ulong	lo;
+		};
+		struct
+		{
+			ushort	hims;
+			ushort	hils;
+			ushort	loms;
+			ushort	lols;
+		};
+	};
+};
+
+void	abort(void);
+
+void
+_addv(Vlong *r, Vlong a, Vlong b)
+{
+	ulong lo, hi;
+
+	lo = a.lo + b.lo;
+	hi = a.hi + b.hi;
+	if(lo < a.lo)
+		hi++;
+	r->lo = lo;
+	r->hi = hi;
+}
+
+void
+_subv(Vlong *r, Vlong a, Vlong b)
+{
+	ulong lo, hi;
+
+	lo = a.lo - b.lo;
+	hi = a.hi - b.hi;
+	if(lo > a.lo)
+		hi--;
+	r->lo = lo;
+	r->hi = hi;
+}
+
+void
+_d2v(Vlong *y, double d)
+{
+	union { double d; struct Vlong; } x;
+	ulong xhi, xlo, ylo, yhi;
+	int sh;
+
+	x.d = d;
+
+	xhi = (x.hi & 0xfffff) | 0x100000;
+	xlo = x.lo;
+	sh = 1075 - ((x.hi >> 20) & 0x7ff);
+
+	ylo = 0;
+	yhi = 0;
+	if(sh >= 0) {
+		/* v = (hi||lo) >> sh */
+		if(sh < 32) {
+			if(sh == 0) {
+				ylo = xlo;
+				yhi = xhi;
+			} else {
+				ylo = (xlo >> sh) | (xhi << (32-sh));
+				yhi = xhi >> sh;
+			}
+		} else {
+			if(sh == 32) {
+				ylo = xhi;
+			} else
+			if(sh < 64) {
+				ylo = xhi >> (sh-32);
+			}
+		}
+	} else {
+		/* v = (hi||lo) << -sh */
+		sh = -sh;
+		if(sh <= 10) {
+			ylo = xlo << sh;
+			yhi = (xhi << sh) | (xlo >> (32-sh));
+		} else {
+			/* overflow */
+			yhi = d;	/* causes something awful */
+		}
+	}
+	if(x.hi & SIGN(32)) {
+		if(ylo != 0) {
+			ylo = -ylo;
+			yhi = ~yhi;
+		} else
+			yhi = -yhi;
+	}
+
+	y->hi = yhi;
+	y->lo = ylo;
+}
+
+void
+_f2v(Vlong *y, float f)
+{
+
+	_d2v(y, f);
+}
+
+double
+_v2d(Vlong x)
+{
+	if(x.hi & SIGN(32)) {
+		if(x.lo) {
+			x.lo = -x.lo;
+			x.hi = ~x.hi;
+		} else
+			x.hi = -x.hi;
+		return -((long)x.hi*4294967296. + x.lo);
+	}
+	return (long)x.hi*4294967296. + x.lo;
+}
+
+float
+_v2f(Vlong x)
+{
+	return _v2d(x);
+}
+
+static void
+dodiv(Vlong num, Vlong den, Vlong *q, Vlong *r)
+{
+	ulong numlo, numhi, denhi, denlo, quohi, quolo, t;
+	int i;
+
+	numhi = num.hi;
+	numlo = num.lo;
+	denhi = den.hi;
+	denlo = den.lo;
+
+	/*
+	 * get a divide by zero
+	 */
+	if(denlo==0 && denhi==0) {
+		numlo = numlo / denlo;
+	}
+
+	/*
+	 * set up the divisor and find the number of iterations needed
+	 */
+	if(numhi >= SIGN(32)) {
+		quohi = SIGN(32);
+		quolo = 0;
+	} else {
+		quohi = numhi;
+		quolo = numlo;
+	}
+	i = 0;
+	while(denhi < quohi || (denhi == quohi && denlo < quolo)) {
+		denhi = (denhi<<1) | (denlo>>31);
+		denlo <<= 1;
+		i++;
+	}
+
+	quohi = 0;
+	quolo = 0;
+	for(; i >= 0; i--) {
+		quohi = (quohi<<1) | (quolo>>31);
+		quolo <<= 1;
+		if(numhi > denhi || (numhi == denhi && numlo >= denlo)) {
+			t = numlo;
+			numlo -= denlo;
+			if(numlo > t)
+				numhi--;
+			numhi -= denhi;
+			quolo |= 1;
+		}
+		denlo = (denlo>>1) | (denhi<<31);
+		denhi >>= 1;
+	}
+
+	if(q) {
+		q->lo = quolo;
+		q->hi = quohi;
+	}
+	if(r) {
+		r->lo = numlo;
+		r->hi = numhi;
+	}
+}
+
+void
+_divvu(Vlong *q, Vlong n, Vlong d)
+{
+
+	if(n.hi == 0 && d.hi == 0) {
+		q->hi = 0;
+		q->lo = n.lo / d.lo;
+		return;
+	}
+	dodiv(n, d, q, 0);
+}
+
+void
+_modvu(Vlong *r, Vlong n, Vlong d)
+{
+
+	if(n.hi == 0 && d.hi == 0) {
+		r->hi = 0;
+		r->lo = n.lo % d.lo;
+		return;
+	}
+	dodiv(n, d, 0, r);
+}
+
+static void
+vneg(Vlong *v)
+{
+
+	if(v->lo == 0) {
+		v->hi = -v->hi;
+		return;
+	}
+	v->lo = -v->lo;
+	v->hi = ~v->hi;
+}
+
+void
+_divv(Vlong *q, Vlong n, Vlong d)
+{
+	long nneg, dneg;
+
+	if(n.hi == (((long)n.lo)>>31) && d.hi == (((long)d.lo)>>31)) {
+		q->lo = (long)n.lo / (long)d.lo;
+		q->hi = ((long)q->lo) >> 31;
+		return;
+	}
+	nneg = n.hi >> 31;
+	if(nneg)
+		vneg(&n);
+	dneg = d.hi >> 31;
+	if(dneg)
+		vneg(&d);
+	dodiv(n, d, q, 0);
+	if(nneg != dneg)
+		vneg(q);
+}
+
+void
+_modv(Vlong *r, Vlong n, Vlong d)
+{
+	long nneg, dneg;
+
+	if(n.hi == (((long)n.lo)>>31) && d.hi == (((long)d.lo)>>31)) {
+		r->lo = (long)n.lo % (long)d.lo;
+		r->hi = ((long)r->lo) >> 31;
+		return;
+	}
+	nneg = n.hi >> 31;
+	if(nneg)
+		vneg(&n);
+	dneg = d.hi >> 31;
+	if(dneg)
+		vneg(&d);
+	dodiv(n, d, 0, r);
+	if(nneg)
+		vneg(r);
+}
+
+void
+_rshav(Vlong *r, Vlong a, int b)
+{
+	long t;
+
+	t = a.hi;
+	if(b >= 32) {
+		r->hi = t>>31;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->lo = t>>31;
+			return;
+		}
+		r->lo = t >> (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->hi = t;
+		r->lo = a.lo;
+		return;
+	}
+	r->hi = t >> b;
+	r->lo = (t << (32-b)) | (a.lo >> b);
+}
+
+void
+_rshlv(Vlong *r, Vlong a, int b)
+{
+	ulong t;
+
+	t = a.hi;
+	if(b >= 32) {
+		r->hi = 0;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->lo = 0;
+			return;
+		}
+		r->lo = t >> (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->hi = t;
+		r->lo = a.lo;
+		return;
+	}
+	r->hi = t >> b;
+	r->lo = (t << (32-b)) | (a.lo >> b);
+}
+
+void
+_lshv(Vlong *r, Vlong a, int b)
+{
+	ulong t;
+
+	t = a.lo;
+	if(b >= 32) {
+		r->lo = 0;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->hi = 0;
+			return;
+		}
+		r->hi = t << (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->lo = t;
+		r->hi = a.hi;
+		return;
+	}
+	r->lo = t << b;
+	r->hi = (t >> (32-b)) | (a.hi << b);
+}
+
+void
+_andv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi & b.hi;
+	r->lo = a.lo & b.lo;
+}
+
+void
+_orv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi | b.hi;
+	r->lo = a.lo | b.lo;
+}
+
+void
+_xorv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi ^ b.hi;
+	r->lo = a.lo ^ b.lo;
+}
+
+void
+_vpp(Vlong *l, Vlong *r)
+{
+
+	l->hi = r->hi;
+	l->lo = r->lo;
+	r->lo++;
+	if(r->lo == 0)
+		r->hi++;
+}
+
+void
+_vmm(Vlong *l, Vlong *r)
+{
+
+	l->hi = r->hi;
+	l->lo = r->lo;
+	if(r->lo == 0)
+		r->hi--;
+	r->lo--;
+}
+
+void
+_ppv(Vlong *l, Vlong *r)
+{
+
+	r->lo++;
+	if(r->lo == 0)
+		r->hi++;
+	l->hi = r->hi;
+	l->lo = r->lo;
+}
+
+void
+_mmv(Vlong *l, Vlong *r)
+{
+
+	if(r->lo == 0)
+		r->hi--;
+	r->lo--;
+	l->hi = r->hi;
+	l->lo = r->lo;
+}
+
+void
+_vasop(Vlong *ret, void *lv, void fn(Vlong*, Vlong, Vlong), int type, Vlong rv)
+{
+	Vlong t, u;
+
+	u = *ret;
+	switch(type) {
+	default:
+		abort();
+		break;
+
+	case 1:	/* schar */
+		t.lo = *(schar*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(schar*)lv = u.lo;
+		break;
+
+	case 2:	/* uchar */
+		t.lo = *(uchar*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(uchar*)lv = u.lo;
+		break;
+
+	case 3:	/* short */
+		t.lo = *(short*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(short*)lv = u.lo;
+		break;
+
+	case 4:	/* ushort */
+		t.lo = *(ushort*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(ushort*)lv = u.lo;
+		break;
+
+	case 9:	/* int */
+		t.lo = *(int*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(int*)lv = u.lo;
+		break;
+
+	case 10:	/* uint */
+		t.lo = *(uint*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(uint*)lv = u.lo;
+		break;
+
+	case 5:	/* long */
+		t.lo = *(long*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(long*)lv = u.lo;
+		break;
+
+	case 6:	/* ulong */
+		t.lo = *(ulong*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(ulong*)lv = u.lo;
+		break;
+
+	case 7:	/* vlong */
+	case 8:	/* uvlong */
+		fn(&u, *(Vlong*)lv, rv);
+		*(Vlong*)lv = u;
+		break;
+	}
+	*ret = u;
+}
+
+void
+_p2v(Vlong *ret, void *p)
+{
+	long t;
+
+	t = (ulong)p;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sl2v(Vlong *ret, long sl)
+{
+	long t;
+
+	t = sl;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_ul2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_si2v(Vlong *ret, int si)
+{
+	long t;
+
+	t = si;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_ui2v(Vlong *ret, uint ui)
+{
+	long t;
+
+	t = ui;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sh2v(Vlong *ret, long sh)
+{
+	long t;
+
+	t = (sh << 16) >> 16;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_uh2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul & 0xffff;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sc2v(Vlong *ret, long uc)
+{
+	long t;
+
+	t = (uc << 24) >> 24;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_uc2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul & 0xff;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+long
+_v2sc(Vlong rv)
+{
+	long t;
+
+	t = rv.lo & 0xff;
+	return (t << 24) >> 24;
+}
+
+long
+_v2uc(Vlong rv)
+{
+
+	return rv.lo & 0xff;
+}
+
+long
+_v2sh(Vlong rv)
+{
+	long t;
+
+	t = rv.lo & 0xffff;
+	return (t << 16) >> 16;
+}
+
+long
+_v2uh(Vlong rv)
+{
+
+	return rv.lo & 0xffff;
+}
+
+long
+_v2sl(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2ul(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2si(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2ui(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+int
+_testv(Vlong rv)
+{
+	return rv.lo || rv.hi;
+}
+
+int
+_eqv(Vlong lv, Vlong rv)
+{
+	return lv.lo == rv.lo && lv.hi == rv.hi;
+}
+
+int
+_nev(Vlong lv, Vlong rv)
+{
+	return lv.lo != rv.lo || lv.hi != rv.hi;
+}
+
+int
+_ltv(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi < (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo < rv.lo);
+}
+
+int
+_lev(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi < (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo <= rv.lo);
+}
+
+int
+_gtv(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi > (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo > rv.lo);
+}
+
+int
+_gev(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi > (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo >= rv.lo);
+}
+
+int
+_lov(Vlong lv, Vlong rv)
+{
+	return lv.hi < rv.hi || 
+		(lv.hi == rv.hi && lv.lo < rv.lo);
+}
+
+int
+_lsv(Vlong lv, Vlong rv)
+{
+	return lv.hi < rv.hi || 
+		(lv.hi == rv.hi && lv.lo <= rv.lo);
+}
+
+int
+_hiv(Vlong lv, Vlong rv)
+{
+	return lv.hi > rv.hi || 
+		(lv.hi == rv.hi && lv.lo > rv.lo);
+}
+
+int
+_hsv(Vlong lv, Vlong rv)
+{
+	return lv.hi > rv.hi || 
+		(lv.hi == rv.hi && lv.lo >= rv.lo);
+}
diff --git a/libkern/vlrt-thumb.c b/libkern/vlrt-thumb.c
new file mode 100644
index 00000000..58cf5591
--- /dev/null
+++ b/libkern/vlrt-thumb.c
@@ -0,0 +1,718 @@
+typedef	unsigned long	ulong;
+typedef	unsigned int	uint;
+typedef	unsigned short	ushort;
+typedef	unsigned char	uchar;
+typedef	signed char	schar;
+
+#define	SIGN(n)	(1UL<<(n-1))
+
+typedef	struct	Vlong	Vlong;
+struct	Vlong
+{
+	union
+	{
+		struct
+		{
+			ulong	hi;
+			ulong	lo;
+		};
+		struct
+		{
+			ushort	hims;
+			ushort	hils;
+			ushort	loms;
+			ushort	lols;
+		};
+	};
+};
+
+void	abort(void);
+
+void
+_addv(Vlong *r, Vlong a, Vlong b)
+{
+	ulong lo, hi;
+
+	lo = a.lo + b.lo;
+	hi = a.hi + b.hi;
+	if(lo < a.lo)
+		hi++;
+	r->lo = lo;
+	r->hi = hi;
+}
+
+void
+_subv(Vlong *r, Vlong a, Vlong b)
+{
+	ulong lo, hi;
+
+	lo = a.lo - b.lo;
+	hi = a.hi - b.hi;
+	if(lo > a.lo)
+		hi--;
+	r->lo = lo;
+	r->hi = hi;
+}
+
+
+void
+_d2v(Vlong *y, double d)
+{
+	union { double d; struct Vlong; } x;
+	ulong xhi, xlo, ylo, yhi;
+	int sh;
+
+	x.d = d;
+
+	xhi = (x.hi & 0xfffff) | 0x100000;
+	xlo = x.lo;
+	sh = 1075 - ((x.hi >> 20) & 0x7ff);
+
+	ylo = 0;
+	yhi = 0;
+	if(sh >= 0) {
+		/* v = (hi||lo) >> sh */
+		if(sh < 32) {
+			if(sh == 0) {
+				ylo = xlo;
+				yhi = xhi;
+			} else {
+				ylo = (xlo >> sh) | (xhi << (32-sh));
+				yhi = xhi >> sh;
+			}
+		} else {
+			if(sh == 32) {
+				ylo = xhi;
+			} else
+			if(sh < 64) {
+				ylo = xhi >> (sh-32);
+			}
+		}
+	} else {
+		/* v = (hi||lo) << -sh */
+		sh = -sh;
+		if(sh <= 10) {
+			ylo = xlo << sh;
+			yhi = (xhi << sh) | (xlo >> (32-sh));
+		} else {
+			/* overflow */
+			yhi = d;	/* causes something awful */
+		}
+	}
+	if(x.hi & SIGN(32)) {
+		if(ylo != 0) {
+			ylo = -ylo;
+			yhi = ~yhi;
+		} else
+			yhi = -yhi;
+	}
+
+	y->hi = yhi;
+	y->lo = ylo;
+}
+
+void
+_f2v(Vlong *y, float f)
+{
+	_d2v(y, f);
+}
+
+double
+_v2d(Vlong x)
+{
+	if(x.hi & SIGN(32)) {
+		if(x.lo) {
+			x.lo = -x.lo;
+			x.hi = ~x.hi;
+		} else
+			x.hi = -x.hi;
+		return -((long)x.hi*4294967296. + x.lo);
+	}
+	return (long)x.hi*4294967296. + x.lo;
+}
+
+float
+_v2f(Vlong x)
+{
+	return _v2d(x);
+}
+
+static void
+dodiv(Vlong num, Vlong den, Vlong *q, Vlong *r)
+{
+	ulong numlo, numhi, denhi, denlo, quohi, quolo, t;
+	int i;
+
+	numhi = num.hi;
+	numlo = num.lo;
+	denhi = den.hi;
+	denlo = den.lo;
+
+	/*
+	 * get a divide by zero
+	 */
+	if(denlo==0 && denhi==0) {
+		numlo = numlo / denlo;
+	}
+
+	/*
+	 * set up the divisor and find the number of iterations needed
+	 */
+	if(numhi >= SIGN(32)) {
+		quohi = SIGN(32);
+		quolo = 0;
+	} else {
+		quohi = numhi;
+		quolo = numlo;
+	}
+	i = 0;
+	while(denhi < quohi || (denhi == quohi && denlo < quolo)) {
+		denhi = (denhi<<1) | (denlo>>31);
+		denlo <<= 1;
+		i++;
+	}
+
+	quohi = 0;
+	quolo = 0;
+	for(; i >= 0; i--) {
+		quohi = (quohi<<1) | (quolo>>31);
+		quolo <<= 1;
+		if(numhi > denhi || (numhi == denhi && numlo >= denlo)) {
+			t = numlo;
+			numlo -= denlo;
+			if(numlo > t)
+				numhi--;
+			numhi -= denhi;
+			quolo |= 1;
+		}
+		denlo = (denlo>>1) | (denhi<<31);
+		denhi >>= 1;
+	}
+
+	if(q) {
+		q->lo = quolo;
+		q->hi = quohi;
+	}
+	if(r) {
+		r->lo = numlo;
+		r->hi = numhi;
+	}
+}
+
+void
+_divvu(Vlong *q, Vlong n, Vlong d)
+{
+
+	if(n.hi == 0 && d.hi == 0) {
+		q->hi = 0;
+		q->lo = n.lo / d.lo;
+		return;
+	}
+	dodiv(n, d, q, 0);
+}
+
+void
+_modvu(Vlong *r, Vlong n, Vlong d)
+{
+
+	if(n.hi == 0 && d.hi == 0) {
+		r->hi = 0;
+		r->lo = n.lo % d.lo;
+		return;
+	}
+	dodiv(n, d, 0, r);
+}
+
+static void
+vneg(Vlong *v)
+{
+
+	if(v->lo == 0) {
+		v->hi = -v->hi;
+		return;
+	}
+	v->lo = -v->lo;
+	v->hi = ~v->hi;
+}
+
+void
+_divv(Vlong *q, Vlong n, Vlong d)
+{
+	long nneg, dneg;
+
+	if(n.hi == (((long)n.lo)>>31) && d.hi == (((long)d.lo)>>31)) {
+		q->lo = (long)n.lo / (long)d.lo;
+		q->hi = ((long)q->lo) >> 31;
+		return;
+	}
+	nneg = n.hi >> 31;
+	if(nneg)
+		vneg(&n);
+	dneg = d.hi >> 31;
+	if(dneg)
+		vneg(&d);
+	dodiv(n, d, q, 0);
+	if(nneg != dneg)
+		vneg(q);
+}
+
+void
+_modv(Vlong *r, Vlong n, Vlong d)
+{
+	long nneg, dneg;
+
+	if(n.hi == (((long)n.lo)>>31) && d.hi == (((long)d.lo)>>31)) {
+		r->lo = (long)n.lo % (long)d.lo;
+		r->hi = ((long)r->lo) >> 31;
+		return;
+	}
+	nneg = n.hi >> 31;
+	if(nneg)
+		vneg(&n);
+	dneg = d.hi >> 31;
+	if(dneg)
+		vneg(&d);
+	dodiv(n, d, 0, r);
+	if(nneg)
+		vneg(r);
+}
+
+void
+_rshav(Vlong *r, Vlong a, int b)
+{
+	long t;
+
+	t = a.hi;
+	if(b >= 32) {
+		r->hi = t>>31;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->lo = t>>31;
+			return;
+		}
+		r->lo = t >> (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->hi = t;
+		r->lo = a.lo;
+		return;
+	}
+	r->hi = t >> b;
+	r->lo = (t << (32-b)) | (a.lo >> b);
+}
+
+void
+_rshlv(Vlong *r, Vlong a, int b)
+{
+	ulong t;
+
+	t = a.hi;
+	if(b >= 32) {
+		r->hi = 0;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->lo = 0;
+			return;
+		}
+		r->lo = t >> (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->hi = t;
+		r->lo = a.lo;
+		return;
+	}
+	r->hi = t >> b;
+	r->lo = (t << (32-b)) | (a.lo >> b);
+}
+
+void
+_lshv(Vlong *r, Vlong a, int b)
+{
+	ulong t;
+
+	t = a.lo;
+	if(b >= 32) {
+		r->lo = 0;
+		if(b >= 64) {
+			/* this is illegal re C standard */
+			r->hi = 0;
+			return;
+		}
+		r->hi = t << (b-32);
+		return;
+	}
+	if(b <= 0) {
+		r->lo = t;
+		r->hi = a.hi;
+		return;
+	}
+	r->lo = t << b;
+	r->hi = (t >> (32-b)) | (a.hi << b);
+}
+
+void
+_andv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi & b.hi;
+	r->lo = a.lo & b.lo;
+}
+
+void
+_orv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi | b.hi;
+	r->lo = a.lo | b.lo;
+}
+
+void
+_xorv(Vlong *r, Vlong a, Vlong b)
+{
+	r->hi = a.hi ^ b.hi;
+	r->lo = a.lo ^ b.lo;
+}
+
+void
+_vpp(Vlong *l, Vlong *r)
+{
+
+	l->hi = r->hi;
+	l->lo = r->lo;
+	r->lo++;
+	if(r->lo == 0)
+		r->hi++;
+}
+
+void
+_vmm(Vlong *l, Vlong *r)
+{
+
+	l->hi = r->hi;
+	l->lo = r->lo;
+	if(r->lo == 0)
+		r->hi--;
+	r->lo--;
+}
+
+void
+_ppv(Vlong *l, Vlong *r)
+{
+
+	r->lo++;
+	if(r->lo == 0)
+		r->hi++;
+	l->hi = r->hi;
+	l->lo = r->lo;
+}
+
+void
+_mmv(Vlong *l, Vlong *r)
+{
+
+	if(r->lo == 0)
+		r->hi--;
+	r->lo--;
+	l->hi = r->hi;
+	l->lo = r->lo;
+}
+
+void
+_vasop(Vlong *ret, void *lv, void fn(Vlong*, Vlong, Vlong), int type, Vlong rv)
+{
+	Vlong t, u;
+
+	u = *ret;
+	switch(type) {
+	default:
+		abort();
+		break;
+
+	case 1:	/* schar */
+		t.lo = *(schar*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(schar*)lv = u.lo;
+		break;
+
+	case 2:	/* uchar */
+		t.lo = *(uchar*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(uchar*)lv = u.lo;
+		break;
+
+	case 3:	/* short */
+		t.lo = *(short*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(short*)lv = u.lo;
+		break;
+
+	case 4:	/* ushort */
+		t.lo = *(ushort*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(ushort*)lv = u.lo;
+		break;
+
+	case 9:	/* int */
+		t.lo = *(int*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(int*)lv = u.lo;
+		break;
+
+	case 10:	/* uint */
+		t.lo = *(uint*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(uint*)lv = u.lo;
+		break;
+
+	case 5:	/* long */
+		t.lo = *(long*)lv;
+		t.hi = t.lo >> 31;
+		fn(&u, t, rv);
+		*(long*)lv = u.lo;
+		break;
+
+	case 6:	/* ulong */
+		t.lo = *(ulong*)lv;
+		t.hi = 0;
+		fn(&u, t, rv);
+		*(ulong*)lv = u.lo;
+		break;
+
+	case 7:	/* vlong */
+	case 8:	/* uvlong */
+		fn(&u, *(Vlong*)lv, rv);
+		*(Vlong*)lv = u;
+		break;
+	}
+	*ret = u;
+}
+
+void
+_p2v(Vlong *ret, void *p)
+{
+	long t;
+
+	t = (ulong)p;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sl2v(Vlong *ret, long sl)
+{
+	long t;
+
+	t = sl;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_ul2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_si2v(Vlong *ret, int si)
+{
+	long t;
+
+	t = si;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_ui2v(Vlong *ret, uint ui)
+{
+	long t;
+
+	t = ui;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sh2v(Vlong *ret, long sh)
+{
+	long t;
+
+	t = (sh << 16) >> 16;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_uh2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul & 0xffff;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+void
+_sc2v(Vlong *ret, long uc)
+{
+	long t;
+
+	t = (uc << 24) >> 24;
+	ret->lo = t;
+	ret->hi = t >> 31;
+}
+
+void
+_uc2v(Vlong *ret, ulong ul)
+{
+	long t;
+
+	t = ul & 0xff;
+	ret->lo = t;
+	ret->hi = 0;
+}
+
+long
+_v2sc(Vlong rv)
+{
+	long t;
+
+	t = rv.lo & 0xff;
+	return (t << 24) >> 24;
+}
+
+long
+_v2uc(Vlong rv)
+{
+
+	return rv.lo & 0xff;
+}
+
+long
+_v2sh(Vlong rv)
+{
+	long t;
+
+	t = rv.lo & 0xffff;
+	return (t << 16) >> 16;
+}
+
+long
+_v2uh(Vlong rv)
+{
+
+	return rv.lo & 0xffff;
+}
+
+long
+_v2sl(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2ul(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2si(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+long
+_v2ui(Vlong rv)
+{
+
+	return rv.lo;
+}
+
+int
+_testv(Vlong rv)
+{
+	return rv.lo || rv.hi;
+}
+
+int
+_eqv(Vlong lv, Vlong rv)
+{
+	return lv.lo == rv.lo && lv.hi == rv.hi;
+}
+
+int
+_nev(Vlong lv, Vlong rv)
+{
+	return lv.lo != rv.lo || lv.hi != rv.hi;
+}
+
+int
+_ltv(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi < (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo < rv.lo);
+}
+
+int
+_lev(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi < (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo <= rv.lo);
+}
+
+int
+_gtv(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi > (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo > rv.lo);
+}
+
+int
+_gev(Vlong lv, Vlong rv)
+{
+	return (long)lv.hi > (long)rv.hi || 
+		(lv.hi == rv.hi && lv.lo >= rv.lo);
+}
+
+int
+_lov(Vlong lv, Vlong rv)
+{
+	return lv.hi < rv.hi || 
+		(lv.hi == rv.hi && lv.lo < rv.lo);
+}
+
+int
+_lsv(Vlong lv, Vlong rv)
+{
+	return lv.hi < rv.hi || 
+		(lv.hi == rv.hi && lv.lo <= rv.lo);
+}
+
+int
+_hiv(Vlong lv, Vlong rv)
+{
+	return lv.hi > rv.hi || 
+		(lv.hi == rv.hi && lv.lo > rv.lo);
+}
+
+int
+_hsv(Vlong lv, Vlong rv)
+{
+	return lv.hi > rv.hi || 
+		(lv.hi == rv.hi && lv.lo >= rv.lo);
+}
diff --git a/libkern/vseprint.c b/libkern/vseprint.c
new file mode 100644
index 00000000..7b23d55b
--- /dev/null
+++ b/libkern/vseprint.c
@@ -0,0 +1,34 @@
+/*
+ * The authors of this software are Rob Pike and Ken Thompson.
+ *              Copyright (c) 2002 by Lucent Technologies.
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose without fee is hereby granted, provided that this entire notice
+ * is included in all copies of any software which is or includes a copy
+ * or modification of this software and in all copies of the supporting
+ * documentation for such software.
+ * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR IMPLIED
+ * WARRANTY.  IN PARTICULAR, NEITHER THE AUTHORS NOR LUCENT TECHNOLOGIES MAKE ANY
+ * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE MERCHANTABILITY
+ * OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR PURPOSE.
+ */
+#include "lib9.h"
+
+char*
+vseprint(char *buf, char *e, char *fmt, va_list args)
+{
+	Fmt f;
+
+	if(e <= buf)
+		return nil;
+	f.runes = 0;
+	f.start = buf;
+	f.to = buf;
+	f.stop = e - 1;
+	f.flush = nil;
+	f.farg = nil;
+	f.nfmt = 0;
+	f.args = args;
+	dofmt(&f, fmt);
+	*(char*)f.to = '\0';
+	return f.to;
+}
diff --git a/libkern/vsmprint.c b/libkern/vsmprint.c
new file mode 100644
index 00000000..44789be5
--- /dev/null
+++ b/libkern/vsmprint.c
@@ -0,0 +1,74 @@
+/*
+ * The authors of this software are Rob Pike and Ken Thompson.
+ *              Copyright (c) 2002 by Lucent Technologies.
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose without fee is hereby granted, provided that this entire notice
+ * is included in all copies of any software which is or includes a copy
+ * or modification of this software and in all copies of the supporting
+ * documentation for such software.
+ * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR IMPLIED
+ * WARRANTY.  IN PARTICULAR, NEITHER THE AUTHORS NOR LUCENT TECHNOLOGIES MAKE ANY
+ * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE MERCHANTABILITY
+ * OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR PURPOSE.
+ */
+#include "lib9.h"
+#include "fmtdef.h"
+
+static int
+fmtStrFlush(Fmt *f)
+{
+	char *s;
+	int n;
+
+	n = (int)f->farg;
+	n += 256;
+	f->farg = (void*)n;
+	s = f->start;
+	f->start = realloc(s, n);
+	if(f->start == nil){
+		free(s);
+		f->to = nil;
+		f->stop = nil;
+		return 0;
+	}
+	f->to = (char*)f->start + ((char*)f->to - s);
+	f->stop = (char*)f->start + n - 1;
+	return 1;
+}
+
+int
+fmtstrinit(Fmt *f)
+{
+	int n;
+
+	f->runes = 0;
+	n = 32;
+	f->start = malloc(n);
+	if(f->start == nil)
+		return -1;
+	f->to = f->start;
+	f->stop = (char*)f->start + n - 1;
+	f->flush = fmtStrFlush;
+	f->farg = (void*)n;
+	f->nfmt = 0;
+	return 0;
+}
+
+/*
+ * print into an allocated string buffer
+ */
+char*
+vsmprint(char *fmt, va_list args)
+{
+	Fmt f;
+	int n;
+
+	if(fmtstrinit(&f) < 0)
+		return nil;
+	f.args = args;
+	n = dofmt(&f, fmt);
+	if(n < 0)
+		return nil;
+	*(char*)f.to = '\0';
+	return f.start;
+}
diff --git a/libkern/vsnprint.c b/libkern/vsnprint.c
new file mode 100644
index 00000000..841d142c
--- /dev/null
+++ b/libkern/vsnprint.c
@@ -0,0 +1,34 @@
+/*
+ * The authors of this software are Rob Pike and Ken Thompson.
+ *              Copyright (c) 2002 by Lucent Technologies.
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose without fee is hereby granted, provided that this entire notice
+ * is included in all copies of any software which is or includes a copy
+ * or modification of this software and in all copies of the supporting
+ * documentation for such software.
+ * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR IMPLIED
+ * WARRANTY.  IN PARTICULAR, NEITHER THE AUTHORS NOR LUCENT TECHNOLOGIES MAKE ANY
+ * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE MERCHANTABILITY
+ * OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR PURPOSE.
+ */
+#include "lib9.h"
+
+int
+vsnprint(char *buf, int len, char *fmt, va_list args)
+{
+	Fmt f;
+
+	if(len <= 0)
+		return -1;
+	f.runes = 0;
+	f.start = buf;
+	f.to = buf;
+	f.stop = buf + len - 1;
+	f.flush = nil;
+	f.farg = nil;
+	f.nfmt = 0;
+	f.args = args;
+	dofmt(&f, fmt);
+	*(char*)f.to = '\0';
+	return (char*)f.to - buf;
+}