New interrupt code from Bruce Evans. In additional to Bruce's attached

list of changes, I've made the following additional changes:

1) i386/include/ipl.h renamed to spl.h as the name conflicts with the
   file of the same name in i386/isa/ipl.h.
2) changed all use of *mask (i.e. netmask, biomask, ttymask, etc) to
   *_imask (net_imask, etc).
3) changed vestige of splnet use in if_is to splimp.
4) got rid of "impmask" completely (Bruce had gotten rid of netmask),
   and are now using net_imask instead.
5) dozens of minor cruft to glue in Bruce's changes.

   These require changes I made to config(8) as well, and thus it must
be rebuilt.

-DG

from Bruce Evans:

sio:
	o No diff is supplied.  Remove the define of setsofttty().  I hope
	  that is enough.

*.s:
	o i386/isa/debug.h no longer exists.  The event counters became too
	  much trouble to maintain.  All function call entry and exception
	  entry counters can be recovered by using profiling kernel (the new
	  profiling supports all entry points; however, it is too slow to
	  leave enabled all the time; it also).  Only BDBTRAP() from debug.h
	  is now used.  That is moved to exception.s.  It might be worth
	  preserving SHOW_BITS() and calling it from _mcount() (if enabled).
	o T_ASTFLT is now only set just before calling trap().
	o All exception handlers set SWI_AST_MASK in cpl as soon as possible
	  after entry and arrange for _doreti to restore it atomically with
	  exiting.  It is not possible to set it atomically with entering
	  the kernel, so it must be checked against the user mode bits in
	  the trap frame before committing to using it.  There is no place
	  to store the old value of cpl for syscalls or traps, so there are
	  some complications restoring it.

Profiling stuff (mostly in *.s):
	o Changes to kern/subr_mcount.c, gcc and gprof are not supplied yet.
	o All interesting labels `foo' are renamed `_foo' and all
	  uninteresting labels `_bar' are renamed `bar'.  A small change
	  to gprof allows ignoring labels not starting with underscores.
	o MCOUNT_LABEL() is to provide names for counters for times spent
	  in exception handlers.
	o FAKE_MCOUNT() is a version of MCOUNT() suitable for exception
	  handlers.  Its arg is the pc where the exception occurred.  The
	  new mcount() pretends that this was a call from that pc to a
	  suitable MCOUNT_LABEL().
	o MEXITCOUNT is to turn off any timer started by MCOUNT().

/usr/src/sys/i386/i386/exception.s:
	o The non-BDB BPTTRAP() macros were doing a sti even when interrupts
	  were disabled when the trap occurred.  The sti (fixed) sti is
	  actually a no-op unless you have my changes to machdep.c that make
	  the debugger trap gates interrupt gates, but fixing that would
	  make the ifdefs messier.  ddb seems to be unharmed by both
	  interrupts always disabled and always enabled (I had the branch in
	  the fix back to front for some time :-().
	o There is no known pushal bug.
	o tf_err can be left as garbage for syscalls.

/usr/src/sys/i386/i386/locore.s:
	o Fix and update BDE_DEBUGGER support.
	o ENTRY(btext) before initialization was dangerous.
	o Warm boot shot was longer than intended.

/usr/src/sys/i386/i386/machdep.c:
	o DON'T APPLY ALL OF THIS DIFF.  It's what I'm using, but may require
	  other changes.
	  Use the following:
		o Remove aston() and setsoftclock().
	  Maybe use the following:
		o No netisr.h.
		o Spelling fix.
		o Delay to read the Rebooting message.
		o Fix for vm system unmapping a reduced area of memory
		  after bounds_check_with_label() reduces the size of
		  a physical i/o for a partition boundary.  A similar
		  fix is required in kern_physio.c.
		o Correct use of __CONCAT.  It never worked here for non-
		  ANSI cpp's.  Is it time to drop support for non-ANSI?
		o gdt_segs init.  0xffffffffUL is bogus because ssd_limit
		  is not 32 bits.  The replacement may have the same
		  value :-), but is more natural.
		o physmem was one page too low.  Confusing variable names.
	  Don't use the following:
		o Better numbers of buffers.  Each 8K page requires up to
		  16 buffer headers.  On my system, this results in 5576
		  buffers containing [up to] 2854912 bytes of memory.
		  The usual allocation of about 384 buffers only holds
		  192K of disk if you use it on an fs with a block size
		  of 512.
		o gdt changes for bdb.
		o *TGT -> *IDT changes for bdb.
		o #ifdefed changes for bdb.

/usr/src/sys/i386/i386/microtime.s:
	o Use the correct asm macros.  I think asm.h was copied from Mach
	  just for microtime and isn't used now.  It certainly doesn't
	  belong in <sys>.  Various macros are also duplicated in
	  sys/i386/boot.h and libc/i386/*.h.
	o Don't switch to and from the IRR; it is guaranteed to be selected
	  (default after ICU init and explicitly selected in isa.c too, and
	  never changed until the old microtime clobbered it).

/usr/src/sys/i386/i386/support.s:
	o Non-essential changes (none related to spls or profiling).
	o Removed slow loads of %gs again.  The LDT support may require
	  not relying on %gs, but loading it is not the way to fix it!
	  Some places (copyin ...) forgot to load it.  Loading it clobbers
	  the user %gs.  trap() still loads it after certain types of
	  faults so that fuword() etc can rely on it without loading it
	  explicitly.  Exception handlers don't restore it.  If we want
	  to preserve the user %gs, then the fastest method is to not
	  touch it except for context switches.  Comparing with
	  VM_MAXUSER_ADDRESS and branching takes only 2 or 4 cycles on
	  a 486, while loading %gs takes 9 cycles and using it takes
	  another.
	o Fixed a signed branch to unsigned.

/usr/src/sys/i386/i386/swtch.s:
	o Move spl0() outside of idle loop.
	o Remove cli/sti from idle loop.  sw1 does a cli, and in the
	  unlikely event of an interrupt occurring and whichqs becoming
	  zero, sw1 will just jump back to _idle.
	o There's no spl0() function in asm any more, so use splz().
	o swtch() doesn't need to be superaligned, at least with the
	  new mcounting.
	o Fixed a signed branch to unsigned.
	o Removed astoff().

/usr/src/sys/i386/i386/trap.c:
	o The decentralized extern decls were inconsistent, of course.
	o Fixed typo MATH_EMULTATE in comments. */
	o Removed unused variables.
	o Old netmask is now impmask; print it instead.  Perhaps we
	  should print some of the new masks.
	o BTW, trap() should not print anything for normal debugger
	  traps.

/usr/src/sys/i386/include/asmacros.h:
	o DON'T APPLY ALL OF THIS DIFF.  Just use some of the null macros
	  as necessary.

/usr/src/sys/i386/include/cpu.h:
	o CLKF_BASEPRI() changes since cpl == SWI_AST_MASK is now normal
	  while the kernel is running.
	o Don't use var++ to set boolean variables.  It fails after a mere
	  4G times :-) and is slower than storing a constant on [3-4]86s.

/usr/src/sys/i386/include/cpufunc.h:
	o DON'T APPLY ALL OF THIS DIFF.  You need mainly the include of
	  <machine/ipl.h>.  Unfortunately, <machine/ipl.h> is needed by
	  almost everything for the inlines.

/usr/src/sys/i386/include/ipl.h:
	o New file.  Defines spl inlines and SWI macros and declares most
	  variables related to hard and soft interrupt masks.

/usr/src/sys/i386/isa/icu.h:
	o Moved definitions to <machine/ipl.h>

/usr/src/sys/i386/isa/icu.s:
	o Software interrupts (SWIs) and delayed hardware interrupts (HWIs)
	  are now handled uniformally, and dispatching them from splx() is
	  more like dispatching them from _doreti.  The dispatcher is
	  essentially *(handler[ffs(ipending & ~cpl)]().
	o More care (not quite enough) is taken to avoid unbounded nesting
	  of interrupts.
	o The interface to softclock() is changed so that a trap frame is
	  not required.
	o Fast interrupt handlers are now handled more uniformally.
	  Configuration is still too early (new handlers would require
	  bits in <machine/ipl.h> and functions to vector.s).
	o splnnn() and splx() are no longer here; they are inline functions
	  (could be macros for other compilers).  splz() is the nontrivial
	  part of the old splx().

/usr/src/sys/i386/isa/ipl.h
	o New file.  Supposed to have only bus-dependent stuff.  Perhaps
	  the h/w masks should be declared here.

/usr/src/sys/i386/isa/isa.c:
	o DON'T APPLY ALL OF THIS DIFF.  You need only things involving
	  *mask and *MASK and comments about them.  netmask is now a pure
	  software mask.  It works like the softclock mask.

/usr/src/sys/i386/isa/vector.s:
	o Reorganize AUTO_EOI* macros.
	o Option FAST_INTR_HANDLER_USERS_ES for people who don't trust
	  fastintr handlers.
	o fastintr handlers need to metamorphose into ordinary interrupt
	  handlers if their SWI bit has become set.  Previously, sio had
	  unintended latency for handling output completions and input
	  of SLIP framing characters because this was not done.

/usr/src/sys/net/netisr.h:
	o The machine-dependent stuff is now imported from <machine/ipl.h>.

/usr/src/sys/sys/systm.h
	o DON'T APPLY ALL OF THIS DIFF.  You need mainly the different
	  splx() prototype.  The spl*() prototypes are duplicated as
	  inlines in <machine/ipl.h> but they need to be duplicated here
	  in case there are no inlines.  I sent systm.h and cpufunc.h
	  to Garrett.  We agree that spl0 should be replaced by splnone
	  and not the other way around like I've done.

/usr/src/sys/kern/kern_clock.c
	o splsoftclock() now lowers cpl so the direct call to softclock()
	  works as intended.
	o softclock() interface changed to avoid passing the whole frame
	  (some machines may need another change for profile_tick()).
	o profiling renamed _profiling to avoid ANSI namespace pollution.
	  (I had to improve the mcount() interface and may as well fix it.)
	  The GUPROF variant doesn't actually reference profiling here,
	  but the 'U' in GUPROF should mean to select the microtimer
	  mcount() and not change the interface.

8 files changed, 355 insertions, 436 deletions

diff --git a/sys/i386/isa/icu.h b/sys/i386/isa/icu.h
index 488ad3e..13216b0 100644
--- a/sys/i386/isa/icu.h
+++ b/sys/i386/isa/icu.h
@@ -34,7 +34,7 @@
  * SUCH DAMAGE.
  *
  *	from: @(#)icu.h	5.6 (Berkeley) 5/9/91
- *	$Id$
+ *	$Id: icu.h,v 1.2 1993/10/16 13:45:51 rgrimes Exp $
  */
 
 /*
@@ -51,12 +51,6 @@
  * Interrupt "level" mechanism variables, masks, and macros
  */
 extern	unsigned imen;		/* interrupt mask enable */
-extern	unsigned cpl;		/* current priority level mask */
-
-extern	unsigned highmask;	/* group of interrupts masked with splhigh() */
-extern	unsigned ttymask;	/* group of interrupts masked with spltty() */
-extern	unsigned biomask;	/* group of interrupts masked with splbio() */
-extern	unsigned netmask;	/* group of interrupts masked with splimp() */
 
 #define	INTREN(s)	(imen &= ~(s), SET_ICUS())
 #define	INTRDIS(s)	(imen |= (s), SET_ICUS())
@@ -74,7 +68,7 @@ extern	unsigned netmask;	/* group of interrupts masked with splimp() */
 #endif
 
 /*
- * Interrupt enable bits -- in order of priority
+ * Interrupt enable bits - in normal order of priority (which we change)
  */
 #define	IRQ0		0x0001		/* highest priority - timer */
 #define	IRQ1		0x0002
@@ -88,7 +82,7 @@ extern	unsigned netmask;	/* group of interrupts masked with splimp() */
 #define	IRQ13		0x2000
 #define	IRQ14		0x4000
 #define	IRQ15		0x8000
-#define	IRQ3		0x0008
+#define	IRQ3		0x0008		/* this is highest after rotation */
 #define	IRQ4		0x0010
 #define	IRQ5		0x0020
 #define	IRQ6		0x0040
diff --git a/sys/i386/isa/icu.s b/sys/i386/isa/icu.s
index 5af03e7..b8bf1a8 100644
--- a/sys/i386/isa/icu.s
+++ b/sys/i386/isa/icu.s
@@ -36,7 +36,7 @@
  *
  *	@(#)icu.s	7.2 (Berkeley) 5/21/91
  *
- *	$Id: icu.s,v 1.6 1993/12/19 00:50:35 wollman Exp $
+ *	$Id: icu.s,v 1.7 1993/12/20 14:58:21 wollman Exp $
  */
 
 /*
@@ -45,215 +45,131 @@
  */
 
 /*
- * XXX - this file is now misnamed.  All spls are now soft and the only thing
- * related to the hardware icu is that the bit numbering is the same in the
- * soft priority masks as in the hard ones.
+ * XXX this file should be named ipl.s.  All spls are now soft and the
+ * only thing related to the hardware icu is that the h/w interrupt
+ * numbers are used without translation in the masks.
  */
 
-#include "sio.h"
-#define	HIGHMASK	0xffff
-#define	SOFTCLOCKMASK	0x8000
+#include "../net/netisr.h"
 
 	.data
-
 	.globl	_cpl
-_cpl:	.long	0xffff			/* current priority (all off) */
-
+_cpl:	.long	HWI_MASK | SWI_MASK	/* current priority (all off) */
 	.globl	_imen
-_imen:	.long	0xffff			/* interrupt mask enable (all off) */
-
-/* 	.globl	_highmask	*/
-_highmask:	.long	HIGHMASK
-
-	.globl	_ttymask, _biomask, _netmask
-_ttymask:	.long	0
-_biomask:	.long	0
-_netmask:	.long	0
-
-	.globl	_ipending, _astpending
+_imen:	.long	HWI_MASK		/* interrupt mask enable (all h/w off) */
+_high_imask:	.long	HWI_MASK | SWI_MASK
+	.globl	_tty_imask
+_tty_imask:	.long	0
+	.globl	_bio_imask
+_bio_imask:	.long	0
+	.globl	_net_imask
+_net_imask:	.long	0
+	.globl	_ipending
 _ipending:	.long	0
+	.globl	_astpending
 _astpending:	.long	0		/* tells us an AST needs to be taken */
-
 	.globl	_netisr
 _netisr:	.long	0		/* set with bits for which queue to service */
-
 vec:
 	.long	vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7
 	.long	vec8, vec9, vec10, vec11, vec12, vec13, vec14, vec15
 
-#define	GENSPL(name, mask, event) \
-	.globl	_spl/**/name ; \
-	ALIGN_TEXT ; \
-_spl/**/name: ; \
-	COUNT_EVENT(_intrcnt_spl, event) ; \
-	movl	_cpl,%eax ; \
-	movl	%eax,%edx ; \
-	orl	mask,%edx ; \
-	movl	%edx,_cpl ; \
-	SHOW_CPL ; \
-	ret
-
-#define	FASTSPL(mask) \
-	movl	mask,_cpl ; \
-	SHOW_CPL
-
-#define	FASTSPL_VARMASK(varmask) \
-	movl	varmask,%eax ; \
-	movl	%eax,_cpl ; \
-	SHOW_CPL
-
 	.text
 
-	ALIGN_TEXT
-unpend_v:
-	COUNT_EVENT(_intrcnt_spl, 0)
-	bsfl	%eax,%eax		# slow, but not worth optimizing
-	btrl	%eax,_ipending
-	jnc	unpend_v_next		# some intr cleared the in-memory bit
-	SHOW_IPENDING
-	movl	Vresume(,%eax,4),%eax
-	testl	%eax,%eax
-	je	noresume
-	jmp	%eax
-
-	ALIGN_TEXT
-/*
- * XXX - must be some fastintr, need to register those too.
- */
-noresume:
-#if NSIO > 0
-	call	_softsio1
-#endif
-unpend_v_next:
-	movl	_cpl,%eax
-	movl	%eax,%edx
-	notl	%eax
-	andl	_ipending,%eax
-	je	none_to_unpend
-	jmp	unpend_v
-
 /*
- * Handle return from interrupt after device handler finishes
- */
-	ALIGN_TEXT
-doreti:
-	COUNT_EVENT(_intrcnt_spl, 1)
-	addl	$4,%esp			# discard unit arg
-	popl	%eax			# get previous priority
-/*
- * Now interrupt frame is a trap frame!
- *
- * XXX - setting up the interrupt frame to be almost a stack frame is mostly
- * a waste of time.
+ * Handle return from interrupts, traps and syscalls.
  */
+	SUPERALIGN_TEXT
+_doreti:
+	FAKE_MCOUNT(_bintr)		/* init "from" _bintr -> _doreti */
+	addl	$4,%esp			/* discard unit number */
+	popl	%eax			/* cpl to restore */
+doreti_next:
+	/*
+	 * Check for pending HWIs and SWIs atomically with restoring cpl
+	 * and exiting.  The check has to be atomic with exiting to stop
+	 * (ipending & ~cpl) changing from zero to nonzero while we're
+	 * looking at it (this wouldn't be fatal but it would increase
+	 * interrupt latency).  Restoring cpl has to be atomic with exiting
+	 * so that the stack cannot pile up (the nesting level of interrupt
+	 * handlers is limited by the number of bits in cpl).
+	 */
+	movl	%eax,%ecx
+	notl	%ecx
+	cli
+	andl	_ipending,%ecx
+	jne	doreti_unpend
+doreti_exit:
 	movl	%eax,_cpl
-	SHOW_CPL
-	movl	%eax,%edx
-	notl	%eax
-	andl	_ipending,%eax
-	jne	unpend_v
-none_to_unpend:
-	testl	%edx,%edx		# returning to zero priority?
-	jne	1f			# nope, going to non-zero priority
-	movl	_netisr,%eax
-	testl	%eax,%eax		# check for softint s/traps
-	jne	2f			# there are some
-	jmp	test_resched		# XXX - schedule jumps better
-	COUNT_EVENT(_intrcnt_spl, 2)	# XXX
-
-	ALIGN_TEXT			# XXX
-1:					# XXX
-	COUNT_EVENT(_intrcnt_spl, 3)
+	MEXITCOUNT
 	popl	%es
 	popl	%ds
 	popal
 	addl	$8,%esp
 	iret
 
-#include "../net/netisr.h"
-
-#define DONET(s, c, event) ; \
-	.globl	c ; \
-	btrl	$s,_netisr ; \
-	jnc	1f ; \
-	COUNT_EVENT(_intrcnt_spl, event) ; \
-	call	c ; \
-1:
-
 	ALIGN_TEXT
-2:
-	COUNT_EVENT(_intrcnt_spl, 4)
-/*
- * XXX - might need extra locking while testing reg copy of netisr, but
- * interrupt routines setting it would not cause any new problems (since we
- * don't loop, fresh bits will not be processed until the next doreti or spl0).
- */
-	testl	$~((1 << NETISR_SCLK) | (1 << NETISR_AST)),%eax
-	je	test_ASTs		# no net stuff, just temporary AST's
-	FASTSPL_VARMASK(_netmask)
-#if 0	
-	DONET(NETISR_RAW, _rawintr, 5)
-#endif
-
-#ifdef	INET
-	DONET(NETISR_IP, _ipintr, 6)
-#endif	/* INET */
-
-#ifdef	IMP
-	DONET(NETISR_IMP, _impintr, 7)
-#endif	/* IMP */
-
-#ifdef	NS
-	DONET(NETISR_NS, _nsintr, 8)
-#endif	/* NS */
-
-#ifdef	ISO
-	DONET(NETISR_ISO, _clnlintr, 9)
-#endif	/* ISO */
+doreti_unpend:
+	/*
+	 * Enabling interrupts is safe because we haven't restored cpl yet.
+	 * The locking from the "btrl" test is probably no longer necessary.
+	 * We won't miss any new pending interrupts because we will check
+	 * for them again.
+	 */
+	sti
+	bsfl	%ecx,%ecx		/* slow, but not worth optimizing */
+	btrl	%ecx,_ipending
+	jnc	doreti_next		/* some intr cleared memory copy */
+	movl	ihandlers(,%ecx,4),%edx
+	testl	%edx,%edx
+	je	doreti_next		/* "can't happen" */
+	cmpl	$NHWI,%ecx
+	jae	doreti_swi
+	cli
+	movl	%eax,_cpl
+	MEXITCOUNT
+	jmp	%edx
 
-#ifdef	CCITT
-	DONET(NETISR_X25, _pkintr, 29)
-	DONET(NETISR_HDLC, _hdintr, 30)
-#endif	/* CCITT */
+	ALIGN_TEXT
+doreti_swi:
+	pushl	%eax
+	/*
+	 * The SWI_AST handler has to run at cpl = SWI_AST_MASK and the
+	 * SWI_CLOCK handler at cpl = SWI_CLOCK_MASK, so we have to restore
+	 * all the h/w bits in cpl now and have to worry about stack growth.
+	 * The worst case is currently (30 Jan 1994) 2 SWI handlers nested
+	 * in dying interrupt frames and about 12 HWIs nested in active
+	 * interrupt frames.  There are only 4 different SWIs and the HWI
+	 * and SWI masks limit the nesting further.
+	 */
+	orl	imasks(,%ecx,4),%eax
+	movl	%eax,_cpl
+	call	%edx
+	popl	%eax
+	jmp	doreti_next
 
-	FASTSPL($0)
-test_ASTs:
-	btrl	$NETISR_SCLK,_netisr
-	jnc	test_resched
-	COUNT_EVENT(_intrcnt_spl, 10)
-	FASTSPL($SOFTCLOCKMASK)
-/*
- * Back to an interrupt frame for a moment.
- */
-	pushl	$0			# previous cpl (probably not used)
-	pushl	$0x7f			# dummy unit number
-	call	_softclock
-	addl	$8,%esp			# discard dummies
-	FASTSPL($0)
-test_resched:
-#ifdef notused1
-	btrl	$NETISR_AST,_netisr
-	jnc	2f
-#endif
-#ifdef notused2
-	cmpl	$0,_want_resched
-	je	2f
-#endif
-	cmpl	$0,_astpending		# XXX - put it back in netisr to
-	je	2f			# reduce the number of tests
+	ALIGN_TEXT
+swi_ast:
+	addl	$8,%esp			/* discard raddr & cpl to get trap frame */
 	testb	$SEL_RPL_MASK,TRAPF_CS_OFF(%esp)
-					# to non-kernel (i.e., user)?
-	je	2f			# nope, leave
-	COUNT_EVENT(_intrcnt_spl, 11)
-	movl	$0,_astpending
+	je	swi_ast_phantom
+	movl	$T_ASTFLT,(2+8+0)*4(%esp)
 	call	_trap
-2:
-	COUNT_EVENT(_intrcnt_spl, 12)
-	popl	%es
-	popl	%ds
-	popal
-	addl	$8,%esp
-	iret
+	subl	%eax,%eax		/* recover cpl */
+	jmp	doreti_next
+
+	ALIGN_TEXT
+swi_ast_phantom:
+	/*
+	 * These happen when there is an interrupt in a trap handler before
+	 * ASTs can be masked or in an lcall handler before they can be
+	 * masked or after they are unmasked.  They could be avoided for
+	 * trap entries by using interrupt gates, and for lcall exits by
+	 * using by using cli, but they are unavoidable for lcall entries.
+	 */
+	cli
+	orl	$SWI_AST_PENDING,_ipending
+	jmp	doreti_exit	/* SWI_AST is highest so we must be done */
 
 /*
  * Interrupt priority mechanism
@@ -262,121 +178,84 @@ test_resched:
  *	-- ipending = active interrupts currently masked by cpl
  */
 
-	GENSPL(bio, _biomask, 13)
-	GENSPL(clock, $HIGHMASK, 14)	/* splclock == splhigh ex for count */
-	GENSPL(high, $HIGHMASK, 15)
-	GENSPL(imp, _netmask, 16)	/* splimp == splnet except for count */
-	GENSPL(net, _netmask, 17)
-	GENSPL(softclock, $SOFTCLOCKMASK, 18)
-	GENSPL(tty, _ttymask, 19)
-
-	.globl _splnone
-	.globl _spl0
-	ALIGN_TEXT
-_splnone:
-_spl0:
-	COUNT_EVENT(_intrcnt_spl, 20)
-in_spl0:
+ENTRY(splz)
+	/*
+	 * The caller has restored cpl and checked that (ipending & ~cpl)
+	 * is nonzero.  We have to repeat the check since if there is an
+	 * interrupt while we're looking, _doreti processing for the
+	 * interrupt will handle all the unmasked pending interrupts
+	 * because we restored early.  We're repeating the calculation
+	 * of (ipending & ~cpl) anyway so that the caller doesn't have
+	 * to pass it, so this only costs one "jne".  "bsfl %ecx,%ecx"
+	 * is undefined when %ecx is 0 so we can't rely on the secondary
+	 * btrl tests.
+	 */
 	movl	_cpl,%eax
-	pushl	%eax			# save old priority
-	testl	$(1 << NETISR_RAW) | (1 << NETISR_IP),_netisr
-	je	over_net_stuff_for_spl0
-	movl	_netmask,%eax		# mask off those network devices
-	movl	%eax,_cpl		# set new priority
-	SHOW_CPL
-/*
- * XXX - what about other net intrs?
- */
-#if 0
-	DONET(NETISR_RAW, _rawintr, 21)
-#endif
-
-#ifdef	INET
-	DONET(NETISR_IP, _ipintr, 22)
-#endif	/* INET */
-
-#ifdef	IMP
-	DONET(NETISR_IMP, _impintr, 23)
-#endif	/* IMP */
-
-#ifdef	NS
-	DONET(NETISR_NS, _nsintr, 24)
-#endif	/* NS */
-
-#ifdef	ISO
-	DONET(NETISR_ISO, _clnlintr, 25)
-#endif	/* ISO */
-
-over_net_stuff_for_spl0:
-	movl	$0,_cpl			# set new priority
-	SHOW_CPL
-	movl	_ipending,%eax
-	testl	%eax,%eax
-	jne	unpend_V
-	popl	%eax			# return old priority
+splz_next:
+	/*
+	 * We don't need any locking here.  (ipending & ~cpl) cannot grow 
+	 * while we're looking at it - any interrupt will shrink it to 0.
+	 */
+	movl	%eax,%ecx
+	notl	%ecx
+	andl	_ipending,%ecx
+	jne	splz_unpend
 	ret
 
-	.globl _splx
 	ALIGN_TEXT
-_splx:
-	COUNT_EVENT(_intrcnt_spl, 26)
-	movl	4(%esp),%eax		# new priority
-	testl	%eax,%eax
-	je	in_spl0			# going to "zero level" is special
-	COUNT_EVENT(_intrcnt_spl, 27)
-	movl	_cpl,%edx		# save old priority
-	movl	%eax,_cpl		# set new priority
-	SHOW_CPL
-	notl	%eax
-	andl	_ipending,%eax
-	jne	unpend_V_result_edx
-	movl	%edx,%eax		# return old priority
-	ret
-
-	ALIGN_TEXT
-unpend_V_result_edx:
-	pushl	%edx
-unpend_V:
-	COUNT_EVENT(_intrcnt_spl, 28)
-	bsfl	%eax,%eax
-	btrl	%eax,_ipending
-	jnc	unpend_V_next
-	SHOW_IPENDING
-	movl	Vresume(,%eax,4),%edx
+splz_unpend:
+	bsfl	%ecx,%ecx
+	btrl	%ecx,_ipending
+	jnc	splz_next
+	movl	ihandlers(,%ecx,4),%edx
 	testl	%edx,%edx
-	je	noresumeV
-/*
- * We would prefer to call the intr handler directly here but that doesn't
- * work for badly behaved handlers that want the interrupt frame.  Also,
- * there's a problem determining the unit number.  We should change the
- * interface so that the unit number is not determined at config time.
- */
-	jmp	*vec(,%eax,4)
+	je	splz_next		/* "can't happen" */
+	cmpl	$NHWI,%ecx
+	jae	splz_swi
+	/*
+	 * We would prefer to call the intr handler directly here but that
+	 * doesn't work for badly behaved handlers that want the interrupt
+	 * frame.  Also, there's a problem determining the unit number.
+	 * We should change the interface so that the unit number is not
+	 * determined at config time.
+	 */
+	jmp	*vec(,%ecx,4)
 
 	ALIGN_TEXT
+splz_swi:
+	cmpl	$SWI_AST,%ecx
+	je	splz_next		/* "can't happen" */
+	pushl	%eax
+	orl	imasks(,%ecx,4),%eax
+	movl	%eax,_cpl
+	call	%edx
+	popl	%eax
+	movl	%eax,_cpl
+	jmp	splz_next
+
 /*
- * XXX - must be some fastintr, need to register those too.
+ * Fake clock IRQ so that it appears to come from our caller and not from
+ * vec0, so that kernel profiling works.
+ * XXX do this more generally (for all vectors; look up the C entry point).
+ * XXX frame bogusness stops us from just jumping to the C entry point.
  */
-noresumeV:
-#if NSIO > 0
-	call	_softsio1
-#endif
-unpend_V_next:
-	movl	_cpl,%eax
-	notl	%eax
-	andl	_ipending,%eax
-	jne	unpend_V
-	popl	%eax
-	ret
+	ALIGN_TEXT
+vec0:
+	popl	%eax			/* return address */
+	pushfl
+#define	KCSEL	8
+	pushl	$KCSEL
+	pushl	%eax
+	cli
+	MEXITCOUNT
+	jmp	_Vclk
 
 #define BUILD_VEC(irq_num) \
 	ALIGN_TEXT ; \
 vec/**/irq_num: ; \
 	int	$ICU_OFFSET + (irq_num) ; \
-	popl	%eax ; \
 	ret
 
-	BUILD_VEC(0)
 	BUILD_VEC(1)
 	BUILD_VEC(2)
 	BUILD_VEC(3)
@@ -392,3 +271,58 @@ vec/**/irq_num: ; \
 	BUILD_VEC(13)
 	BUILD_VEC(14)
 	BUILD_VEC(15)
+
+	ALIGN_TEXT
+swi_clock:
+	MCOUNT
+	subl	%eax,%eax
+	cmpl	$_splz,(%esp)		/* XXX call from splz()? */
+	jae	1f			/* yes, usermode = 0 */
+	movl	4+4+TRAPF_CS_OFF(%esp),%eax	/* no, check trap frame */
+	andl	$SEL_RPL_MASK,%eax
+1:
+	pushl	%eax
+	call	_softclock
+	addl	$4,%esp
+	ret
+
+#define DONET(s, c, event) ; \
+	.globl	c ; \
+	btrl	$s,_netisr ; \
+	jnc	9f ; \
+	call	c ; \
+9:
+
+	ALIGN_TEXT
+swi_net:
+	MCOUNT
+#if 0	
+	DONET(NETISR_RAW, _rawintr,netisr_raw)
+#endif
+#ifdef INET
+	DONET(NETISR_IP, _ipintr,netisr_ip)
+#endif
+#ifdef IMP
+	DONET(NETISR_IMP, _impintr,netisr_imp)
+#endif
+#ifdef NS
+	DONET(NETISR_NS, _nsintr,netisr_ns)
+#endif
+#ifdef ISO
+	DONET(NETISR_ISO, _clnlintr,netisr_iso)
+#endif
+#ifdef	CCITT
+	DONET(NETISR_X25, _pkintr, 29)
+	DONET(NETISR_HDLC, _hdintr, 30)
+#endif
+	ret
+
+	ALIGN_TEXT
+swi_tty:
+	MCOUNT
+#include "sio.h"
+#if NSIO > 0
+	jmp	_siopoll
+#else
+	ret
+#endif
diff --git a/sys/i386/isa/if_is.c b/sys/i386/isa/if_is.c
index 341885f..9f0e6ad 100644
--- a/sys/i386/isa/if_is.c
+++ b/sys/i386/isa/if_is.c
@@ -483,7 +483,7 @@ is_init(unit)
 	/* Address not known */
  	if (ifp->if_addrlist == (struct ifaddr *)0) return;
 
-	s = splnet();
+	s = splimp();
 
 	/* 
 	 * Lance must be stopped
@@ -984,7 +984,7 @@ is_ioctl(ifp, cmd, data)
 	struct ifreq *ifr = (struct ifreq *)data;
 	int s, error = 0;
 
-	s = splnet();
+	s = splimp();
 
 	switch (cmd) {
 
diff --git a/sys/i386/isa/isa.c b/sys/i386/isa/isa.c
index 972c2f7..835aa98 100644
--- a/sys/i386/isa/isa.c
+++ b/sys/i386/isa/isa.c
@@ -34,7 +34,7 @@
  * SUCH DAMAGE.
  *
  *	from: @(#)isa.c	7.2 (Berkeley) 5/13/91
- *	$Id: isa.c,v 1.13 1994/01/17 05:49:20 rgrimes Exp $
+ *	$Id: isa.c,v 1.14 1994/01/22 21:52:04 rgrimes Exp $
  */
 
 /*
@@ -213,38 +213,45 @@ isa_configure() {
 	printf("Probing for devices on the ISA bus:\n");
 	for (dvp = isa_devtab_tty; dvp->id_driver; dvp++) {
 		if (!haveseen_isadev(dvp))
-			config_isadev(dvp,&ttymask);
+			config_isadev(dvp,&tty_imask);
 	}
 	for (dvp = isa_devtab_bio; dvp->id_driver; dvp++) {
 		if (!haveseen_isadev(dvp))
-			config_isadev(dvp,&biomask);
+			config_isadev(dvp,&bio_imask);
 	}
 	for (dvp = isa_devtab_net; dvp->id_driver; dvp++) {
 		if (!haveseen_isadev(dvp))
-			config_isadev(dvp,&netmask);
+			config_isadev(dvp,&net_imask);
 	}
 	for (dvp = isa_devtab_null; dvp->id_driver; dvp++) {
 		if (!haveseen_isadev(dvp))
 			config_isadev(dvp,(u_int *) NULL);
 	}
+	bio_imask |= SWI_CLOCK_MASK;
+	net_imask |= SWI_NET_MASK;
+	tty_imask |= SWI_TTY_MASK;
+
 /*
- * XXX We should really add the tty device to netmask when the line is
+ * XXX we should really add the tty device to net_imask when the line is
  * switched to SLIPDISC, and then remove it when it is switched away from
- * SLIPDISC.  No need to block out ALL ttys during a splnet when only one
+ * SLIPDISC.  No need to block out ALL ttys during a splimp when only one
  * of them is running slip.
+ *
+ * XXX actually, blocking all ttys during a splimp doesn't matter so much 
+ * with sio because the serial interrupt layer doesn't use tty_imask.  Only
+ * non-serial ttys suffer.  It's more stupid that ALL 'net's are blocked
+ * during spltty.
  */
 #include "sl.h"
 #if NSL > 0
-	netmask |= ttymask;
-	ttymask |= netmask;
+	net_imask |= tty_imask;
+	tty_imask = net_imask;
+#endif
+	/* bio_imask |= tty_imask ;  can some tty devices use buffers? */
+#ifdef DIAGNOSTIC
+	printf("bio_imask %x tty_imask %x net_imask %x\n",
+	       bio_imask, tty_imask, net_imask);
 #endif
-	/* if netmask == 0, then the loopback code can do some really
-	 * bad things.
-	 */
-	if (netmask == 0)
-		netmask = 0x10000;
-	/* biomask |= ttymask ;  can some tty devices use buffers? */
-	printf("biomask %x ttymask %x netmask %x\n", biomask, ttymask, netmask);
 	splnone();
 }
 
@@ -337,15 +344,12 @@ extern inthand_t
 	IDTVEC(intr8), IDTVEC(intr9), IDTVEC(intr10), IDTVEC(intr11),
 	IDTVEC(intr12), IDTVEC(intr13), IDTVEC(intr14), IDTVEC(intr15);
 
-static inthand_func_t defvec[16] = {
+static inthand_func_t defvec[ICU_LEN] = {
 	&IDTVEC(intr0), &IDTVEC(intr1), &IDTVEC(intr2), &IDTVEC(intr3),
 	&IDTVEC(intr4), &IDTVEC(intr5), &IDTVEC(intr6), &IDTVEC(intr7),
 	&IDTVEC(intr8), &IDTVEC(intr9), &IDTVEC(intr10), &IDTVEC(intr11),
 	&IDTVEC(intr12), &IDTVEC(intr13), &IDTVEC(intr14), &IDTVEC(intr15) };
 
-/* out of range default interrupt vector gate entry */
-extern inthand_t IDTVEC(intrdefault);
-
 /*
  * Fill in default interrupt table (in case of spuruious interrupt
  * during configuration of kernel, setup interrupt control unit
@@ -356,12 +360,8 @@ isa_defaultirq()
 	int i;
 
 	/* icu vectors */
-	for (i = NRSVIDT ; i < NRSVIDT+ICU_LEN ; i++)
-		setidt(i, defvec[i],  SDT_SYS386IGT, SEL_KPL);
-  
-	/* out of range vectors */
-	for (i = NRSVIDT; i < NIDT; i++)
-		setidt(i, &IDTVEC(intrdefault), SDT_SYS386IGT, SEL_KPL);
+	for (i = 0; i < ICU_LEN; i++)
+		setidt(ICU_OFFSET + i, defvec[i], SDT_SYS386IGT, SEL_KPL);
 
 	/* initialize 8259's */
 	outb(IO_ICU1, 0x11);		/* reset; program device, four bytes */
diff --git a/sys/i386/isa/npx.c b/sys/i386/isa/npx.c
index 796dfbb..00424bf 100644
--- a/sys/i386/isa/npx.c
+++ b/sys/i386/isa/npx.c
@@ -32,7 +32,7 @@
  * SUCH DAMAGE.
  *
  *	from: @(#)npx.c	7.2 (Berkeley) 5/12/91
- *	$Id: npx.c,v 1.5 1993/11/03 23:32:35 paul Exp $
+ *	$Id: npx.c,v 1.6 1994/01/03 07:55:43 davidg Exp $
  */
 
 #include "npx.h"
@@ -114,7 +114,7 @@ struct	isa_driver npxdriver = {
 	npxprobe, npxattach, "npx",
 };
 
-u_int	npx0mask;
+u_int	npx0_imask;
 struct proc	*npxproc;
 
 static	bool_t			npx_ex16;
@@ -292,7 +292,7 @@ npxprobe1(dvp)
 				 * Bad, we are stuck with IRQ13.
 				 */
 				npx_irq13 = 1;
-				npx0mask = dvp->id_irq;	/* npxattach too late */
+				npx0_imask = dvp->id_irq;	/* npxattach too late */
 				return (IO_NPXSIZE);
 			}
 			/*
@@ -528,8 +528,8 @@ npxsave(addr)
 	old_icu1_mask = inb(IO_ICU1 + 1);
 	old_icu2_mask = inb(IO_ICU2 + 1);
 	save_idt_npxintr = idt[npx_intrno];
-	outb(IO_ICU1 + 1, old_icu1_mask & ~(IRQ_SLAVE | npx0mask));
-	outb(IO_ICU2 + 1, old_icu2_mask & ~(npx0mask >> 8));
+	outb(IO_ICU1 + 1, old_icu1_mask & ~(IRQ_SLAVE | npx0_imask));
+	outb(IO_ICU2 + 1, old_icu2_mask & ~(npx0_imask >> 8));
 	idt[npx_intrno] = npx_idt_probeintr;
 	enable_intr();
 	stop_emulating();
@@ -541,10 +541,10 @@ npxsave(addr)
 	icu1_mask = inb(IO_ICU1 + 1);	/* masks may have changed */
 	icu2_mask = inb(IO_ICU2 + 1);
 	outb(IO_ICU1 + 1,
-	     (icu1_mask & ~npx0mask) | (old_icu1_mask & npx0mask));
+	     (icu1_mask & ~npx0_imask) | (old_icu1_mask & npx0_imask));
 	outb(IO_ICU2 + 1,
-	     (icu2_mask & ~(npx0mask >> 8))
-	     | (old_icu2_mask & (npx0mask >> 8)));
+	     (icu2_mask & ~(npx0_imask >> 8))
+	     | (old_icu2_mask & (npx0_imask >> 8)));
 	idt[npx_intrno] = save_idt_npxintr;
 	enable_intr();		/* back to usual state */
 }
diff --git a/sys/i386/isa/sio.c b/sys/i386/isa/sio.c
index 305256d..603c963 100644
--- a/sys/i386/isa/sio.c
+++ b/sys/i386/isa/sio.c
@@ -31,7 +31,7 @@
  * SUCH DAMAGE.
  *
  *	from: @(#)com.c	7.5 (Berkeley) 5/16/91
- *	$Id: sio.c,v 1.40 1994/03/26 13:40:18 ache Exp $
+ *	$Id: sio.c,v 1.41 1994/04/01 16:47:01 ache Exp $
  */
 
 #include "sio.h"
@@ -98,7 +98,6 @@
 #endif
 
 #define	com_scr		7	/* scratch register for 16450-16550 (R/W) */
-#define setsofttty()    (ipending |= 1 << 4)     /* XXX */
 
 /*
  * Input buffer watermarks.
diff --git a/sys/i386/isa/sound/soundcard.c b/sys/i386/isa/sound/soundcard.c
index dae5a54..c2b39d9 100644
--- a/sys/i386/isa/sound/soundcard.c
+++ b/sys/i386/isa/sound/soundcard.c
@@ -34,15 +34,15 @@
 
 #include "dev_table.h"
 
-u_int	snd1mask;
-u_int	snd2mask;
-u_int	snd3mask;
-u_int	snd4mask;
-u_int	snd5mask;
-u_int	snd6mask;
-u_int	snd7mask;
-u_int	snd8mask;
-u_int	snd9mask;
+u_int	snd1_imask;
+u_int	snd2_imask;
+u_int	snd3_imask;
+u_int	snd4_imask;
+u_int	snd5_imask;
+u_int	snd6_imask;
+u_int	snd7_imask;
+u_int	snd8_imask;
+u_int	snd9_imask;
 
 #define FIX_RETURN(ret) {if ((ret)<0) return -(ret); else return 0;}
 
diff --git a/sys/i386/isa/vector.s b/sys/i386/isa/vector.s
index d434c76..7135ae7 100644
--- a/sys/i386/isa/vector.s
+++ b/sys/i386/isa/vector.s
@@ -1,6 +1,6 @@
 /*
  *	from: vector.s, 386BSD 0.1 unknown origin
- *	$Id: vector.s,v 1.5 1993/12/20 15:08:33 wollman Exp $
+ *	$Id: vector.s,v 1.6 1994/01/10 23:15:09 ache Exp $
  */
 
 #include "i386/isa/icu.h"
@@ -12,24 +12,44 @@
 #define	IRQ_BIT(irq_num)	(1 << ((irq_num) % 8))
 #define	IRQ_BYTE(irq_num)	((irq_num) / 8)
 
+#ifdef AUTO_EOI_1
+#define	ENABLE_ICU1		/* use auto-EOI to reduce i/o */
+#else
 #define	ENABLE_ICU1 \
 	movb	$ICU_EOI,%al ;	/* as soon as possible send EOI ... */ \
 	FASTER_NOP ;		/* ... ASAP ... */ \
 	outb	%al,$IO_ICU1	/* ... to clear in service bit */
-#ifdef AUTO_EOI_1
-#undef	ENABLE_ICU1		/* we now use auto-EOI to reduce i/o */
-#define	ENABLE_ICU1
 #endif
 
+#ifdef AUTO_EOI_2
+/*
+ * The data sheet says no auto-EOI on slave, but it sometimes works.
+ */
+#define	ENABLE_ICU1_AND_2	ENABLE_ICU1
+#else
 #define	ENABLE_ICU1_AND_2 \
 	movb	$ICU_EOI,%al ;	/* as above */ \
 	FASTER_NOP ; \
 	outb	%al,$IO_ICU2 ;	/* but do second icu first */ \
 	FASTER_NOP ; \
 	outb	%al,$IO_ICU1	/* then first icu */
-#ifdef AUTO_EOI_2
-#undef	ENABLE_ICU1_AND_2	/* data sheet says no auto-EOI on slave ... */
-#define	ENABLE_ICU1_AND_2	/* ... but it works */
+#endif
+
+#ifdef FAST_INTR_HANDLER_USES_ES
+#define	ACTUALLY_PUSHED		1
+#define	MAYBE_MOVW_AX_ES	movl	%ax,%es
+#define	MAYBE_POPL_ES		popl	%es
+#define	MAYBE_PUSHL_ES		pushl	%es
+#else
+/*
+ * We can usually skip loading %es for fastintr handlers.  %es should
+ * only be used for string instructions, and fastintr handlers shouldn't
+ * do anything slow enough to justify using a string instruction.
+ */
+#define	ACTUALLY_PUSHED		0
+#define	MAYBE_MOVW_AX_ES
+#define	MAYBE_POPL_ES
+#define	MAYBE_PUSHL_ES
 #endif
 
 /*
@@ -82,39 +102,63 @@
 	pushl	%ecx ; \
 	pushl	%edx ; \
 	pushl	%ds ; \
-	/* pushl	%es ; know compiler doesn't do string insns */ \
+	MAYBE_PUSHL_ES ; \
 	movl	$KDSEL,%eax ; \
 	movl	%ax,%ds ; \
-	/* movl	%ax,%es ; */ \
-	SHOW_CLI ;		/* although it interferes with "ASAP" */ \
+	MAYBE_MOVW_AX_ES ; \
+	FAKE_MCOUNT((4+ACTUALLY_PUSHED)*4(%esp)) ; \
 	pushl	$unit ; \
 	call	handler ;	/* do the work ASAP */ \
 	enable_icus ;		/* (re)enable ASAP (helps edge trigger?) */ \
 	addl	$4,%esp ; \
 	incl	_cnt+V_INTR ;	/* book-keeping can wait */ \
-	COUNT_EVENT(_intrcnt_actv, id_num) ; \
-	SHOW_STI ; \
-	/* popl	%es ; */ \
+	incl	_intrcnt_actv + (id_num) * 4 ; \
+	movl	_cpl,%eax ;	/* are we unmasking pending HWIs or SWIs? */ \
+	notl	%eax ; \
+	andl	_ipending,%eax ; \
+	jne	1f ;		/* yes, handle them */ \
+	MEXITCOUNT ; \
+	MAYBE_POPL_ES ; \
 	popl	%ds ; \
-	popl	%edx; \
-	popl	%ecx; \
-	popl	%eax; \
-	iret
+	popl	%edx ; \
+	popl	%ecx ; \
+	popl	%eax ; \
+	iret ; \
+; \
+	ALIGN_TEXT ; \
+1: ; \
+	movl	_cpl,%eax ; \
+	movl	$HWI_MASK|SWI_MASK,_cpl ;	/* limit nesting ... */ \
+	sti ;			/* ... to do this as early as possible */ \
+	MAYBE_POPL_ES ;		/* discard most of thin frame ... */ \
+	popl	%ecx ;		/* ... original %ds ... */ \
+	popl	%edx ; \
+	xchgl	%eax,(1+ACTUALLY_PUSHED)*4(%esp) ; /* orig %eax; save cpl */ \
+	pushal ;		/* build fat frame (grrr) ... */ \
+	pushl	%ecx ;		/* ... actually %ds ... */ \
+	pushl	%es ; \
+	movl	$KDSEL,%eax ; \
+	movl	%ax,%es ; \
+	movl	(2+8+0)*4(%esp),%ecx ;	/* ... %ecx from thin frame ... */ \
+	movl	%ecx,(2+6)*4(%esp) ;	/* ... to fat frame ... */ \
+	movl	(2+8+1)*4(%esp),%eax ;	/* ... cpl from thin frame */ \
+	pushl	%eax ; \
+	subl	$4,%esp ;	/* junk for unit number */ \
+	MEXITCOUNT ; \
+	jmp	_doreti
 
 #define	INTR(unit, irq_num, id_num, mask, handler, icu, enable_icus, reg, stray) \
-	pushl	$0 ;		/* dummy error code */ \
-	pushl	$T_ASTFLT ; \
+	pushl	$0 ;		/* dumby error code */ \
+	pushl	$0 ;		/* dumby trap type */ \
 	pushal ; \
-	pushl	%ds ; 		/* save our data and extra segments ... */ \
+	pushl	%ds ;		/* save our data and extra segments ... */ \
 	pushl	%es ; \
 	movl	$KDSEL,%eax ;	/* ... and reload with kernel's own ... */ \
-	movl	%ax,%ds ; 	/* ... early in case SHOW_A_LOT is on */ \
+	movl	%ax,%ds ;	/* ... early for obsolete reasons */ \
 	movl	%ax,%es ; \
-	SHOW_CLI ;		/* interrupt did an implicit cli */ \
 	movb	_imen + IRQ_BYTE(irq_num),%al ; \
 	orb	$IRQ_BIT(irq_num),%al ; \
 	movb	%al,_imen + IRQ_BYTE(irq_num) ; \
-	SHOW_IMEN ; \
 	FASTER_NOP ; \
 	outb	%al,$icu+1 ; \
 	enable_icus ; \
@@ -123,32 +167,32 @@
 	testb	$IRQ_BIT(irq_num),%reg ; \
 	jne	2f ; \
 1: ; \
-	COUNT_EVENT(_intrcnt_actv, id_num) ; \
+	FAKE_MCOUNT(12*4(%esp)) ;	/* XXX late to avoid double count */ \
+	incl	_intrcnt_actv + (id_num) * 4 ; \
 	movl	_cpl,%eax ; \
 	pushl	%eax ; \
 	pushl	$unit ; \
 	orl	mask,%eax ; \
 	movl	%eax,_cpl ; \
-	SHOW_CPL ; \
-	SHOW_STI ; \
 	sti ; \
 	call	handler ; \
 	movb	_imen + IRQ_BYTE(irq_num),%al ; \
 	andb	$~IRQ_BIT(irq_num),%al ; \
 	movb	%al,_imen + IRQ_BYTE(irq_num) ; \
-	SHOW_IMEN ; \
 	FASTER_NOP ; \
 	outb	%al,$icu+1 ; \
-	jmp	doreti ; \
+	MEXITCOUNT ; \
+	/* We could usually avoid the following jmp by inlining some of */ \
+	/* _doreti, but it's probably better to use less cache. */ \
+	jmp	_doreti ; \
 ; \
 	ALIGN_TEXT ; \
 2: ; \
-	COUNT_EVENT(_intrcnt_pend, id_num) ; \
+	/* XXX skip mcounting here to avoid double count */ \
 	movl	$1b,%eax ;	/* register resume address */ \
 				/* XXX - someday do it at attach time */ \
-	movl	%eax,Vresume + (irq_num) * 4 ;	\
+	movl	%eax,ihandlers + (irq_num) * 4 ;	\
 	orb	$IRQ_BIT(irq_num),_ipending + IRQ_BYTE(irq_num) ; \
-	SHOW_IPENDING ; \
 	popl	%es ; \
 	popl	%ds ; \
 	popal ; \
@@ -191,7 +235,7 @@
 	.globl	_V/**/name ; \
 	SUPERALIGN_TEXT ; \
 _V/**/name: ; \
-	FAST_INTR(unit, irq_num, id_num, handler, ENABLE_ICU/**/icu_enables)
+	FAST_INTR(unit, irq_num,id_num, handler, ENABLE_ICU/**/icu_enables)
 
 #undef BUILD_VECTOR
 #define	BUILD_VECTOR(name, unit, irq_num, id_num, mask, handler, \
@@ -201,9 +245,10 @@ _V/**/name: ; \
 	.globl	_V/**/name ; \
 	SUPERALIGN_TEXT ; \
 _V/**/name: ; \
-	INTR(unit,irq_num,id_num, mask, handler, IO_ICU/**/icu_num, \
+	INTR(unit,irq_num, id_num, mask, handler, IO_ICU/**/icu_num, \
 	     ENABLE_ICU/**/icu_enables, reg,)
 
+MCOUNT_LABEL(bintr)
 	BUILD_VECTORS
 
 	/* hardware interrupt catcher (IDT 32 - 47) */
@@ -211,7 +256,7 @@ _V/**/name: ; \
 
 #define	STRAYINTR(irq_num, icu_num, icu_enables, reg) \
 IDTVEC(intr/**/irq_num) ; \
-	INTR(irq_num,irq_num,irq_num, _highmask,  _isa_strayintr, \
+	INTR(irq_num,irq_num,irq_num, _high_imask,  _isa_strayintr, \
 		  IO_ICU/**/icu_num, ENABLE_ICU/**/icu_enables, reg,stray)
 
 /*
@@ -241,6 +286,7 @@ IDTVEC(intr/**/irq_num) ; \
 	STRAYINTR(4,1,1, al)
 	STRAYINTR(5,1,1, al)
 	STRAYINTR(6,1,1, al)
+	STRAYINTR(7,1,1, al)
 	STRAYINTR(8,2,1_AND_2, ah)
 	STRAYINTR(9,2,1_AND_2, ah)
 	STRAYINTR(10,2,1_AND_2, ah)
@@ -249,11 +295,11 @@ IDTVEC(intr/**/irq_num) ; \
 	STRAYINTR(13,2,1_AND_2, ah)
 	STRAYINTR(14,2,1_AND_2, ah)
 	STRAYINTR(15,2,1_AND_2, ah)
-IDTVEC(intrdefault)
-	STRAYINTR(7,1,1, al)	/* XXX */
 #if 0
 	INTRSTRAY(255, _highmask, 255) ; call	_isa_strayintr ; INTREXIT2
 #endif
+MCOUNT_LABEL(eintr)
+
 /*
  * These are the interrupt counters, I moved them here from icu.s so that
  * they are with the name table.  rgrimes
@@ -263,7 +309,15 @@ IDTVEC(intrdefault)
  * work with vmstat.
  */
 	.data
-Vresume:        .space  32 * 4  /* where to resume intr handler after unpend */
+ihandlers:			/* addresses of interrupt handlers */
+	.space	NHWI*4		/* actually resumption addresses for HWI's */
+	.long	swi_tty, swi_net, 0, 0, 0, 0, 0, 0
+	.long	0, 0, 0, 0, 0, 0, swi_clock, swi_ast
+imasks:				/* masks for interrupt handlers */
+	.space	NHWI*4		/* padding; HWI masks are elsewhere */
+	.long	SWI_TTY_MASK, SWI_NET_MASK, 0, 0, 0, 0, 0, 0
+	.long	0, 0, 0, 0, 0, 0, SWI_CLOCK_MASK, SWI_AST_MASK
+
 	.globl	_intrcnt
 _intrcnt:			/* used by vmstat to calc size of table */
 	.globl	_intrcnt_bad7
@@ -274,14 +328,8 @@ _intrcnt_bad15:	.space	4	/* glitches on irq 15 */
 _intrcnt_stray:	.space	4	/* total count of stray interrupts */
 	.globl	_intrcnt_actv
 _intrcnt_actv:	.space	NR_REAL_INT_HANDLERS * 4	/* active interrupts */
-	.globl	_intrcnt_pend
-_intrcnt_pend:	.space	NR_REAL_INT_HANDLERS * 4	/* pending interrupts */
 	.globl	_eintrcnt
 _eintrcnt:			/* used by vmstat to calc size of table */
-	.globl	_intrcnt_spl
-_intrcnt_spl:	.space	32 * 4	/* XXX 32 should not be hard coded ? */
-	.globl	_intrcnt_show
-_intrcnt_show:	.space	8 * 4	/* XXX 16 should not be hard coded ? */
 
 /*
  * Build the interrupt name table for vmstat
@@ -296,8 +344,9 @@ _intrcnt_show:	.space	8 * 4	/* XXX 16 should not be hard coded ? */
 	.ascii	"name irq" ; \
 	.asciz	"irq_num"
 /*
- * XXX - use the STRING and CONCAT macros from <sys/cdefs.h> to stringize
- * and concatenate names above and elsewhere.
+ * XXX - use the __STRING and __CONCAT macros from <sys/cdefs.h> to stringize
+ * and concatenate names above and elsewhere.  Note that __CONCAT doesn't
+ * work when nested.
  */
 
 	.text
@@ -308,61 +357,4 @@ _intrnames:
 	BUILD_VECTOR(stray,,,,,,,,)
 	BUILD_VECTORS
 
-#undef BUILD_FAST_VECTOR
-#define BUILD_FAST_VECTOR	BUILD_VECTOR
-
-#undef BUILD_VECTOR
-#define	BUILD_VECTOR(name, unit, irq_num, id_num, mask, handler, \
-		     icu_num, icu_enables, reg) \
-	.asciz	"name pend"
-
-	BUILD_VECTORS
 _eintrnames:
-
-/*
- * now the spl names
- */
-	.asciz	"unpend_v"
-	.asciz	"doreti"
-	.asciz	"p0!ni"
-	.asciz	"!p0!ni"
-	.asciz	"p0ni"
-	.asciz	"netisr_raw"
-	.asciz	"netisr_ip"
-	.asciz	"netisr_imp"
-	.asciz	"netisr_ns"
-	.asciz	"netisr_iso"
-	.asciz	"softclock"		/* 10 */
-	.asciz	"trap"
-	.asciz	"doreti_exit2"
-	.asciz	"splbio"
-	.asciz	"splclock"
-	.asciz	"splhigh"
-	.asciz	"splimp"
-	.asciz	"splnet"
-	.asciz	"splsoftclock"
-	.asciz	"spltty"
-	.asciz	"spl0"			/* 20 */
-	.asciz	"netisr_raw2"
-	.asciz	"netisr_ip2"
-	.asciz	"netisr_imp2"
-	.asciz	"netisr_ns2"
-	.asciz	"netisr_iso2"
-	.asciz	"splx"
-	.asciz	"splx!0"
-	.asciz	"unpend_V"
-	.asciz	"netisr_x25"
-	.asciz	"netisr_hdlc"
-	.asciz	"spl31"
-/*
- * now the mask names
- */
-	.asciz	"cli"
-	.asciz	"cpl"
-	.asciz	"imen"
-	.asciz	"ipending"
-	.asciz	"sti"
-	.asciz	"mask5"		/* mask5-mask7 are spares */
-	.asciz	"mask6"
-	.asciz	"mask7"
-