Use PCPU_GET, PCPU_PTR and PCPU_SET to access all per-cpu variables

other then curproc.

57 files changed, 477 insertions, 438 deletions

diff --git a/sys/alpha/alpha/interrupt.c b/sys/alpha/alpha/interrupt.c
index 76321a1..6311937 100644
--- a/sys/alpha/alpha/interrupt.c
+++ b/sys/alpha/alpha/interrupt.c
@@ -94,7 +94,7 @@ interrupt(a0, a1, a2, framep)
 	 */
 	globalp = (struct globaldata *) alpha_pal_rdval();
 
-	atomic_add_int(&PCPU_GET(intr_nesting_level), 1);
+	atomic_add_int(PCPU_PTR(intr_nesting_level), 1);
 	{
 		struct proc *p = curproc;
 		if (!p) p = &proc0;
@@ -116,7 +116,7 @@ interrupt(a0, a1, a2, framep)
 		CTR0(KTR_INTR, "clock interrupt");
 		if (PCPU_GET(cpuno) != hwrpb->rpb_primary_cpu_id) {
 			CTR0(KTR_INTR, "ignoring clock on secondary");
-			atomic_subtract_int(&PCPU_GET(intr_nesting_level), 1);
+			atomic_subtract_int(PCPU_PTR(intr_nesting_level), 1);
 			return;
 		}
 			
@@ -152,7 +152,7 @@ interrupt(a0, a1, a2, framep)
 		    a0, a1, a2);
 		/* NOTREACHED */
 	}
-	atomic_subtract_int(&PCPU_GET(intr_nesting_level), 1);
+	atomic_subtract_int(PCPU_PTR(intr_nesting_level), 1);
 }
 
 void
diff --git a/sys/alpha/alpha/machdep.c b/sys/alpha/alpha/machdep.c
index 1d1e26b..1976cd9 100644
--- a/sys/alpha/alpha/machdep.c
+++ b/sys/alpha/alpha/machdep.c
@@ -2068,7 +2068,7 @@ alpha_fpstate_check(struct proc *p)
 	 */
 #ifndef SMP
 	if (p->p_addr->u_pcb.pcb_hw.apcb_flags & ALPHA_PCB_FLAGS_FEN)
-		if (p != fpcurproc)
+		if (p != PCPU_GET(fpcurproc))
 			panic("alpha_check_fpcurproc: bogus");
 #endif
 }
@@ -2089,7 +2089,7 @@ alpha_fpstate_check(struct proc *p)
 void
 alpha_fpstate_save(struct proc *p, int write)
 {
-	if (p == fpcurproc) {
+	if (p == PCPU_GET(fpcurproc)) {
 		/*
 		 * If curproc != fpcurproc, then we need to enable FEN 
 		 * so that we can dump the fp state.
@@ -2107,11 +2107,11 @@ alpha_fpstate_save(struct proc *p, int write)
 			 * PALcode to disable FEN, otherwise we must
 			 * clear the FEN bit in fpcurproc's pcb.
 			 */
-			if (fpcurproc == curproc)
+			if (PCPU_GET(fpcurproc) == curproc)
 				alpha_pal_wrfen(0);
 			else
-				CLEAR_FEN(fpcurproc);
-			fpcurproc = NULL;
+				CLEAR_FEN(PCPU_GET(fpcurproc));
+			PCPU_SET(fpcurproc, NULL);
 		} else {
 			/*
 			 * Make sure that we leave FEN enabled if
@@ -2119,7 +2119,7 @@ alpha_fpstate_save(struct proc *p, int write)
 			 * one process with FEN enabled. Note that FEN 
 			 * must already be set in fpcurproc's pcb.
 			 */
-			if (curproc != fpcurproc)
+			if (curproc != PCPU_GET(fpcurproc))
 				alpha_pal_wrfen(0);
 		}
 	}
@@ -2133,7 +2133,7 @@ alpha_fpstate_save(struct proc *p, int write)
 void
 alpha_fpstate_drop(struct proc *p)
 {
-	if (p == fpcurproc) {
+	if (p == PCPU_GET(fpcurproc)) {
 		if (p == curproc) {
 			/*
 			 * Disable FEN via the PALcode. This will
@@ -2146,7 +2146,7 @@ alpha_fpstate_drop(struct proc *p)
 			 */
 			CLEAR_FEN(p);
 		}
-		fpcurproc = NULL;
+		PCPU_SET(fpcurproc, NULL);
 	}
 }
 
@@ -2161,19 +2161,19 @@ alpha_fpstate_switch(struct proc *p)
 	 * Enable FEN so that we can access the fp registers.
 	 */
 	alpha_pal_wrfen(1);
-	if (fpcurproc) {
+	if (PCPU_GET(fpcurproc)) {
 		/*
 		 * Dump the old fp state if its valid.
 		 */
-		savefpstate(&fpcurproc->p_addr->u_pcb.pcb_fp);
-		CLEAR_FEN(fpcurproc);
+		savefpstate(&PCPU_GET(fpcurproc)->p_addr->u_pcb.pcb_fp);
+		CLEAR_FEN(PCPU_GET(fpcurproc));
 	}
 
 	/*
 	 * Remember the new FP owner and reload its state.
 	 */
-	fpcurproc = p;
-	restorefpstate(&fpcurproc->p_addr->u_pcb.pcb_fp);
+	PCPU_SET(fpcurproc, p);
+	restorefpstate(&PCPU_GET(fpcurproc)->p_addr->u_pcb.pcb_fp);
 
 	/*
 	 * If the new owner is curproc, leave FEN enabled, otherwise
diff --git a/sys/alpha/alpha/mp_machdep.c b/sys/alpha/alpha/mp_machdep.c
index c2e2fcc..2ece633 100644
--- a/sys/alpha/alpha/mp_machdep.c
+++ b/sys/alpha/alpha/mp_machdep.c
@@ -598,7 +598,7 @@ forward_statclock(int pscnt)
 	
 	map = 0;
 	for (id = 0; id < mp_ncpus; id++) {
-		if (id == cpuid)
+		if (id == PCPU_GET(cpuid))
 			continue;
 		if (((1 << id) & checkstate_probed_cpus) == 0)
 			continue;
@@ -676,7 +676,7 @@ forward_hardclock(int pscnt)
 	
 	map = 0;
 	for (id = 0; id < mp_ncpus; id++) {
-		if (id == cpuid)
+		if (id == PCPU_GET(cpuid))
 			continue;
 		if (((1 << id) & checkstate_probed_cpus) == 0)
 			continue;
@@ -1022,7 +1022,7 @@ atomic_readandclear(u_int64_t* p)
 void
 smp_handle_ipi(struct trapframe *frame)
 {
-	u_int64_t ipis = atomic_readandclear(&PCPU_GET(pending_ipis));
+	u_int64_t ipis = atomic_readandclear(PCPU_PTR(pending_ipis));
 	u_int64_t ipi;
 	int cpuno = PCPU_GET(cpuno);
 
diff --git a/sys/alpha/alpha/prom.c b/sys/alpha/alpha/prom.c
index 1958d79..2ddf62d 100644
--- a/sys/alpha/alpha/prom.c
+++ b/sys/alpha/alpha/prom.c
@@ -64,19 +64,21 @@ static pt_entry_t *
 rom_lev1map()
 {
 	struct alpha_pcb *apcb;
+	struct pcb *cpcb;
 
 	/*
 	 * We may be called before the first context switch
 	 * after alpha_init(), in which case we just need
 	 * to use the kernel Lev1map.
 	 */
-	if (curpcb == 0)
+	if (PCPU_GET(curpcb) == 0)
 		return (Lev1map);
 
 	/*
 	 * Find the level 1 map that we're currently running on.
 	 */
-	apcb = (struct alpha_pcb *)ALPHA_PHYS_TO_K0SEG((vm_offset_t) curpcb);
+	cpcb = PCPU_GET(curpcb);
+	apcb = (struct alpha_pcb *)ALPHA_PHYS_TO_K0SEG((vm_offset_t)cpcb);
 
 	return ((pt_entry_t *)ALPHA_PHYS_TO_K0SEG(alpha_ptob(apcb->apcb_ptbr)));
 }
@@ -198,12 +200,12 @@ enter_prom()
 		/*
 		 * SimOS console uses floating point.
 		 */
-		if (curproc != fpcurproc) {
+		if (curproc != PCPU_GET(fpcurproc)) {
 			alpha_pal_wrfen(1);
-			if (fpcurproc)
-				savefpstate(&fpcurproc->p_addr->u_pcb.pcb_fp);
-			fpcurproc = curproc;
-			restorefpstate(&fpcurproc->p_addr->u_pcb.pcb_fp);
+			if (PCPU_GET(fpcurproc))
+				savefpstate(&PCPU_GET(fpcurproc)->p_addr->u_pcb.pcb_fp);
+			PCPU_SET(fpcurproc, curproc);
+			restorefpstate(&PCPU_GET(fpcurproc)->p_addr->u_pcb.pcb_fp);
 		}
 #endif
 		if (!pmap_uses_prom_console())
diff --git a/sys/alpha/alpha/trap.c b/sys/alpha/alpha/trap.c
index 698c09b..78de845 100644
--- a/sys/alpha/alpha/trap.c
+++ b/sys/alpha/alpha/trap.c
@@ -365,7 +365,7 @@ trap(a0, a1, a2, entry, framep)
 			 * on exit from the kernel, if proc == fpcurproc,
 			 * FP is enabled.
 			 */
-			if (fpcurproc == p) {
+			if (PCPU_GET(fpcurproc) == p) {
 				printf("trap: fp disabled for fpcurproc == %p",
 				    p);
 				goto dopanic;
diff --git a/sys/amd64/amd64/bios.c b/sys/amd64/amd64/bios.c
index a22f1549..e90de29 100644
--- a/sys/amd64/amd64/bios.c
+++ b/sys/amd64/amd64/bios.c
@@ -258,7 +258,7 @@ set_bios_selectors(struct bios_segments *seg, int flags)
     union descriptor *p_gdt;
 
 #ifdef SMP
-    p_gdt = &gdt[cpuid * NGDT];
+    p_gdt = &gdt[PCPU_GET(cpuid) * NGDT];
 #else
     p_gdt = gdt;
 #endif
diff --git a/sys/amd64/amd64/db_interface.c b/sys/amd64/amd64/db_interface.c
index fe94264..ee4ab19 100644
--- a/sys/amd64/amd64/db_interface.c
+++ b/sys/amd64/amd64/db_interface.c
@@ -142,11 +142,12 @@ kdb_trap(type, code, regs)
 #ifdef CPUSTOP_ON_DDBBREAK
 
 #if defined(VERBOSE_CPUSTOP_ON_DDBBREAK)
-	db_printf("\nCPU%d stopping CPUs: 0x%08x...", cpuid, other_cpus);
+	db_printf("\nCPU%d stopping CPUs: 0x%08x...", PCPU_GET(cpuid),
+	    PCPU_GET(other_cpus));
 #endif /* VERBOSE_CPUSTOP_ON_DDBBREAK */
 
 	/* We stop all CPUs except ourselves (obviously) */
-	stop_cpus(other_cpus);
+	stop_cpus(PCPU_GET(other_cpus));
 
 #if defined(VERBOSE_CPUSTOP_ON_DDBBREAK)
 	db_printf(" stopped.\n");
@@ -171,13 +172,14 @@ kdb_trap(type, code, regs)
 #ifdef CPUSTOP_ON_DDBBREAK
 
 #if defined(VERBOSE_CPUSTOP_ON_DDBBREAK)
-	db_printf("\nCPU%d restarting CPUs: 0x%08x...", cpuid, stopped_cpus);
+	db_printf("\nCPU%d restarting CPUs: 0x%08x...", PCPU_GET(cpuid),
+	    stopped_cpus);
 #endif /* VERBOSE_CPUSTOP_ON_DDBBREAK */
 
 	/* Restart all the CPUs we previously stopped */
-	if (stopped_cpus != other_cpus && smp_started != 0) {
+	if (stopped_cpus != PCPU_GET(other_cpus) && smp_started != 0) {
 		db_printf("whoa, other_cpus: 0x%08x, stopped_cpus: 0x%08x\n",
-			  other_cpus, stopped_cpus);
+			  PCPU_GET(other_cpus), stopped_cpus);
 		panic("stop_cpus() failed");
 	}
 	restart_cpus(stopped_cpus);
diff --git a/sys/amd64/amd64/fpu.c b/sys/amd64/amd64/fpu.c
index 2672a63..b79d511 100644
--- a/sys/amd64/amd64/fpu.c
+++ b/sys/amd64/amd64/fpu.c
@@ -493,8 +493,8 @@ npxinit(control)
 	npxsave(&dummy);
 	stop_emulating();
 	fldcw(&control);
-	if (curpcb != NULL)
-		fnsave(&curpcb->pcb_savefpu);
+	if (PCPU_GET(curpcb) != NULL)
+		fnsave(&PCPU_GET(curpcb)->pcb_savefpu);
 	start_emulating();
 }
 
@@ -506,14 +506,14 @@ npxexit(p)
 	struct proc *p;
 {
 
-	if (p == npxproc)
-		npxsave(&curpcb->pcb_savefpu);
+	if (p == PCPU_GET(npxproc))
+		npxsave(&PCPU_GET(curpcb)->pcb_savefpu);
 #ifdef NPX_DEBUG
 	if (npx_exists) {
 		u_int	masked_exceptions;
 
-		masked_exceptions = curpcb->pcb_savefpu.sv_env.en_cw
-				    & curpcb->pcb_savefpu.sv_env.en_sw & 0x7f;
+		masked_exceptions = PCPU_GET(curpcb)->pcb_savefpu.sv_env.en_cw
+		    &PCPU_GET(curpcb)->pcb_savefpu.sv_env.en_sw & 0x7f;
 		/*
 		 * Log exceptions that would have trapped with the old
 		 * control word (overflow, divide by 0, and invalid operand).
@@ -722,19 +722,19 @@ npx_intr(dummy)
 	u_short control;
 	struct intrframe *frame;
 
-	if (npxproc == NULL || !npx_exists) {
+	if (PCPU_GET(npxproc) == NULL || !npx_exists) {
 		printf("npxintr: npxproc = %p, curproc = %p, npx_exists = %d\n",
-		       npxproc, curproc, npx_exists);
+		       PCPU_GET(npxproc), curproc, npx_exists);
 		panic("npxintr from nowhere");
 	}
-	if (npxproc != curproc) {
+	if (PCPU_GET(npxproc) != curproc) {
 		printf("npxintr: npxproc = %p, curproc = %p, npx_exists = %d\n",
-		       npxproc, curproc, npx_exists);
+		       PCPU_GET(npxproc), curproc, npx_exists);
 		panic("npxintr from non-current process");
 	}
 
 	outb(0xf0, 0);
-	fnstsw(&curpcb->pcb_savefpu.sv_ex_sw);
+	fnstsw(&PCPU_GET(curpcb)->pcb_savefpu.sv_ex_sw);
 	fnstcw(&control);
 	fnclex();
 
@@ -760,8 +760,8 @@ npx_intr(dummy)
 		 * this exception.
 		 */
 		code = 
-		    fpetable[(curpcb->pcb_savefpu.sv_ex_sw & ~control & 0x3f) |
-			(curpcb->pcb_savefpu.sv_ex_sw & 0x40)];
+		    fpetable[(PCPU_GET(curpcb)->pcb_savefpu.sv_ex_sw & ~control & 0x3f) |
+			(PCPU_GET(curpcb)->pcb_savefpu.sv_ex_sw & 0x40)];
 		trapsignal(curproc, SIGFPE, code);
 	} else {
 		/*
@@ -794,9 +794,9 @@ npxdna()
 {
 	if (!npx_exists)
 		return (0);
-	if (npxproc != NULL) {
+	if (PCPU_GET(npxproc) != NULL) {
 		printf("npxdna: npxproc = %p, curproc = %p\n",
-		       npxproc, curproc);
+		       PCPU_GET(npxproc), curproc);
 		panic("npxdna");
 	}
 	stop_emulating();
@@ -804,7 +804,7 @@ npxdna()
 	 * Record new context early in case frstor causes an IRQ13.
 	 */
 	PCPU_SET(npxproc, CURPROC);
-	curpcb->pcb_savefpu.sv_ex_sw = 0;
+	PCPU_GET(curpcb)->pcb_savefpu.sv_ex_sw = 0;
 	/*
 	 * The following frstor may cause an IRQ13 when the state being
 	 * restored has a pending error.  The error will appear to have been
@@ -817,7 +817,7 @@ npxdna()
 	 * fnsave are broken, so our treatment breaks fnclex if it is the
 	 * first FPU instruction after a context switch.
 	 */
-	frstor(&curpcb->pcb_savefpu);
+	frstor(&PCPU_GET(curpcb)->pcb_savefpu);
 
 	return (1);
 }
diff --git a/sys/amd64/amd64/machdep.c b/sys/amd64/amd64/machdep.c
index c1c6f29..5abdc05 100644
--- a/sys/amd64/amd64/machdep.c
+++ b/sys/amd64/amd64/machdep.c
@@ -1051,7 +1051,7 @@ setregs(p, entry, stack, ps_strings)
 	regs->tf_ebx = ps_strings;
 
 	/* reset %gs as well */
-	if (pcb == curpcb)
+	if (pcb == PCPU_GET(curpcb))
 		load_gs(_udatasel);
 	else
 		pcb->pcb_gs = _udatasel;
@@ -1067,7 +1067,7 @@ setregs(p, entry, stack, ps_strings)
                 pcb->pcb_dr3 = 0;
                 pcb->pcb_dr6 = 0;
                 pcb->pcb_dr7 = 0;
-                if (pcb == curpcb) {
+                if (pcb == PCPU_GET(curpcb)) {
 		        /*
 			 * Clear the debug registers on the running
 			 * CPU, otherwise they will end up affecting
@@ -1970,13 +1970,14 @@ init386(first)
 	initializecpu();	/* Initialize CPU registers */
 
 	/* make an initial tss so cpu can get interrupt stack on syscall! */
-	common_tss.tss_esp0 = (int) proc0.p_addr + UPAGES*PAGE_SIZE - 16;
-	common_tss.tss_ss0 = GSEL(GDATA_SEL, SEL_KPL);
+	PCPU_SET(common_tss.tss_esp0,
+	    (int) proc0.p_addr + UPAGES*PAGE_SIZE - 16);
+	PCPU_SET(common_tss.tss_ss0, GSEL(GDATA_SEL, SEL_KPL));
 	gsel_tss = GSEL(GPROC0_SEL, SEL_KPL);
 	private_tss = 0;
-	tss_gdt = &gdt[GPROC0_SEL].sd;
-	common_tssd = *tss_gdt;
-	common_tss.tss_ioopt = (sizeof common_tss) << 16;
+	PCPU_SET(tss_gdt, &gdt[GPROC0_SEL].sd);
+	PCPU_SET(common_tssd, *PCPU_GET(tss_gdt));
+	PCPU_SET(common_tss.tss_ioopt, (sizeof (struct i386tss)) << 16);
 	ltr(gsel_tss);
 
 	dblfault_tss.tss_esp = dblfault_tss.tss_esp0 = dblfault_tss.tss_esp1 =
diff --git a/sys/amd64/amd64/mp_machdep.c b/sys/amd64/amd64/mp_machdep.c
index 6c3abeb..79fc383 100644
--- a/sys/amd64/amd64/mp_machdep.c
+++ b/sys/amd64/amd64/mp_machdep.c
@@ -484,11 +484,11 @@ init_secondary(void)
 
 	gsel_tss = GSEL(GPROC0_SEL, SEL_KPL);
 	gdt[myid * NGDT + GPROC0_SEL].sd.sd_type = SDT_SYS386TSS;
-	common_tss.tss_esp0 = 0;	/* not used until after switch */
-	common_tss.tss_ss0 = GSEL(GDATA_SEL, SEL_KPL);
-	common_tss.tss_ioopt = (sizeof common_tss) << 16;
-	tss_gdt = &gdt[myid * NGDT + GPROC0_SEL].sd;
-	common_tssd = *tss_gdt;
+	PCPU_SET(common_tss.tss_esp0, 0); /* not used until after switch */
+	PCPU_SET(common_tss.tss_ss0, GSEL(GDATA_SEL, SEL_KPL));
+	PCPU_SET(common_tss.tss_ioopt, (sizeof (struct i386tss)) << 16);
+	PCPU_SET(tss_gdt, &gdt[myid * NGDT + GPROC0_SEL].sd);
+	PCPU_SET(common_tssd, *PCPU_GET(tss_gdt));
 	ltr(gsel_tss);
 
 	pmap_set_opt();
@@ -2045,7 +2045,7 @@ start_all_aps(u_int boot_addr)
 	}
 
 	/* build our map of 'other' CPUs */
-	other_cpus = all_cpus & ~(1 << cpuid);
+	PCPU_SET(other_cpus, all_cpus & ~(1 << PCPU_GET(cpuid)));
 
 	/* fill in our (BSP) APIC version */
 	cpu_apic_versions[0] = lapic.version;
@@ -2398,9 +2398,9 @@ ap_init(void)
 #endif
 
 	/* Build our map of 'other' CPUs. */
-	other_cpus = all_cpus & ~(1 << cpuid);
+	PCPU_SET(other_cpus, all_cpus & ~(1 << PCPU_GET(cpuid)));
 
-	printf("SMP: AP CPU #%d Launched!\n", cpuid);
+	printf("SMP: AP CPU #%d Launched!\n", PCPU_GET(cpuid));
 
 	/* set up CPU registers and state */
 	cpu_setregs();
@@ -2410,8 +2410,8 @@ ap_init(void)
 
 	/* A quick check from sanity claus */
 	apic_id = (apic_id_to_logical[(lapic.id & 0x0f000000) >> 24]);
-	if (cpuid != apic_id) {
-		printf("SMP: cpuid = %d\n", cpuid);
+	if (PCPU_GET(cpuid) != apic_id) {
+		printf("SMP: cpuid = %d\n", PCPU_GET(cpuid));
 		printf("SMP: apic_id = %d\n", apic_id);
 		printf("PTD[MPPTDI] = %p\n", (void *)PTD[MPPTDI]);
 		panic("cpuid mismatch! boom!!");
@@ -2445,10 +2445,10 @@ ap_init(void)
 	 * Set curproc to our per-cpu idleproc so that mutexes have
 	 * something unique to lock with.
 	 */
-	PCPU_SET(curproc,idleproc);
+	PCPU_SET(curproc, PCPU_GET(idleproc));
 
-	microuptime(&switchtime);
-	switchticks = ticks;
+	microuptime(PCPU_PTR(switchtime));
+	PCPU_SET(switchticks, ticks);
 
 	/* ok, now grab sched_lock and enter the scheduler */
 	enable_intr();
@@ -2610,7 +2610,7 @@ forward_statclock(int pscnt)
 
 	/* Step 1: Probe state   (user, cpu, interrupt, spinlock, idle ) */
 	
-	map = other_cpus & ~stopped_cpus ;
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 	checkstate_probed_cpus = 0;
 	if (map != 0)
 		selected_apic_ipi(map,
@@ -2636,7 +2636,7 @@ forward_statclock(int pscnt)
 	
 	map = 0;
 	for (id = 0; id < mp_ncpus; id++) {
-		if (id == cpuid)
+		if (id == PCPU_GET(cpuid))
 			continue;
 		if (((1 << id) & checkstate_probed_cpus) == 0)
 			continue;
@@ -2685,7 +2685,7 @@ forward_hardclock(int pscnt)
 
 	/* Step 1: Probe state   (user, cpu, interrupt, spinlock, idle) */
 	
-	map = other_cpus & ~stopped_cpus ;
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 	checkstate_probed_cpus = 0;
 	if (map != 0)
 		selected_apic_ipi(map,
@@ -2712,7 +2712,7 @@ forward_hardclock(int pscnt)
 	
 	map = 0;
 	for (id = 0; id < mp_ncpus; id++) {
-		if (id == cpuid)
+		if (id == PCPU_GET(cpuid))
 			continue;
 		if (((1 << id) & checkstate_probed_cpus) == 0)
 			continue;
@@ -2813,8 +2813,8 @@ forward_roundrobin(void)
 		return;
 	if (!forward_roundrobin_enabled)
 		return;
-	resched_cpus |= other_cpus;
-	map = other_cpus & ~stopped_cpus ;
+	resched_cpus |= PCPU_GET(other_cpus);
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 #if 1
 	selected_apic_ipi(map, XCPUAST_OFFSET, APIC_DELMODE_FIXED);
 #else
diff --git a/sys/amd64/amd64/mptable.c b/sys/amd64/amd64/mptable.c
index 6c3abeb..79fc383 100644
--- a/sys/amd64/amd64/mptable.c
+++ b/sys/amd64/amd64/mptable.c
@@ -484,11 +484,11 @@ init_secondary(void)
 
 	gsel_tss = GSEL(GPROC0_SEL, SEL_KPL);
 	gdt[myid * NGDT + GPROC0_SEL].sd.sd_type = SDT_SYS386TSS;
-	common_tss.tss_esp0 = 0;	/* not used until after switch */
-	common_tss.tss_ss0 = GSEL(GDATA_SEL, SEL_KPL);
-	common_tss.tss_ioopt = (sizeof common_tss) << 16;
-	tss_gdt = &gdt[myid * NGDT + GPROC0_SEL].sd;
-	common_tssd = *tss_gdt;
+	PCPU_SET(common_tss.tss_esp0, 0); /* not used until after switch */
+	PCPU_SET(common_tss.tss_ss0, GSEL(GDATA_SEL, SEL_KPL));
+	PCPU_SET(common_tss.tss_ioopt, (sizeof (struct i386tss)) << 16);
+	PCPU_SET(tss_gdt, &gdt[myid * NGDT + GPROC0_SEL].sd);
+	PCPU_SET(common_tssd, *PCPU_GET(tss_gdt));
 	ltr(gsel_tss);
 
 	pmap_set_opt();
@@ -2045,7 +2045,7 @@ start_all_aps(u_int boot_addr)
 	}
 
 	/* build our map of 'other' CPUs */
-	other_cpus = all_cpus & ~(1 << cpuid);
+	PCPU_SET(other_cpus, all_cpus & ~(1 << PCPU_GET(cpuid)));
 
 	/* fill in our (BSP) APIC version */
 	cpu_apic_versions[0] = lapic.version;
@@ -2398,9 +2398,9 @@ ap_init(void)
 #endif
 
 	/* Build our map of 'other' CPUs. */
-	other_cpus = all_cpus & ~(1 << cpuid);
+	PCPU_SET(other_cpus, all_cpus & ~(1 << PCPU_GET(cpuid)));
 
-	printf("SMP: AP CPU #%d Launched!\n", cpuid);
+	printf("SMP: AP CPU #%d Launched!\n", PCPU_GET(cpuid));
 
 	/* set up CPU registers and state */
 	cpu_setregs();
@@ -2410,8 +2410,8 @@ ap_init(void)
 
 	/* A quick check from sanity claus */
 	apic_id = (apic_id_to_logical[(lapic.id & 0x0f000000) >> 24]);
-	if (cpuid != apic_id) {
-		printf("SMP: cpuid = %d\n", cpuid);
+	if (PCPU_GET(cpuid) != apic_id) {
+		printf("SMP: cpuid = %d\n", PCPU_GET(cpuid));
 		printf("SMP: apic_id = %d\n", apic_id);
 		printf("PTD[MPPTDI] = %p\n", (void *)PTD[MPPTDI]);
 		panic("cpuid mismatch! boom!!");
@@ -2445,10 +2445,10 @@ ap_init(void)
 	 * Set curproc to our per-cpu idleproc so that mutexes have
 	 * something unique to lock with.
 	 */
-	PCPU_SET(curproc,idleproc);
+	PCPU_SET(curproc, PCPU_GET(idleproc));
 
-	microuptime(&switchtime);
-	switchticks = ticks;
+	microuptime(PCPU_PTR(switchtime));
+	PCPU_SET(switchticks, ticks);
 
 	/* ok, now grab sched_lock and enter the scheduler */
 	enable_intr();
@@ -2610,7 +2610,7 @@ forward_statclock(int pscnt)
 
 	/* Step 1: Probe state   (user, cpu, interrupt, spinlock, idle ) */
 	
-	map = other_cpus & ~stopped_cpus ;
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 	checkstate_probed_cpus = 0;
 	if (map != 0)
 		selected_apic_ipi(map,
@@ -2636,7 +2636,7 @@ forward_statclock(int pscnt)
 	
 	map = 0;
 	for (id = 0; id < mp_ncpus; id++) {
-		if (id == cpuid)
+		if (id == PCPU_GET(cpuid))
 			continue;
 		if (((1 << id) & checkstate_probed_cpus) == 0)
 			continue;
@@ -2685,7 +2685,7 @@ forward_hardclock(int pscnt)
 
 	/* Step 1: Probe state   (user, cpu, interrupt, spinlock, idle) */
 	
-	map = other_cpus & ~stopped_cpus ;
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 	checkstate_probed_cpus = 0;
 	if (map != 0)
 		selected_apic_ipi(map,
@@ -2712,7 +2712,7 @@ forward_hardclock(int pscnt)
 	
 	map = 0;
 	for (id = 0; id < mp_ncpus; id++) {
-		if (id == cpuid)
+		if (id == PCPU_GET(cpuid))
 			continue;
 		if (((1 << id) & checkstate_probed_cpus) == 0)
 			continue;
@@ -2813,8 +2813,8 @@ forward_roundrobin(void)
 		return;
 	if (!forward_roundrobin_enabled)
 		return;
-	resched_cpus |= other_cpus;
-	map = other_cpus & ~stopped_cpus ;
+	resched_cpus |= PCPU_GET(other_cpus);
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 #if 1
 	selected_apic_ipi(map, XCPUAST_OFFSET, APIC_DELMODE_FIXED);
 #else
diff --git a/sys/amd64/amd64/pmap.c b/sys/amd64/amd64/pmap.c
index ef4d3f0..aa5f94e 100644
--- a/sys/amd64/amd64/pmap.c
+++ b/sys/amd64/amd64/pmap.c
@@ -453,7 +453,7 @@ pmap_set_opt(void)
 {
 	if (pseflag && (cpu_feature & CPUID_PSE)) {
 		load_cr4(rcr4() | CR4_PSE);
-		if (pdir4mb && cpuid == 0) {	/* only on BSP */
+		if (pdir4mb && PCPU_GET(cpuid) == 0) {	/* only on BSP */
 			kernel_pmap->pm_pdir[KPTDI] =
 			    PTD[KPTDI] = (pd_entry_t)pdir4mb;
 			cpu_invltlb();
@@ -581,9 +581,9 @@ static __inline void
 pmap_TLB_invalidate(pmap_t pmap, vm_offset_t va)
 {
 #if defined(SMP)
-	if (pmap->pm_active & (1 << cpuid))
+	if (pmap->pm_active & (1 << PCPU_GET(cpuid)))
 		cpu_invlpg((void *)va);
-	if (pmap->pm_active & other_cpus)
+	if (pmap->pm_active & PCPU_GET(other_cpus))
 		smp_invltlb();
 #else
 	if (pmap->pm_active)
@@ -595,9 +595,9 @@ static __inline void
 pmap_TLB_invalidate_all(pmap_t pmap)
 {
 #if defined(SMP)
-	if (pmap->pm_active & (1 << cpuid))
+	if (pmap->pm_active & (1 << PCPU_GET(cpuid)))
 		cpu_invltlb();
-	if (pmap->pm_active & other_cpus)
+	if (pmap->pm_active & PCPU_GET(other_cpus))
 		smp_invltlb();
 #else
 	if (pmap->pm_active)
@@ -652,11 +652,11 @@ pmap_pte_quick(pmap, va)
 		}
 		newpf = pde & PG_FRAME;
 #ifdef SMP
-		if ( ((* (unsigned *) prv_PMAP1) & PG_FRAME) != newpf) {
-			* (unsigned *) prv_PMAP1 = newpf | PG_RW | PG_V;
-			cpu_invlpg(prv_PADDR1);
+		if ( ((* (unsigned *) PCPU_GET(prv_PMAP1)) & PG_FRAME) != newpf) {
+			* (unsigned *) PCPU_GET(prv_PMAP1) = newpf | PG_RW | PG_V;
+			cpu_invlpg(PCPU_GET(prv_PADDR1));
 		}
-		return (unsigned *)(prv_PADDR1 + (index & (NPTEPG - 1)));
+		return (unsigned *)(PCPU_GET(prv_PADDR1) + (index & (NPTEPG - 1)));
 #else
 		if ( ((* (unsigned *) PMAP1) & PG_FRAME) != newpf) {
 			* (unsigned *) PMAP1 = newpf | PG_RW | PG_V;
@@ -1985,11 +1985,11 @@ pmap_enter(pmap_t pmap, vm_offset_t va, vm_page_t m, vm_prot_t prot,
 				pmap->pm_pdir[PTDPTDI], origpte, va);
 		}
 		if (smp_active) {
-			pdeaddr = (vm_offset_t *) IdlePTDS[cpuid];
+			pdeaddr = (vm_offset_t *) IdlePTDS[PCPU_GET(cpuid)];
 			if (((newpte = pdeaddr[va >> PDRSHIFT]) & PG_V) == 0) {
 				if ((vm_offset_t) my_idlePTD != (vm_offset_t) vtophys(pdeaddr))
 					printf("pde mismatch: %x, %x\n", my_idlePTD, pdeaddr);
-				printf("cpuid: %d, pdeaddr: 0x%x\n", cpuid, pdeaddr);
+				printf("cpuid: %d, pdeaddr: 0x%x\n", PCPU_GET(cpuid), pdeaddr);
 				panic("pmap_enter: invalid kernel page table page(1), pdir=%p, npde=%p, pde=%p, va=%p\n",
 					pmap->pm_pdir[PTDPTDI], newpte, origpte, va);
 			}
@@ -2048,7 +2048,7 @@ pmap_enter(pmap_t pmap, vm_offset_t va, vm_page_t m, vm_prot_t prot,
 				*pte |= PG_RW;
 #ifdef SMP
 				cpu_invlpg((void *)va);
-				if (pmap->pm_active & other_cpus)
+				if (pmap->pm_active & PCPU_GET(other_cpus))
 					smp_invltlb();
 #else
 				invltlb_1pg(va);
@@ -2122,7 +2122,7 @@ validate:
 		/*if (origpte)*/ {
 #ifdef SMP
 			cpu_invlpg((void *)va);
-			if (pmap->pm_active & other_cpus)
+			if (pmap->pm_active & PCPU_GET(other_cpus))
 				smp_invltlb();
 #else
 			invltlb_1pg(va);
@@ -2666,20 +2666,20 @@ pmap_zero_page(phys)
 	vm_offset_t phys;
 {
 #ifdef SMP
-	if (*(int *) prv_CMAP3)
+	if (*(int *) PCPU_GET(prv_CMAP3))
 		panic("pmap_zero_page: prv_CMAP3 busy");
 
-	*(int *) prv_CMAP3 = PG_V | PG_RW | (phys & PG_FRAME) | PG_A | PG_M;
-	cpu_invlpg(prv_CADDR3);
+	*(int *) PCPU_GET(prv_CMAP3) = PG_V | PG_RW | (phys & PG_FRAME) | PG_A | PG_M;
+	cpu_invlpg(PCPU_GET(prv_CADDR3));
 
 #if defined(I686_CPU)
 	if (cpu_class == CPUCLASS_686)
-		i686_pagezero(prv_CADDR3);
+		i686_pagezero(PCPU_GET(prv_CADDR3));
 	else
 #endif
-		bzero(prv_CADDR3, PAGE_SIZE);
+		bzero(PCPU_GET(prv_CADDR3), PAGE_SIZE);
 
-	*(int *) prv_CMAP3 = 0;
+	*(int *) PCPU_GET(prv_CMAP3) = 0;
 #else
 	if (*(int *) CMAP2)
 		panic("pmap_zero_page: CMAP2 busy");
@@ -2710,20 +2710,20 @@ pmap_zero_page_area(phys, off, size)
 	int size;
 {
 #ifdef SMP
-	if (*(int *) prv_CMAP3)
+	if (*(int *) PCPU_GET(prv_CMAP3))
 		panic("pmap_zero_page: prv_CMAP3 busy");
 
-	*(int *) prv_CMAP3 = PG_V | PG_RW | (phys & PG_FRAME) | PG_A | PG_M;
-	cpu_invlpg(prv_CADDR3);
+	*(int *) PCPU_GET(prv_CMAP3) = PG_V | PG_RW | (phys & PG_FRAME) | PG_A | PG_M;
+	cpu_invlpg(PCPU_GET(prv_CADDR3));
 
 #if defined(I686_CPU)
 	if (cpu_class == CPUCLASS_686 && off == 0 && size == PAGE_SIZE)
-		i686_pagezero(prv_CADDR3);
+		i686_pagezero(PCPU_GET(prv_CADDR3));
 	else
 #endif
-		bzero((char *)prv_CADDR3 + off, size);
+		bzero((char *)PCPU_GET(prv_CADDR3) + off, size);
 
-	*(int *) prv_CMAP3 = 0;
+	*(int *) PCPU_GET(prv_CMAP3) = 0;
 #else
 	if (*(int *) CMAP2)
 		panic("pmap_zero_page: CMAP2 busy");
@@ -2753,21 +2753,22 @@ pmap_copy_page(src, dst)
 	vm_offset_t dst;
 {
 #ifdef SMP
-	if (*(int *) prv_CMAP1)
+	if (*(int *) PCPU_GET(prv_CMAP1))
 		panic("pmap_copy_page: prv_CMAP1 busy");
-	if (*(int *) prv_CMAP2)
+	if (*(int *) PCPU_GET(prv_CMAP2))
 		panic("pmap_copy_page: prv_CMAP2 busy");
 
-	*(int *) prv_CMAP1 = PG_V | (src & PG_FRAME) | PG_A;
-	*(int *) prv_CMAP2 = PG_V | PG_RW | (dst & PG_FRAME) | PG_A | PG_M;
+	*(int *) PCPU_GET(prv_CMAP1) = PG_V | (src & PG_FRAME) | PG_A;
+	*(int *) PCPU_GET(prv_CMAP2) = PG_V | PG_RW | (dst & PG_FRAME) | PG_A | PG_M;
 
-	cpu_invlpg(prv_CADDR1);
-	cpu_invlpg(prv_CADDR2);
+	cpu_invlpg(PCPU_GET(prv_CADDR1));
+	cpu_invlpg(PCPU_GET(prv_CADDR2));
 
-	bcopy(prv_CADDR1, prv_CADDR2, PAGE_SIZE);
+	bcopy(PCPU_GET(prv_CADDR1), PCPU_GET(prv_CADDR2), PAGE_SIZE);
+
+	*(int *) PCPU_GET(prv_CMAP1) = 0;
+	*(int *) PCPU_GET(prv_CMAP2) = 0;
 
-	*(int *) prv_CMAP1 = 0;
-	*(int *) prv_CMAP2 = 0;
 #else
 	if (*(int *) CMAP1 || *(int *) CMAP2)
 		panic("pmap_copy_page: CMAP busy");
@@ -3294,7 +3295,7 @@ pmap_activate(struct proc *p)
 
 	pmap = vmspace_pmap(p->p_vmspace);
 #if defined(SMP)
-	pmap->pm_active |= 1 << cpuid;
+	pmap->pm_active |= 1 << PCPU_GET(cpuid);
 #else
 	pmap->pm_active |= 1;
 #endif
diff --git a/sys/amd64/amd64/sys_machdep.c b/sys/amd64/amd64/sys_machdep.c
index d09d882..e242b73 100644
--- a/sys/amd64/amd64/sys_machdep.c
+++ b/sys/amd64/amd64/sys_machdep.c
@@ -252,12 +252,12 @@ set_user_ldt(struct pcb *pcb)
 {
 	struct pcb_ldt *pcb_ldt;
 
-	if (pcb != curpcb)
+	if (pcb != PCPU_GET(curpcb))
 		return;
 
 	pcb_ldt = pcb->pcb_ldt;
 #ifdef SMP
-	gdt[cpuid * NGDT + GUSERLDT_SEL].sd = pcb_ldt->ldt_sd;
+	gdt[PCPU_GET(cpuid) * NGDT + GUSERLDT_SEL].sd = pcb_ldt->ldt_sd;
 #else
 	gdt[GUSERLDT_SEL].sd = pcb_ldt->ldt_sd;
 #endif
@@ -308,7 +308,7 @@ user_ldt_free(struct pcb *pcb)
 	if (pcb_ldt == NULL)
 		return;
 
-	if (pcb == curpcb) {
+	if (pcb == PCPU_GET(curpcb)) {
 		lldt(_default_ldt);
 		PCPU_SET(currentldt, _default_ldt);
 	}
diff --git a/sys/amd64/amd64/trap.c b/sys/amd64/amd64/trap.c
index db02d51..959c64f 100644
--- a/sys/amd64/amd64/trap.c
+++ b/sys/amd64/amd64/trap.c
@@ -480,7 +480,7 @@ restart:
 			if (in_vm86call)
 				break;
 
-			if (intr_nesting_level != 0)
+			if (PCPU_GET(intr_nesting_level) != 0)
 				break;
 
 			/*
@@ -493,7 +493,7 @@ restart:
 			 * a signal.
 			 */
 			if (frame.tf_eip == (int)cpu_switch_load_gs) {
-				curpcb->pcb_gs = 0;
+				PCPU_GET(curpcb)->pcb_gs = 0;
 				psignal(p, SIGBUS);
 				goto out;
 			}
@@ -519,13 +519,15 @@ restart:
 			if (frame.tf_eip == (int)doreti_popl_es) {
 				frame.tf_eip = (int)doreti_popl_es_fault;
 				goto out;
-				}
+			}
 			if (frame.tf_eip == (int)doreti_popl_fs) {
 				frame.tf_eip = (int)doreti_popl_fs_fault;
 				goto out;
 			}
-			if (curpcb && curpcb->pcb_onfault) {
-				frame.tf_eip = (int)curpcb->pcb_onfault;
+			if (PCPU_GET(curpcb) != NULL &&
+			    PCPU_GET(curpcb)->pcb_onfault != NULL) {
+				frame.tf_eip =
+				    (int)PCPU_GET(curpcb)->pcb_onfault;
 				goto out;
 			}
 			break;
@@ -685,8 +687,9 @@ trap_pfault(frame, usermode, eva)
 
 		if (p == NULL ||
 		    (!usermode && va < VM_MAXUSER_ADDRESS &&
-		     (intr_nesting_level != 0 || curpcb == NULL ||
-		      curpcb->pcb_onfault == NULL))) {
+		     (PCPU_GET(intr_nesting_level) != 0 ||
+		      PCPU_GET(curpcb) == NULL ||
+		      PCPU_GET(curpcb)->pcb_onfault == NULL))) {
 			trap_fatal(frame, eva);
 			return (-1);
 		}
@@ -748,8 +751,10 @@ trap_pfault(frame, usermode, eva)
 		return (0);
 nogo:
 	if (!usermode) {
-		if (intr_nesting_level == 0 && curpcb && curpcb->pcb_onfault) {
-			frame->tf_eip = (int)curpcb->pcb_onfault;
+		if (PCPU_GET(intr_nesting_level) == 0 &&
+		    PCPU_GET(curpcb) != NULL &&
+		    PCPU_GET(curpcb)->pcb_onfault != NULL) {
+			frame->tf_eip = (int)PCPU_GET(curpcb)->pcb_onfault;
 			return (0);
 		}
 		trap_fatal(frame, eva);
@@ -853,8 +858,10 @@ trap_pfault(frame, usermode, eva)
 		return (0);
 nogo:
 	if (!usermode) {
-		if (intr_nesting_level == 0 && curpcb && curpcb->pcb_onfault) {
-			frame->tf_eip = (int)curpcb->pcb_onfault;
+		if (PCPU_GET(intr_nesting_level) == 0 &&
+		    PCPU_GET(curpcb) != NULL &&
+		    PCPU_GET(curpcb)->pcb_onfault != NULL) {
+			frame->tf_eip = (int)PCPU_GET(curpcb)->pcb_onfault;
 			return (0);
 		}
 		trap_fatal(frame, eva);
@@ -886,7 +893,7 @@ trap_fatal(frame, eva)
 			ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel");
 #ifdef SMP
 	/* two seperate prints in case of a trap on an unmapped page */
-	printf("cpuid = %d; ", cpuid);
+	printf("cpuid = %d; ", PCPU_GET(cpuid));
 	printf("lapic.id = %08x\n", lapic.id);
 #endif
 	if (type == T_PAGEFLT) {
@@ -964,12 +971,12 @@ void
 dblfault_handler()
 {
 	printf("\nFatal double fault:\n");
-	printf("eip = 0x%x\n", common_tss.tss_eip);
-	printf("esp = 0x%x\n", common_tss.tss_esp);
-	printf("ebp = 0x%x\n", common_tss.tss_ebp);
+	printf("eip = 0x%x\n", PCPU_GET(common_tss.tss_eip));
+	printf("esp = 0x%x\n", PCPU_GET(common_tss.tss_esp));
+	printf("ebp = 0x%x\n", PCPU_GET(common_tss.tss_ebp));
 #ifdef SMP
 	/* two seperate prints in case of a trap on an unmapped page */
-	printf("cpuid = %d; ", cpuid);
+	printf("cpuid = %d; ", PCPU_GET(cpuid));
 	printf("lapic.id = %08x\n", lapic.id);
 #endif
 	panic("double fault");
diff --git a/sys/amd64/amd64/vm_machdep.c b/sys/amd64/amd64/vm_machdep.c
index 0a3bc51..552967c 100644
--- a/sys/amd64/amd64/vm_machdep.c
+++ b/sys/amd64/amd64/vm_machdep.c
@@ -143,7 +143,7 @@ cpu_fork(p1, p2, flags)
 
 #if NNPX > 0
 	/* Ensure that p1's pcb is up to date. */
-	if (npxproc == p1)
+	if (PCPU_GET(npxproc) == p1)
 		npxsave(&p1->p_addr->u_pcb.pcb_savefpu);
 #endif
 
@@ -442,23 +442,23 @@ cpu_reset()
 
 		u_int map;
 		int cnt;
-		printf("cpu_reset called on cpu#%d\n",cpuid);
+		printf("cpu_reset called on cpu#%d\n", PCPU_GET(cpuid));
 
-		map = other_cpus & ~ stopped_cpus;
+		map = PCPU_GET(other_cpus) & ~ stopped_cpus;
 
 		if (map != 0) {
 			printf("cpu_reset: Stopping other CPUs\n");
 			stop_cpus(map);		/* Stop all other CPUs */
 		}
 
-		if (cpuid == 0) {
+		if (PCPU_GET(cpuid) == 0) {
 			DELAY(1000000);
 			cpu_reset_real();
 			/* NOTREACHED */
 		} else {
 			/* We are not BSP (CPU #0) */
 
-			cpu_reset_proxyid = cpuid;
+			cpu_reset_proxyid = PCPU_GET(cpuid);
 			cpustop_restartfunc = cpu_reset_proxy;
 			cpu_reset_proxy_active = 0;
 			printf("cpu_reset: Restarting BSP\n");
diff --git a/sys/amd64/include/cpu.h b/sys/amd64/include/cpu.h
index 4c6cb43..6c4162b 100644
--- a/sys/amd64/include/cpu.h
+++ b/sys/amd64/include/cpu.h
@@ -62,7 +62,7 @@
 #define	CLKF_USERMODE(framep) \
 	((ISPL((framep)->cf_cs) == SEL_UPL) || (framep->cf_eflags & PSL_VM))
 
-#define CLKF_INTR(framep)	(intr_nesting_level >= 2)
+#define CLKF_INTR(framep)	(PCPU_GET(intr_nesting_level) >= 2)
 #define	CLKF_PC(framep)		((framep)->cf_eip)
 
 /*
@@ -82,7 +82,7 @@
 #define	need_resched() do {						\
 	PCPU_SET(astpending, AST_RESCHED|AST_PENDING);			\
 } while (0)
-#define	resched_wanted()	(astpending & AST_RESCHED)
+#define	resched_wanted()	(PCPU_GET(astpending) & AST_RESCHED)
 
 /*
  * Arrange to handle pending profiling ticks before returning to user mode.
@@ -105,7 +105,7 @@
  */
 #define	signotify(p)	aston()
 #define	aston() do {							\
-	PCPU_SET(astpending, astpending | AST_PENDING);			\
+	PCPU_SET(astpending, PCPU_GET(astpending) | AST_PENDING);	\
 } while (0)
 #define astoff()
 
diff --git a/sys/amd64/include/mptable.h b/sys/amd64/include/mptable.h
index 6c3abeb..79fc383 100644
--- a/sys/amd64/include/mptable.h
+++ b/sys/amd64/include/mptable.h
@@ -484,11 +484,11 @@ init_secondary(void)
 
 	gsel_tss = GSEL(GPROC0_SEL, SEL_KPL);
 	gdt[myid * NGDT + GPROC0_SEL].sd.sd_type = SDT_SYS386TSS;
-	common_tss.tss_esp0 = 0;	/* not used until after switch */
-	common_tss.tss_ss0 = GSEL(GDATA_SEL, SEL_KPL);
-	common_tss.tss_ioopt = (sizeof common_tss) << 16;
-	tss_gdt = &gdt[myid * NGDT + GPROC0_SEL].sd;
-	common_tssd = *tss_gdt;
+	PCPU_SET(common_tss.tss_esp0, 0); /* not used until after switch */
+	PCPU_SET(common_tss.tss_ss0, GSEL(GDATA_SEL, SEL_KPL));
+	PCPU_SET(common_tss.tss_ioopt, (sizeof (struct i386tss)) << 16);
+	PCPU_SET(tss_gdt, &gdt[myid * NGDT + GPROC0_SEL].sd);
+	PCPU_SET(common_tssd, *PCPU_GET(tss_gdt));
 	ltr(gsel_tss);
 
 	pmap_set_opt();
@@ -2045,7 +2045,7 @@ start_all_aps(u_int boot_addr)
 	}
 
 	/* build our map of 'other' CPUs */
-	other_cpus = all_cpus & ~(1 << cpuid);
+	PCPU_SET(other_cpus, all_cpus & ~(1 << PCPU_GET(cpuid)));
 
 	/* fill in our (BSP) APIC version */
 	cpu_apic_versions[0] = lapic.version;
@@ -2398,9 +2398,9 @@ ap_init(void)
 #endif
 
 	/* Build our map of 'other' CPUs. */
-	other_cpus = all_cpus & ~(1 << cpuid);
+	PCPU_SET(other_cpus, all_cpus & ~(1 << PCPU_GET(cpuid)));
 
-	printf("SMP: AP CPU #%d Launched!\n", cpuid);
+	printf("SMP: AP CPU #%d Launched!\n", PCPU_GET(cpuid));
 
 	/* set up CPU registers and state */
 	cpu_setregs();
@@ -2410,8 +2410,8 @@ ap_init(void)
 
 	/* A quick check from sanity claus */
 	apic_id = (apic_id_to_logical[(lapic.id & 0x0f000000) >> 24]);
-	if (cpuid != apic_id) {
-		printf("SMP: cpuid = %d\n", cpuid);
+	if (PCPU_GET(cpuid) != apic_id) {
+		printf("SMP: cpuid = %d\n", PCPU_GET(cpuid));
 		printf("SMP: apic_id = %d\n", apic_id);
 		printf("PTD[MPPTDI] = %p\n", (void *)PTD[MPPTDI]);
 		panic("cpuid mismatch! boom!!");
@@ -2445,10 +2445,10 @@ ap_init(void)
 	 * Set curproc to our per-cpu idleproc so that mutexes have
 	 * something unique to lock with.
 	 */
-	PCPU_SET(curproc,idleproc);
+	PCPU_SET(curproc, PCPU_GET(idleproc));
 
-	microuptime(&switchtime);
-	switchticks = ticks;
+	microuptime(PCPU_PTR(switchtime));
+	PCPU_SET(switchticks, ticks);
 
 	/* ok, now grab sched_lock and enter the scheduler */
 	enable_intr();
@@ -2610,7 +2610,7 @@ forward_statclock(int pscnt)
 
 	/* Step 1: Probe state   (user, cpu, interrupt, spinlock, idle ) */
 	
-	map = other_cpus & ~stopped_cpus ;
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 	checkstate_probed_cpus = 0;
 	if (map != 0)
 		selected_apic_ipi(map,
@@ -2636,7 +2636,7 @@ forward_statclock(int pscnt)
 	
 	map = 0;
 	for (id = 0; id < mp_ncpus; id++) {
-		if (id == cpuid)
+		if (id == PCPU_GET(cpuid))
 			continue;
 		if (((1 << id) & checkstate_probed_cpus) == 0)
 			continue;
@@ -2685,7 +2685,7 @@ forward_hardclock(int pscnt)
 
 	/* Step 1: Probe state   (user, cpu, interrupt, spinlock, idle) */
 	
-	map = other_cpus & ~stopped_cpus ;
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 	checkstate_probed_cpus = 0;
 	if (map != 0)
 		selected_apic_ipi(map,
@@ -2712,7 +2712,7 @@ forward_hardclock(int pscnt)
 	
 	map = 0;
 	for (id = 0; id < mp_ncpus; id++) {
-		if (id == cpuid)
+		if (id == PCPU_GET(cpuid))
 			continue;
 		if (((1 << id) & checkstate_probed_cpus) == 0)
 			continue;
@@ -2813,8 +2813,8 @@ forward_roundrobin(void)
 		return;
 	if (!forward_roundrobin_enabled)
 		return;
-	resched_cpus |= other_cpus;
-	map = other_cpus & ~stopped_cpus ;
+	resched_cpus |= PCPU_GET(other_cpus);
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 #if 1
 	selected_apic_ipi(map, XCPUAST_OFFSET, APIC_DELMODE_FIXED);
 #else
diff --git a/sys/amd64/isa/npx.c b/sys/amd64/isa/npx.c
index 2672a63..b79d511 100644
--- a/sys/amd64/isa/npx.c
+++ b/sys/amd64/isa/npx.c
@@ -493,8 +493,8 @@ npxinit(control)
 	npxsave(&dummy);
 	stop_emulating();
 	fldcw(&control);
-	if (curpcb != NULL)
-		fnsave(&curpcb->pcb_savefpu);
+	if (PCPU_GET(curpcb) != NULL)
+		fnsave(&PCPU_GET(curpcb)->pcb_savefpu);
 	start_emulating();
 }
 
@@ -506,14 +506,14 @@ npxexit(p)
 	struct proc *p;
 {
 
-	if (p == npxproc)
-		npxsave(&curpcb->pcb_savefpu);
+	if (p == PCPU_GET(npxproc))
+		npxsave(&PCPU_GET(curpcb)->pcb_savefpu);
 #ifdef NPX_DEBUG
 	if (npx_exists) {
 		u_int	masked_exceptions;
 
-		masked_exceptions = curpcb->pcb_savefpu.sv_env.en_cw
-				    & curpcb->pcb_savefpu.sv_env.en_sw & 0x7f;
+		masked_exceptions = PCPU_GET(curpcb)->pcb_savefpu.sv_env.en_cw
+		    &PCPU_GET(curpcb)->pcb_savefpu.sv_env.en_sw & 0x7f;
 		/*
 		 * Log exceptions that would have trapped with the old
 		 * control word (overflow, divide by 0, and invalid operand).
@@ -722,19 +722,19 @@ npx_intr(dummy)
 	u_short control;
 	struct intrframe *frame;
 
-	if (npxproc == NULL || !npx_exists) {
+	if (PCPU_GET(npxproc) == NULL || !npx_exists) {
 		printf("npxintr: npxproc = %p, curproc = %p, npx_exists = %d\n",
-		       npxproc, curproc, npx_exists);
+		       PCPU_GET(npxproc), curproc, npx_exists);
 		panic("npxintr from nowhere");
 	}
-	if (npxproc != curproc) {
+	if (PCPU_GET(npxproc) != curproc) {
 		printf("npxintr: npxproc = %p, curproc = %p, npx_exists = %d\n",
-		       npxproc, curproc, npx_exists);
+		       PCPU_GET(npxproc), curproc, npx_exists);
 		panic("npxintr from non-current process");
 	}
 
 	outb(0xf0, 0);
-	fnstsw(&curpcb->pcb_savefpu.sv_ex_sw);
+	fnstsw(&PCPU_GET(curpcb)->pcb_savefpu.sv_ex_sw);
 	fnstcw(&control);
 	fnclex();
 
@@ -760,8 +760,8 @@ npx_intr(dummy)
 		 * this exception.
 		 */
 		code = 
-		    fpetable[(curpcb->pcb_savefpu.sv_ex_sw & ~control & 0x3f) |
-			(curpcb->pcb_savefpu.sv_ex_sw & 0x40)];
+		    fpetable[(PCPU_GET(curpcb)->pcb_savefpu.sv_ex_sw & ~control & 0x3f) |
+			(PCPU_GET(curpcb)->pcb_savefpu.sv_ex_sw & 0x40)];
 		trapsignal(curproc, SIGFPE, code);
 	} else {
 		/*
@@ -794,9 +794,9 @@ npxdna()
 {
 	if (!npx_exists)
 		return (0);
-	if (npxproc != NULL) {
+	if (PCPU_GET(npxproc) != NULL) {
 		printf("npxdna: npxproc = %p, curproc = %p\n",
-		       npxproc, curproc);
+		       PCPU_GET(npxproc), curproc);
 		panic("npxdna");
 	}
 	stop_emulating();
@@ -804,7 +804,7 @@ npxdna()
 	 * Record new context early in case frstor causes an IRQ13.
 	 */
 	PCPU_SET(npxproc, CURPROC);
-	curpcb->pcb_savefpu.sv_ex_sw = 0;
+	PCPU_GET(curpcb)->pcb_savefpu.sv_ex_sw = 0;
 	/*
 	 * The following frstor may cause an IRQ13 when the state being
 	 * restored has a pending error.  The error will appear to have been
@@ -817,7 +817,7 @@ npxdna()
 	 * fnsave are broken, so our treatment breaks fnclex if it is the
 	 * first FPU instruction after a context switch.
 	 */
-	frstor(&curpcb->pcb_savefpu);
+	frstor(&PCPU_GET(curpcb)->pcb_savefpu);
 
 	return (1);
 }
diff --git a/sys/dev/usb/ohci.c b/sys/dev/usb/ohci.c
index f559faa..c9062a4 100644
--- a/sys/dev/usb/ohci.c
+++ b/sys/dev/usb/ohci.c
@@ -1889,7 +1889,7 @@ ohci_abort_xfer(xfer, status)
 		timeout(ohci_abort_xfer_end, xfer, hz / USB_FRAMES_PER_SECOND);
 	} else {
 #if defined(DIAGNOSTIC) && defined(__i386__) && defined(__FreeBSD__)
-		KASSERT(intr_nesting_level == 0,
+		KASSERT(PCPU_GET(intr_nesting_level) == 0,
 	        	("ohci_abort_req in interrupt context"));
 #endif
 		usb_delay_ms(opipe->pipe.device->bus, 1);
diff --git a/sys/dev/usb/uhci.c b/sys/dev/usb/uhci.c
index 99c85c4..7098db1 100644
--- a/sys/dev/usb/uhci.c
+++ b/sys/dev/usb/uhci.c
@@ -1610,7 +1610,7 @@ uhci_abort_xfer(usbd_xfer_handle xfer, usbd_status status)
 		timeout(uhci_abort_xfer_end, xfer, hz / USB_FRAMES_PER_SECOND);
 	} else {
 #if defined(DIAGNOSTIC) && defined(__i386__) && defined(__FreeBSD__)
-		KASSERT(intr_nesting_level == 0,
+		KASSERT(PCPU_GET(intr_nesting_level) == 0,
 	        	("ohci_abort_req in interrupt context"));
 #endif
 		usb_delay_ms(xfer->pipe->device->bus, 1);
diff --git a/sys/dev/usb/usbdi.c b/sys/dev/usb/usbdi.c
index 7dc8a0e..c39a856 100644
--- a/sys/dev/usb/usbdi.c
+++ b/sys/dev/usb/usbdi.c
@@ -931,7 +931,7 @@ usbd_do_request_flags(dev, req, data, flags, actlen)
 
 #ifdef DIAGNOSTIC
 #if defined(__i386__) && defined(__FreeBSD__)
-	KASSERT(intr_nesting_level == 0,
+	KASSERT(PCPU_GET(intr_nesting_level) == 0,
 	       	("usbd_do_request: in interrupt context"));
 #endif
 	if (dev->bus->intr_context) {
diff --git a/sys/dev/vinum/vinumhdr.h b/sys/dev/vinum/vinumhdr.h
index be30beb..8d53f0f 100644
--- a/sys/dev/vinum/vinumhdr.h
+++ b/sys/dev/vinum/vinumhdr.h
@@ -86,7 +86,7 @@ caddr_t MMalloc (int size, char *, int);
 void FFree (void *mem, char *, int);
 #define LOCKDRIVE(d) lockdrive (d, __FILE__, __LINE__)
 #else
-#define Malloc(x)  malloc((x), M_DEVBUF, intr_nesting_level == 0)
+#define Malloc(x)  malloc((x), M_DEVBUF, PCPU_GET(intr_nesting_level) == 0)
 #define Free(x)    free((x), M_DEVBUF)
 #define LOCKDRIVE(d) lockdrive (d)
 #endif
diff --git a/sys/dev/vinum/vinummemory.c b/sys/dev/vinum/vinummemory.c
index 8e501a2..e28f7ae 100644
--- a/sys/dev/vinum/vinummemory.c
+++ b/sys/dev/vinum/vinummemory.c
@@ -151,7 +151,7 @@ MMalloc(int size, char *file, int line)
 	return 0;					    /* can't continue */
     }
     /* Wait for malloc if we can */
-    result = malloc(size, M_DEVBUF, intr_nesting_level == 0 ? M_WAITOK : M_NOWAIT);
+    result = malloc(size, M_DEVBUF, PCPU_GET(intr_nesting_level) == 0 ? M_WAITOK : M_NOWAIT);
     if (result == NULL)
 	log(LOG_ERR, "vinum: can't allocate %d bytes from %s:%d\n", size, file, line);
     else {
diff --git a/sys/i386/i386/bios.c b/sys/i386/i386/bios.c
index a22f1549..e90de29 100644
--- a/sys/i386/i386/bios.c
+++ b/sys/i386/i386/bios.c
@@ -258,7 +258,7 @@ set_bios_selectors(struct bios_segments *seg, int flags)
     union descriptor *p_gdt;
 
 #ifdef SMP
-    p_gdt = &gdt[cpuid * NGDT];
+    p_gdt = &gdt[PCPU_GET(cpuid) * NGDT];
 #else
     p_gdt = gdt;
 #endif
diff --git a/sys/i386/i386/db_interface.c b/sys/i386/i386/db_interface.c
index fe94264..ee4ab19 100644
--- a/sys/i386/i386/db_interface.c
+++ b/sys/i386/i386/db_interface.c
@@ -142,11 +142,12 @@ kdb_trap(type, code, regs)
 #ifdef CPUSTOP_ON_DDBBREAK
 
 #if defined(VERBOSE_CPUSTOP_ON_DDBBREAK)
-	db_printf("\nCPU%d stopping CPUs: 0x%08x...", cpuid, other_cpus);
+	db_printf("\nCPU%d stopping CPUs: 0x%08x...", PCPU_GET(cpuid),
+	    PCPU_GET(other_cpus));
 #endif /* VERBOSE_CPUSTOP_ON_DDBBREAK */
 
 	/* We stop all CPUs except ourselves (obviously) */
-	stop_cpus(other_cpus);
+	stop_cpus(PCPU_GET(other_cpus));
 
 #if defined(VERBOSE_CPUSTOP_ON_DDBBREAK)
 	db_printf(" stopped.\n");
@@ -171,13 +172,14 @@ kdb_trap(type, code, regs)
 #ifdef CPUSTOP_ON_DDBBREAK
 
 #if defined(VERBOSE_CPUSTOP_ON_DDBBREAK)
-	db_printf("\nCPU%d restarting CPUs: 0x%08x...", cpuid, stopped_cpus);
+	db_printf("\nCPU%d restarting CPUs: 0x%08x...", PCPU_GET(cpuid),
+	    stopped_cpus);
 #endif /* VERBOSE_CPUSTOP_ON_DDBBREAK */
 
 	/* Restart all the CPUs we previously stopped */
-	if (stopped_cpus != other_cpus && smp_started != 0) {
+	if (stopped_cpus != PCPU_GET(other_cpus) && smp_started != 0) {
 		db_printf("whoa, other_cpus: 0x%08x, stopped_cpus: 0x%08x\n",
-			  other_cpus, stopped_cpus);
+			  PCPU_GET(other_cpus), stopped_cpus);
 		panic("stop_cpus() failed");
 	}
 	restart_cpus(stopped_cpus);
diff --git a/sys/i386/i386/machdep.c b/sys/i386/i386/machdep.c
index c1c6f29..5abdc05 100644
--- a/sys/i386/i386/machdep.c
+++ b/sys/i386/i386/machdep.c
@@ -1051,7 +1051,7 @@ setregs(p, entry, stack, ps_strings)
 	regs->tf_ebx = ps_strings;
 
 	/* reset %gs as well */
-	if (pcb == curpcb)
+	if (pcb == PCPU_GET(curpcb))
 		load_gs(_udatasel);
 	else
 		pcb->pcb_gs = _udatasel;
@@ -1067,7 +1067,7 @@ setregs(p, entry, stack, ps_strings)
                 pcb->pcb_dr3 = 0;
                 pcb->pcb_dr6 = 0;
                 pcb->pcb_dr7 = 0;
-                if (pcb == curpcb) {
+                if (pcb == PCPU_GET(curpcb)) {
 		        /*
 			 * Clear the debug registers on the running
 			 * CPU, otherwise they will end up affecting
@@ -1970,13 +1970,14 @@ init386(first)
 	initializecpu();	/* Initialize CPU registers */
 
 	/* make an initial tss so cpu can get interrupt stack on syscall! */
-	common_tss.tss_esp0 = (int) proc0.p_addr + UPAGES*PAGE_SIZE - 16;
-	common_tss.tss_ss0 = GSEL(GDATA_SEL, SEL_KPL);
+	PCPU_SET(common_tss.tss_esp0,
+	    (int) proc0.p_addr + UPAGES*PAGE_SIZE - 16);
+	PCPU_SET(common_tss.tss_ss0, GSEL(GDATA_SEL, SEL_KPL));
 	gsel_tss = GSEL(GPROC0_SEL, SEL_KPL);
 	private_tss = 0;
-	tss_gdt = &gdt[GPROC0_SEL].sd;
-	common_tssd = *tss_gdt;
-	common_tss.tss_ioopt = (sizeof common_tss) << 16;
+	PCPU_SET(tss_gdt, &gdt[GPROC0_SEL].sd);
+	PCPU_SET(common_tssd, *PCPU_GET(tss_gdt));
+	PCPU_SET(common_tss.tss_ioopt, (sizeof (struct i386tss)) << 16);
 	ltr(gsel_tss);
 
 	dblfault_tss.tss_esp = dblfault_tss.tss_esp0 = dblfault_tss.tss_esp1 =
diff --git a/sys/i386/i386/mp_machdep.c b/sys/i386/i386/mp_machdep.c
index 6c3abeb..79fc383 100644
--- a/sys/i386/i386/mp_machdep.c
+++ b/sys/i386/i386/mp_machdep.c
@@ -484,11 +484,11 @@ init_secondary(void)
 
 	gsel_tss = GSEL(GPROC0_SEL, SEL_KPL);
 	gdt[myid * NGDT + GPROC0_SEL].sd.sd_type = SDT_SYS386TSS;
-	common_tss.tss_esp0 = 0;	/* not used until after switch */
-	common_tss.tss_ss0 = GSEL(GDATA_SEL, SEL_KPL);
-	common_tss.tss_ioopt = (sizeof common_tss) << 16;
-	tss_gdt = &gdt[myid * NGDT + GPROC0_SEL].sd;
-	common_tssd = *tss_gdt;
+	PCPU_SET(common_tss.tss_esp0, 0); /* not used until after switch */
+	PCPU_SET(common_tss.tss_ss0, GSEL(GDATA_SEL, SEL_KPL));
+	PCPU_SET(common_tss.tss_ioopt, (sizeof (struct i386tss)) << 16);
+	PCPU_SET(tss_gdt, &gdt[myid * NGDT + GPROC0_SEL].sd);
+	PCPU_SET(common_tssd, *PCPU_GET(tss_gdt));
 	ltr(gsel_tss);
 
 	pmap_set_opt();
@@ -2045,7 +2045,7 @@ start_all_aps(u_int boot_addr)
 	}
 
 	/* build our map of 'other' CPUs */
-	other_cpus = all_cpus & ~(1 << cpuid);
+	PCPU_SET(other_cpus, all_cpus & ~(1 << PCPU_GET(cpuid)));
 
 	/* fill in our (BSP) APIC version */
 	cpu_apic_versions[0] = lapic.version;
@@ -2398,9 +2398,9 @@ ap_init(void)
 #endif
 
 	/* Build our map of 'other' CPUs. */
-	other_cpus = all_cpus & ~(1 << cpuid);
+	PCPU_SET(other_cpus, all_cpus & ~(1 << PCPU_GET(cpuid)));
 
-	printf("SMP: AP CPU #%d Launched!\n", cpuid);
+	printf("SMP: AP CPU #%d Launched!\n", PCPU_GET(cpuid));
 
 	/* set up CPU registers and state */
 	cpu_setregs();
@@ -2410,8 +2410,8 @@ ap_init(void)
 
 	/* A quick check from sanity claus */
 	apic_id = (apic_id_to_logical[(lapic.id & 0x0f000000) >> 24]);
-	if (cpuid != apic_id) {
-		printf("SMP: cpuid = %d\n", cpuid);
+	if (PCPU_GET(cpuid) != apic_id) {
+		printf("SMP: cpuid = %d\n", PCPU_GET(cpuid));
 		printf("SMP: apic_id = %d\n", apic_id);
 		printf("PTD[MPPTDI] = %p\n", (void *)PTD[MPPTDI]);
 		panic("cpuid mismatch! boom!!");
@@ -2445,10 +2445,10 @@ ap_init(void)
 	 * Set curproc to our per-cpu idleproc so that mutexes have
 	 * something unique to lock with.
 	 */
-	PCPU_SET(curproc,idleproc);
+	PCPU_SET(curproc, PCPU_GET(idleproc));
 
-	microuptime(&switchtime);
-	switchticks = ticks;
+	microuptime(PCPU_PTR(switchtime));
+	PCPU_SET(switchticks, ticks);
 
 	/* ok, now grab sched_lock and enter the scheduler */
 	enable_intr();
@@ -2610,7 +2610,7 @@ forward_statclock(int pscnt)
 
 	/* Step 1: Probe state   (user, cpu, interrupt, spinlock, idle ) */
 	
-	map = other_cpus & ~stopped_cpus ;
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 	checkstate_probed_cpus = 0;
 	if (map != 0)
 		selected_apic_ipi(map,
@@ -2636,7 +2636,7 @@ forward_statclock(int pscnt)
 	
 	map = 0;
 	for (id = 0; id < mp_ncpus; id++) {
-		if (id == cpuid)
+		if (id == PCPU_GET(cpuid))
 			continue;
 		if (((1 << id) & checkstate_probed_cpus) == 0)
 			continue;
@@ -2685,7 +2685,7 @@ forward_hardclock(int pscnt)
 
 	/* Step 1: Probe state   (user, cpu, interrupt, spinlock, idle) */
 	
-	map = other_cpus & ~stopped_cpus ;
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 	checkstate_probed_cpus = 0;
 	if (map != 0)
 		selected_apic_ipi(map,
@@ -2712,7 +2712,7 @@ forward_hardclock(int pscnt)
 	
 	map = 0;
 	for (id = 0; id < mp_ncpus; id++) {
-		if (id == cpuid)
+		if (id == PCPU_GET(cpuid))
 			continue;
 		if (((1 << id) & checkstate_probed_cpus) == 0)
 			continue;
@@ -2813,8 +2813,8 @@ forward_roundrobin(void)
 		return;
 	if (!forward_roundrobin_enabled)
 		return;
-	resched_cpus |= other_cpus;
-	map = other_cpus & ~stopped_cpus ;
+	resched_cpus |= PCPU_GET(other_cpus);
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 #if 1
 	selected_apic_ipi(map, XCPUAST_OFFSET, APIC_DELMODE_FIXED);
 #else
diff --git a/sys/i386/i386/mpapic.c b/sys/i386/i386/mpapic.c
index 8b706c0..33aab28 100644
--- a/sys/i386/i386/mpapic.c
+++ b/sys/i386/i386/mpapic.c
@@ -62,7 +62,7 @@ apic_initialize(void)
 	/* setup LVT1 as ExtINT */
 	temp = lapic.lvt_lint0;
 	temp &= ~(APIC_LVT_M | APIC_LVT_TM | APIC_LVT_IIPP | APIC_LVT_DM);
-	if (cpuid == 0)
+	if (PCPU_GET(cpuid) == 0)
 		temp |= 0x00000700;	/* process ExtInts */
 	else
 		temp |= 0x00010700;	/* mask ExtInts */
@@ -94,7 +94,7 @@ apic_initialize(void)
 	temp |= (XSPURIOUSINT_OFFSET & APIC_SVR_VEC_PROG);
 
 #if defined(TEST_TEST1)
-	if (cpuid == GUARD_CPU) {
+	if (PCPU_GET(cpuid) == GUARD_CPU) {
 		temp &= ~APIC_SVR_SWEN;	/* software DISABLE APIC */
 	}
 #endif  /** TEST_TEST1 */
@@ -112,7 +112,7 @@ apic_initialize(void)
 void
 apic_dump(char* str)
 {
-	printf("SMP: CPU%d %s:\n", cpuid, str);
+	printf("SMP: CPU%d %s:\n", PCPU_GET(cpuid), str);
 	printf("     lint0: 0x%08x lint1: 0x%08x TPR: 0x%08x SVR: 0x%08x\n",
 		lapic.lvt_lint0, lapic.lvt_lint1, lapic.tpr, lapic.svr);
 }
diff --git a/sys/i386/i386/mptable.c b/sys/i386/i386/mptable.c
index 6c3abeb..79fc383 100644
--- a/sys/i386/i386/mptable.c
+++ b/sys/i386/i386/mptable.c
@@ -484,11 +484,11 @@ init_secondary(void)
 
 	gsel_tss = GSEL(GPROC0_SEL, SEL_KPL);
 	gdt[myid * NGDT + GPROC0_SEL].sd.sd_type = SDT_SYS386TSS;
-	common_tss.tss_esp0 = 0;	/* not used until after switch */
-	common_tss.tss_ss0 = GSEL(GDATA_SEL, SEL_KPL);
-	common_tss.tss_ioopt = (sizeof common_tss) << 16;
-	tss_gdt = &gdt[myid * NGDT + GPROC0_SEL].sd;
-	common_tssd = *tss_gdt;
+	PCPU_SET(common_tss.tss_esp0, 0); /* not used until after switch */
+	PCPU_SET(common_tss.tss_ss0, GSEL(GDATA_SEL, SEL_KPL));
+	PCPU_SET(common_tss.tss_ioopt, (sizeof (struct i386tss)) << 16);
+	PCPU_SET(tss_gdt, &gdt[myid * NGDT + GPROC0_SEL].sd);
+	PCPU_SET(common_tssd, *PCPU_GET(tss_gdt));
 	ltr(gsel_tss);
 
 	pmap_set_opt();
@@ -2045,7 +2045,7 @@ start_all_aps(u_int boot_addr)
 	}
 
 	/* build our map of 'other' CPUs */
-	other_cpus = all_cpus & ~(1 << cpuid);
+	PCPU_SET(other_cpus, all_cpus & ~(1 << PCPU_GET(cpuid)));
 
 	/* fill in our (BSP) APIC version */
 	cpu_apic_versions[0] = lapic.version;
@@ -2398,9 +2398,9 @@ ap_init(void)
 #endif
 
 	/* Build our map of 'other' CPUs. */
-	other_cpus = all_cpus & ~(1 << cpuid);
+	PCPU_SET(other_cpus, all_cpus & ~(1 << PCPU_GET(cpuid)));
 
-	printf("SMP: AP CPU #%d Launched!\n", cpuid);
+	printf("SMP: AP CPU #%d Launched!\n", PCPU_GET(cpuid));
 
 	/* set up CPU registers and state */
 	cpu_setregs();
@@ -2410,8 +2410,8 @@ ap_init(void)
 
 	/* A quick check from sanity claus */
 	apic_id = (apic_id_to_logical[(lapic.id & 0x0f000000) >> 24]);
-	if (cpuid != apic_id) {
-		printf("SMP: cpuid = %d\n", cpuid);
+	if (PCPU_GET(cpuid) != apic_id) {
+		printf("SMP: cpuid = %d\n", PCPU_GET(cpuid));
 		printf("SMP: apic_id = %d\n", apic_id);
 		printf("PTD[MPPTDI] = %p\n", (void *)PTD[MPPTDI]);
 		panic("cpuid mismatch! boom!!");
@@ -2445,10 +2445,10 @@ ap_init(void)
 	 * Set curproc to our per-cpu idleproc so that mutexes have
 	 * something unique to lock with.
 	 */
-	PCPU_SET(curproc,idleproc);
+	PCPU_SET(curproc, PCPU_GET(idleproc));
 
-	microuptime(&switchtime);
-	switchticks = ticks;
+	microuptime(PCPU_PTR(switchtime));
+	PCPU_SET(switchticks, ticks);
 
 	/* ok, now grab sched_lock and enter the scheduler */
 	enable_intr();
@@ -2610,7 +2610,7 @@ forward_statclock(int pscnt)
 
 	/* Step 1: Probe state   (user, cpu, interrupt, spinlock, idle ) */
 	
-	map = other_cpus & ~stopped_cpus ;
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 	checkstate_probed_cpus = 0;
 	if (map != 0)
 		selected_apic_ipi(map,
@@ -2636,7 +2636,7 @@ forward_statclock(int pscnt)
 	
 	map = 0;
 	for (id = 0; id < mp_ncpus; id++) {
-		if (id == cpuid)
+		if (id == PCPU_GET(cpuid))
 			continue;
 		if (((1 << id) & checkstate_probed_cpus) == 0)
 			continue;
@@ -2685,7 +2685,7 @@ forward_hardclock(int pscnt)
 
 	/* Step 1: Probe state   (user, cpu, interrupt, spinlock, idle) */
 	
-	map = other_cpus & ~stopped_cpus ;
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 	checkstate_probed_cpus = 0;
 	if (map != 0)
 		selected_apic_ipi(map,
@@ -2712,7 +2712,7 @@ forward_hardclock(int pscnt)
 	
 	map = 0;
 	for (id = 0; id < mp_ncpus; id++) {
-		if (id == cpuid)
+		if (id == PCPU_GET(cpuid))
 			continue;
 		if (((1 << id) & checkstate_probed_cpus) == 0)
 			continue;
@@ -2813,8 +2813,8 @@ forward_roundrobin(void)
 		return;
 	if (!forward_roundrobin_enabled)
 		return;
-	resched_cpus |= other_cpus;
-	map = other_cpus & ~stopped_cpus ;
+	resched_cpus |= PCPU_GET(other_cpus);
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 #if 1
 	selected_apic_ipi(map, XCPUAST_OFFSET, APIC_DELMODE_FIXED);
 #else
diff --git a/sys/i386/i386/pmap.c b/sys/i386/i386/pmap.c
index ef4d3f0..aa5f94e 100644
--- a/sys/i386/i386/pmap.c
+++ b/sys/i386/i386/pmap.c
@@ -453,7 +453,7 @@ pmap_set_opt(void)
 {
 	if (pseflag && (cpu_feature & CPUID_PSE)) {
 		load_cr4(rcr4() | CR4_PSE);
-		if (pdir4mb && cpuid == 0) {	/* only on BSP */
+		if (pdir4mb && PCPU_GET(cpuid) == 0) {	/* only on BSP */
 			kernel_pmap->pm_pdir[KPTDI] =
 			    PTD[KPTDI] = (pd_entry_t)pdir4mb;
 			cpu_invltlb();
@@ -581,9 +581,9 @@ static __inline void
 pmap_TLB_invalidate(pmap_t pmap, vm_offset_t va)
 {
 #if defined(SMP)
-	if (pmap->pm_active & (1 << cpuid))
+	if (pmap->pm_active & (1 << PCPU_GET(cpuid)))
 		cpu_invlpg((void *)va);
-	if (pmap->pm_active & other_cpus)
+	if (pmap->pm_active & PCPU_GET(other_cpus))
 		smp_invltlb();
 #else
 	if (pmap->pm_active)
@@ -595,9 +595,9 @@ static __inline void
 pmap_TLB_invalidate_all(pmap_t pmap)
 {
 #if defined(SMP)
-	if (pmap->pm_active & (1 << cpuid))
+	if (pmap->pm_active & (1 << PCPU_GET(cpuid)))
 		cpu_invltlb();
-	if (pmap->pm_active & other_cpus)
+	if (pmap->pm_active & PCPU_GET(other_cpus))
 		smp_invltlb();
 #else
 	if (pmap->pm_active)
@@ -652,11 +652,11 @@ pmap_pte_quick(pmap, va)
 		}
 		newpf = pde & PG_FRAME;
 #ifdef SMP
-		if ( ((* (unsigned *) prv_PMAP1) & PG_FRAME) != newpf) {
-			* (unsigned *) prv_PMAP1 = newpf | PG_RW | PG_V;
-			cpu_invlpg(prv_PADDR1);
+		if ( ((* (unsigned *) PCPU_GET(prv_PMAP1)) & PG_FRAME) != newpf) {
+			* (unsigned *) PCPU_GET(prv_PMAP1) = newpf | PG_RW | PG_V;
+			cpu_invlpg(PCPU_GET(prv_PADDR1));
 		}
-		return (unsigned *)(prv_PADDR1 + (index & (NPTEPG - 1)));
+		return (unsigned *)(PCPU_GET(prv_PADDR1) + (index & (NPTEPG - 1)));
 #else
 		if ( ((* (unsigned *) PMAP1) & PG_FRAME) != newpf) {
 			* (unsigned *) PMAP1 = newpf | PG_RW | PG_V;
@@ -1985,11 +1985,11 @@ pmap_enter(pmap_t pmap, vm_offset_t va, vm_page_t m, vm_prot_t prot,
 				pmap->pm_pdir[PTDPTDI], origpte, va);
 		}
 		if (smp_active) {
-			pdeaddr = (vm_offset_t *) IdlePTDS[cpuid];
+			pdeaddr = (vm_offset_t *) IdlePTDS[PCPU_GET(cpuid)];
 			if (((newpte = pdeaddr[va >> PDRSHIFT]) & PG_V) == 0) {
 				if ((vm_offset_t) my_idlePTD != (vm_offset_t) vtophys(pdeaddr))
 					printf("pde mismatch: %x, %x\n", my_idlePTD, pdeaddr);
-				printf("cpuid: %d, pdeaddr: 0x%x\n", cpuid, pdeaddr);
+				printf("cpuid: %d, pdeaddr: 0x%x\n", PCPU_GET(cpuid), pdeaddr);
 				panic("pmap_enter: invalid kernel page table page(1), pdir=%p, npde=%p, pde=%p, va=%p\n",
 					pmap->pm_pdir[PTDPTDI], newpte, origpte, va);
 			}
@@ -2048,7 +2048,7 @@ pmap_enter(pmap_t pmap, vm_offset_t va, vm_page_t m, vm_prot_t prot,
 				*pte |= PG_RW;
 #ifdef SMP
 				cpu_invlpg((void *)va);
-				if (pmap->pm_active & other_cpus)
+				if (pmap->pm_active & PCPU_GET(other_cpus))
 					smp_invltlb();
 #else
 				invltlb_1pg(va);
@@ -2122,7 +2122,7 @@ validate:
 		/*if (origpte)*/ {
 #ifdef SMP
 			cpu_invlpg((void *)va);
-			if (pmap->pm_active & other_cpus)
+			if (pmap->pm_active & PCPU_GET(other_cpus))
 				smp_invltlb();
 #else
 			invltlb_1pg(va);
@@ -2666,20 +2666,20 @@ pmap_zero_page(phys)
 	vm_offset_t phys;
 {
 #ifdef SMP
-	if (*(int *) prv_CMAP3)
+	if (*(int *) PCPU_GET(prv_CMAP3))
 		panic("pmap_zero_page: prv_CMAP3 busy");
 
-	*(int *) prv_CMAP3 = PG_V | PG_RW | (phys & PG_FRAME) | PG_A | PG_M;
-	cpu_invlpg(prv_CADDR3);
+	*(int *) PCPU_GET(prv_CMAP3) = PG_V | PG_RW | (phys & PG_FRAME) | PG_A | PG_M;
+	cpu_invlpg(PCPU_GET(prv_CADDR3));
 
 #if defined(I686_CPU)
 	if (cpu_class == CPUCLASS_686)
-		i686_pagezero(prv_CADDR3);
+		i686_pagezero(PCPU_GET(prv_CADDR3));
 	else
 #endif
-		bzero(prv_CADDR3, PAGE_SIZE);
+		bzero(PCPU_GET(prv_CADDR3), PAGE_SIZE);
 
-	*(int *) prv_CMAP3 = 0;
+	*(int *) PCPU_GET(prv_CMAP3) = 0;
 #else
 	if (*(int *) CMAP2)
 		panic("pmap_zero_page: CMAP2 busy");
@@ -2710,20 +2710,20 @@ pmap_zero_page_area(phys, off, size)
 	int size;
 {
 #ifdef SMP
-	if (*(int *) prv_CMAP3)
+	if (*(int *) PCPU_GET(prv_CMAP3))
 		panic("pmap_zero_page: prv_CMAP3 busy");
 
-	*(int *) prv_CMAP3 = PG_V | PG_RW | (phys & PG_FRAME) | PG_A | PG_M;
-	cpu_invlpg(prv_CADDR3);
+	*(int *) PCPU_GET(prv_CMAP3) = PG_V | PG_RW | (phys & PG_FRAME) | PG_A | PG_M;
+	cpu_invlpg(PCPU_GET(prv_CADDR3));
 
 #if defined(I686_CPU)
 	if (cpu_class == CPUCLASS_686 && off == 0 && size == PAGE_SIZE)
-		i686_pagezero(prv_CADDR3);
+		i686_pagezero(PCPU_GET(prv_CADDR3));
 	else
 #endif
-		bzero((char *)prv_CADDR3 + off, size);
+		bzero((char *)PCPU_GET(prv_CADDR3) + off, size);
 
-	*(int *) prv_CMAP3 = 0;
+	*(int *) PCPU_GET(prv_CMAP3) = 0;
 #else
 	if (*(int *) CMAP2)
 		panic("pmap_zero_page: CMAP2 busy");
@@ -2753,21 +2753,22 @@ pmap_copy_page(src, dst)
 	vm_offset_t dst;
 {
 #ifdef SMP
-	if (*(int *) prv_CMAP1)
+	if (*(int *) PCPU_GET(prv_CMAP1))
 		panic("pmap_copy_page: prv_CMAP1 busy");
-	if (*(int *) prv_CMAP2)
+	if (*(int *) PCPU_GET(prv_CMAP2))
 		panic("pmap_copy_page: prv_CMAP2 busy");
 
-	*(int *) prv_CMAP1 = PG_V | (src & PG_FRAME) | PG_A;
-	*(int *) prv_CMAP2 = PG_V | PG_RW | (dst & PG_FRAME) | PG_A | PG_M;
+	*(int *) PCPU_GET(prv_CMAP1) = PG_V | (src & PG_FRAME) | PG_A;
+	*(int *) PCPU_GET(prv_CMAP2) = PG_V | PG_RW | (dst & PG_FRAME) | PG_A | PG_M;
 
-	cpu_invlpg(prv_CADDR1);
-	cpu_invlpg(prv_CADDR2);
+	cpu_invlpg(PCPU_GET(prv_CADDR1));
+	cpu_invlpg(PCPU_GET(prv_CADDR2));
 
-	bcopy(prv_CADDR1, prv_CADDR2, PAGE_SIZE);
+	bcopy(PCPU_GET(prv_CADDR1), PCPU_GET(prv_CADDR2), PAGE_SIZE);
+
+	*(int *) PCPU_GET(prv_CMAP1) = 0;
+	*(int *) PCPU_GET(prv_CMAP2) = 0;
 
-	*(int *) prv_CMAP1 = 0;
-	*(int *) prv_CMAP2 = 0;
 #else
 	if (*(int *) CMAP1 || *(int *) CMAP2)
 		panic("pmap_copy_page: CMAP busy");
@@ -3294,7 +3295,7 @@ pmap_activate(struct proc *p)
 
 	pmap = vmspace_pmap(p->p_vmspace);
 #if defined(SMP)
-	pmap->pm_active |= 1 << cpuid;
+	pmap->pm_active |= 1 << PCPU_GET(cpuid);
 #else
 	pmap->pm_active |= 1;
 #endif
diff --git a/sys/i386/i386/sys_machdep.c b/sys/i386/i386/sys_machdep.c
index d09d882..e242b73 100644
--- a/sys/i386/i386/sys_machdep.c
+++ b/sys/i386/i386/sys_machdep.c
@@ -252,12 +252,12 @@ set_user_ldt(struct pcb *pcb)
 {
 	struct pcb_ldt *pcb_ldt;
 
-	if (pcb != curpcb)
+	if (pcb != PCPU_GET(curpcb))
 		return;
 
 	pcb_ldt = pcb->pcb_ldt;
 #ifdef SMP
-	gdt[cpuid * NGDT + GUSERLDT_SEL].sd = pcb_ldt->ldt_sd;
+	gdt[PCPU_GET(cpuid) * NGDT + GUSERLDT_SEL].sd = pcb_ldt->ldt_sd;
 #else
 	gdt[GUSERLDT_SEL].sd = pcb_ldt->ldt_sd;
 #endif
@@ -308,7 +308,7 @@ user_ldt_free(struct pcb *pcb)
 	if (pcb_ldt == NULL)
 		return;
 
-	if (pcb == curpcb) {
+	if (pcb == PCPU_GET(curpcb)) {
 		lldt(_default_ldt);
 		PCPU_SET(currentldt, _default_ldt);
 	}
diff --git a/sys/i386/i386/trap.c b/sys/i386/i386/trap.c
index db02d51..959c64f 100644
--- a/sys/i386/i386/trap.c
+++ b/sys/i386/i386/trap.c
@@ -480,7 +480,7 @@ restart:
 			if (in_vm86call)
 				break;
 
-			if (intr_nesting_level != 0)
+			if (PCPU_GET(intr_nesting_level) != 0)
 				break;
 
 			/*
@@ -493,7 +493,7 @@ restart:
 			 * a signal.
 			 */
 			if (frame.tf_eip == (int)cpu_switch_load_gs) {
-				curpcb->pcb_gs = 0;
+				PCPU_GET(curpcb)->pcb_gs = 0;
 				psignal(p, SIGBUS);
 				goto out;
 			}
@@ -519,13 +519,15 @@ restart:
 			if (frame.tf_eip == (int)doreti_popl_es) {
 				frame.tf_eip = (int)doreti_popl_es_fault;
 				goto out;
-				}
+			}
 			if (frame.tf_eip == (int)doreti_popl_fs) {
 				frame.tf_eip = (int)doreti_popl_fs_fault;
 				goto out;
 			}
-			if (curpcb && curpcb->pcb_onfault) {
-				frame.tf_eip = (int)curpcb->pcb_onfault;
+			if (PCPU_GET(curpcb) != NULL &&
+			    PCPU_GET(curpcb)->pcb_onfault != NULL) {
+				frame.tf_eip =
+				    (int)PCPU_GET(curpcb)->pcb_onfault;
 				goto out;
 			}
 			break;
@@ -685,8 +687,9 @@ trap_pfault(frame, usermode, eva)
 
 		if (p == NULL ||
 		    (!usermode && va < VM_MAXUSER_ADDRESS &&
-		     (intr_nesting_level != 0 || curpcb == NULL ||
-		      curpcb->pcb_onfault == NULL))) {
+		     (PCPU_GET(intr_nesting_level) != 0 ||
+		      PCPU_GET(curpcb) == NULL ||
+		      PCPU_GET(curpcb)->pcb_onfault == NULL))) {
 			trap_fatal(frame, eva);
 			return (-1);
 		}
@@ -748,8 +751,10 @@ trap_pfault(frame, usermode, eva)
 		return (0);
 nogo:
 	if (!usermode) {
-		if (intr_nesting_level == 0 && curpcb && curpcb->pcb_onfault) {
-			frame->tf_eip = (int)curpcb->pcb_onfault;
+		if (PCPU_GET(intr_nesting_level) == 0 &&
+		    PCPU_GET(curpcb) != NULL &&
+		    PCPU_GET(curpcb)->pcb_onfault != NULL) {
+			frame->tf_eip = (int)PCPU_GET(curpcb)->pcb_onfault;
 			return (0);
 		}
 		trap_fatal(frame, eva);
@@ -853,8 +858,10 @@ trap_pfault(frame, usermode, eva)
 		return (0);
 nogo:
 	if (!usermode) {
-		if (intr_nesting_level == 0 && curpcb && curpcb->pcb_onfault) {
-			frame->tf_eip = (int)curpcb->pcb_onfault;
+		if (PCPU_GET(intr_nesting_level) == 0 &&
+		    PCPU_GET(curpcb) != NULL &&
+		    PCPU_GET(curpcb)->pcb_onfault != NULL) {
+			frame->tf_eip = (int)PCPU_GET(curpcb)->pcb_onfault;
 			return (0);
 		}
 		trap_fatal(frame, eva);
@@ -886,7 +893,7 @@ trap_fatal(frame, eva)
 			ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel");
 #ifdef SMP
 	/* two seperate prints in case of a trap on an unmapped page */
-	printf("cpuid = %d; ", cpuid);
+	printf("cpuid = %d; ", PCPU_GET(cpuid));
 	printf("lapic.id = %08x\n", lapic.id);
 #endif
 	if (type == T_PAGEFLT) {
@@ -964,12 +971,12 @@ void
 dblfault_handler()
 {
 	printf("\nFatal double fault:\n");
-	printf("eip = 0x%x\n", common_tss.tss_eip);
-	printf("esp = 0x%x\n", common_tss.tss_esp);
-	printf("ebp = 0x%x\n", common_tss.tss_ebp);
+	printf("eip = 0x%x\n", PCPU_GET(common_tss.tss_eip));
+	printf("esp = 0x%x\n", PCPU_GET(common_tss.tss_esp));
+	printf("ebp = 0x%x\n", PCPU_GET(common_tss.tss_ebp));
 #ifdef SMP
 	/* two seperate prints in case of a trap on an unmapped page */
-	printf("cpuid = %d; ", cpuid);
+	printf("cpuid = %d; ", PCPU_GET(cpuid));
 	printf("lapic.id = %08x\n", lapic.id);
 #endif
 	panic("double fault");
diff --git a/sys/i386/i386/vm86.c b/sys/i386/i386/vm86.c
index 08555b6..b081eef 100644
--- a/sys/i386/i386/vm86.c
+++ b/sys/i386/i386/vm86.c
@@ -143,9 +143,9 @@ vm86_emulate(vmf)
 	 * the extension is not present.  (This check should not be needed,
 	 * as we can't enter vm86 mode until we set up an extension area)
 	 */
-	if (curpcb->pcb_ext == 0)
+	if (PCPU_GET(curpcb)->pcb_ext == 0)
 		return (SIGBUS);
-	vm86 = &curpcb->pcb_ext->ext_vm86;
+	vm86 = &PCPU_GET(curpcb)->pcb_ext->ext_vm86;
 
 	if (vmf->vmf_eflags & PSL_T)
 		retcode = SIGTRAP;
@@ -507,7 +507,7 @@ static void
 vm86_initflags(struct vm86frame *vmf)
 {
 	int eflags = vmf->vmf_eflags;
-	struct vm86_kernel *vm86 = &curpcb->pcb_ext->ext_vm86;
+	struct vm86_kernel *vm86 = &PCPU_GET(curpcb)->pcb_ext->ext_vm86;
 
 	if (vm86->vm86_has_vme) {
 		eflags = (vmf->vmf_eflags & ~VME_USERCHANGE) |
diff --git a/sys/i386/i386/vm_machdep.c b/sys/i386/i386/vm_machdep.c
index 0a3bc51..552967c 100644
--- a/sys/i386/i386/vm_machdep.c
+++ b/sys/i386/i386/vm_machdep.c
@@ -143,7 +143,7 @@ cpu_fork(p1, p2, flags)
 
 #if NNPX > 0
 	/* Ensure that p1's pcb is up to date. */
-	if (npxproc == p1)
+	if (PCPU_GET(npxproc) == p1)
 		npxsave(&p1->p_addr->u_pcb.pcb_savefpu);
 #endif
 
@@ -442,23 +442,23 @@ cpu_reset()
 
 		u_int map;
 		int cnt;
-		printf("cpu_reset called on cpu#%d\n",cpuid);
+		printf("cpu_reset called on cpu#%d\n", PCPU_GET(cpuid));
 
-		map = other_cpus & ~ stopped_cpus;
+		map = PCPU_GET(other_cpus) & ~ stopped_cpus;
 
 		if (map != 0) {
 			printf("cpu_reset: Stopping other CPUs\n");
 			stop_cpus(map);		/* Stop all other CPUs */
 		}
 
-		if (cpuid == 0) {
+		if (PCPU_GET(cpuid) == 0) {
 			DELAY(1000000);
 			cpu_reset_real();
 			/* NOTREACHED */
 		} else {
 			/* We are not BSP (CPU #0) */
 
-			cpu_reset_proxyid = cpuid;
+			cpu_reset_proxyid = PCPU_GET(cpuid);
 			cpustop_restartfunc = cpu_reset_proxy;
 			cpu_reset_proxy_active = 0;
 			printf("cpu_reset: Restarting BSP\n");
diff --git a/sys/i386/include/cpu.h b/sys/i386/include/cpu.h
index 4c6cb43..6c4162b 100644
--- a/sys/i386/include/cpu.h
+++ b/sys/i386/include/cpu.h
@@ -62,7 +62,7 @@
 #define	CLKF_USERMODE(framep) \
 	((ISPL((framep)->cf_cs) == SEL_UPL) || (framep->cf_eflags & PSL_VM))
 
-#define CLKF_INTR(framep)	(intr_nesting_level >= 2)
+#define CLKF_INTR(framep)	(PCPU_GET(intr_nesting_level) >= 2)
 #define	CLKF_PC(framep)		((framep)->cf_eip)
 
 /*
@@ -82,7 +82,7 @@
 #define	need_resched() do {						\
 	PCPU_SET(astpending, AST_RESCHED|AST_PENDING);			\
 } while (0)
-#define	resched_wanted()	(astpending & AST_RESCHED)
+#define	resched_wanted()	(PCPU_GET(astpending) & AST_RESCHED)
 
 /*
  * Arrange to handle pending profiling ticks before returning to user mode.
@@ -105,7 +105,7 @@
  */
 #define	signotify(p)	aston()
 #define	aston() do {							\
-	PCPU_SET(astpending, astpending | AST_PENDING);			\
+	PCPU_SET(astpending, PCPU_GET(astpending) | AST_PENDING);	\
 } while (0)
 #define astoff()
 
diff --git a/sys/i386/include/mptable.h b/sys/i386/include/mptable.h
index 6c3abeb..79fc383 100644
--- a/sys/i386/include/mptable.h
+++ b/sys/i386/include/mptable.h
@@ -484,11 +484,11 @@ init_secondary(void)
 
 	gsel_tss = GSEL(GPROC0_SEL, SEL_KPL);
 	gdt[myid * NGDT + GPROC0_SEL].sd.sd_type = SDT_SYS386TSS;
-	common_tss.tss_esp0 = 0;	/* not used until after switch */
-	common_tss.tss_ss0 = GSEL(GDATA_SEL, SEL_KPL);
-	common_tss.tss_ioopt = (sizeof common_tss) << 16;
-	tss_gdt = &gdt[myid * NGDT + GPROC0_SEL].sd;
-	common_tssd = *tss_gdt;
+	PCPU_SET(common_tss.tss_esp0, 0); /* not used until after switch */
+	PCPU_SET(common_tss.tss_ss0, GSEL(GDATA_SEL, SEL_KPL));
+	PCPU_SET(common_tss.tss_ioopt, (sizeof (struct i386tss)) << 16);
+	PCPU_SET(tss_gdt, &gdt[myid * NGDT + GPROC0_SEL].sd);
+	PCPU_SET(common_tssd, *PCPU_GET(tss_gdt));
 	ltr(gsel_tss);
 
 	pmap_set_opt();
@@ -2045,7 +2045,7 @@ start_all_aps(u_int boot_addr)
 	}
 
 	/* build our map of 'other' CPUs */
-	other_cpus = all_cpus & ~(1 << cpuid);
+	PCPU_SET(other_cpus, all_cpus & ~(1 << PCPU_GET(cpuid)));
 
 	/* fill in our (BSP) APIC version */
 	cpu_apic_versions[0] = lapic.version;
@@ -2398,9 +2398,9 @@ ap_init(void)
 #endif
 
 	/* Build our map of 'other' CPUs. */
-	other_cpus = all_cpus & ~(1 << cpuid);
+	PCPU_SET(other_cpus, all_cpus & ~(1 << PCPU_GET(cpuid)));
 
-	printf("SMP: AP CPU #%d Launched!\n", cpuid);
+	printf("SMP: AP CPU #%d Launched!\n", PCPU_GET(cpuid));
 
 	/* set up CPU registers and state */
 	cpu_setregs();
@@ -2410,8 +2410,8 @@ ap_init(void)
 
 	/* A quick check from sanity claus */
 	apic_id = (apic_id_to_logical[(lapic.id & 0x0f000000) >> 24]);
-	if (cpuid != apic_id) {
-		printf("SMP: cpuid = %d\n", cpuid);
+	if (PCPU_GET(cpuid) != apic_id) {
+		printf("SMP: cpuid = %d\n", PCPU_GET(cpuid));
 		printf("SMP: apic_id = %d\n", apic_id);
 		printf("PTD[MPPTDI] = %p\n", (void *)PTD[MPPTDI]);
 		panic("cpuid mismatch! boom!!");
@@ -2445,10 +2445,10 @@ ap_init(void)
 	 * Set curproc to our per-cpu idleproc so that mutexes have
 	 * something unique to lock with.
 	 */
-	PCPU_SET(curproc,idleproc);
+	PCPU_SET(curproc, PCPU_GET(idleproc));
 
-	microuptime(&switchtime);
-	switchticks = ticks;
+	microuptime(PCPU_PTR(switchtime));
+	PCPU_SET(switchticks, ticks);
 
 	/* ok, now grab sched_lock and enter the scheduler */
 	enable_intr();
@@ -2610,7 +2610,7 @@ forward_statclock(int pscnt)
 
 	/* Step 1: Probe state   (user, cpu, interrupt, spinlock, idle ) */
 	
-	map = other_cpus & ~stopped_cpus ;
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 	checkstate_probed_cpus = 0;
 	if (map != 0)
 		selected_apic_ipi(map,
@@ -2636,7 +2636,7 @@ forward_statclock(int pscnt)
 	
 	map = 0;
 	for (id = 0; id < mp_ncpus; id++) {
-		if (id == cpuid)
+		if (id == PCPU_GET(cpuid))
 			continue;
 		if (((1 << id) & checkstate_probed_cpus) == 0)
 			continue;
@@ -2685,7 +2685,7 @@ forward_hardclock(int pscnt)
 
 	/* Step 1: Probe state   (user, cpu, interrupt, spinlock, idle) */
 	
-	map = other_cpus & ~stopped_cpus ;
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 	checkstate_probed_cpus = 0;
 	if (map != 0)
 		selected_apic_ipi(map,
@@ -2712,7 +2712,7 @@ forward_hardclock(int pscnt)
 	
 	map = 0;
 	for (id = 0; id < mp_ncpus; id++) {
-		if (id == cpuid)
+		if (id == PCPU_GET(cpuid))
 			continue;
 		if (((1 << id) & checkstate_probed_cpus) == 0)
 			continue;
@@ -2813,8 +2813,8 @@ forward_roundrobin(void)
 		return;
 	if (!forward_roundrobin_enabled)
 		return;
-	resched_cpus |= other_cpus;
-	map = other_cpus & ~stopped_cpus ;
+	resched_cpus |= PCPU_GET(other_cpus);
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 #if 1
 	selected_apic_ipi(map, XCPUAST_OFFSET, APIC_DELMODE_FIXED);
 #else
diff --git a/sys/i386/isa/npx.c b/sys/i386/isa/npx.c
index 2672a63..b79d511 100644
--- a/sys/i386/isa/npx.c
+++ b/sys/i386/isa/npx.c
@@ -493,8 +493,8 @@ npxinit(control)
 	npxsave(&dummy);
 	stop_emulating();
 	fldcw(&control);
-	if (curpcb != NULL)
-		fnsave(&curpcb->pcb_savefpu);
+	if (PCPU_GET(curpcb) != NULL)
+		fnsave(&PCPU_GET(curpcb)->pcb_savefpu);
 	start_emulating();
 }
 
@@ -506,14 +506,14 @@ npxexit(p)
 	struct proc *p;
 {
 
-	if (p == npxproc)
-		npxsave(&curpcb->pcb_savefpu);
+	if (p == PCPU_GET(npxproc))
+		npxsave(&PCPU_GET(curpcb)->pcb_savefpu);
 #ifdef NPX_DEBUG
 	if (npx_exists) {
 		u_int	masked_exceptions;
 
-		masked_exceptions = curpcb->pcb_savefpu.sv_env.en_cw
-				    & curpcb->pcb_savefpu.sv_env.en_sw & 0x7f;
+		masked_exceptions = PCPU_GET(curpcb)->pcb_savefpu.sv_env.en_cw
+		    &PCPU_GET(curpcb)->pcb_savefpu.sv_env.en_sw & 0x7f;
 		/*
 		 * Log exceptions that would have trapped with the old
 		 * control word (overflow, divide by 0, and invalid operand).
@@ -722,19 +722,19 @@ npx_intr(dummy)
 	u_short control;
 	struct intrframe *frame;
 
-	if (npxproc == NULL || !npx_exists) {
+	if (PCPU_GET(npxproc) == NULL || !npx_exists) {
 		printf("npxintr: npxproc = %p, curproc = %p, npx_exists = %d\n",
-		       npxproc, curproc, npx_exists);
+		       PCPU_GET(npxproc), curproc, npx_exists);
 		panic("npxintr from nowhere");
 	}
-	if (npxproc != curproc) {
+	if (PCPU_GET(npxproc) != curproc) {
 		printf("npxintr: npxproc = %p, curproc = %p, npx_exists = %d\n",
-		       npxproc, curproc, npx_exists);
+		       PCPU_GET(npxproc), curproc, npx_exists);
 		panic("npxintr from non-current process");
 	}
 
 	outb(0xf0, 0);
-	fnstsw(&curpcb->pcb_savefpu.sv_ex_sw);
+	fnstsw(&PCPU_GET(curpcb)->pcb_savefpu.sv_ex_sw);
 	fnstcw(&control);
 	fnclex();
 
@@ -760,8 +760,8 @@ npx_intr(dummy)
 		 * this exception.
 		 */
 		code = 
-		    fpetable[(curpcb->pcb_savefpu.sv_ex_sw & ~control & 0x3f) |
-			(curpcb->pcb_savefpu.sv_ex_sw & 0x40)];
+		    fpetable[(PCPU_GET(curpcb)->pcb_savefpu.sv_ex_sw & ~control & 0x3f) |
+			(PCPU_GET(curpcb)->pcb_savefpu.sv_ex_sw & 0x40)];
 		trapsignal(curproc, SIGFPE, code);
 	} else {
 		/*
@@ -794,9 +794,9 @@ npxdna()
 {
 	if (!npx_exists)
 		return (0);
-	if (npxproc != NULL) {
+	if (PCPU_GET(npxproc) != NULL) {
 		printf("npxdna: npxproc = %p, curproc = %p\n",
-		       npxproc, curproc);
+		       PCPU_GET(npxproc), curproc);
 		panic("npxdna");
 	}
 	stop_emulating();
@@ -804,7 +804,7 @@ npxdna()
 	 * Record new context early in case frstor causes an IRQ13.
 	 */
 	PCPU_SET(npxproc, CURPROC);
-	curpcb->pcb_savefpu.sv_ex_sw = 0;
+	PCPU_GET(curpcb)->pcb_savefpu.sv_ex_sw = 0;
 	/*
 	 * The following frstor may cause an IRQ13 when the state being
 	 * restored has a pending error.  The error will appear to have been
@@ -817,7 +817,7 @@ npxdna()
 	 * fnsave are broken, so our treatment breaks fnclex if it is the
 	 * first FPU instruction after a context switch.
 	 */
-	frstor(&curpcb->pcb_savefpu);
+	frstor(&PCPU_GET(curpcb)->pcb_savefpu);
 
 	return (1);
 }
diff --git a/sys/ia64/ia64/interrupt.c b/sys/ia64/ia64/interrupt.c
index 16fbfdb..698b46c 100644
--- a/sys/ia64/ia64/interrupt.c
+++ b/sys/ia64/ia64/interrupt.c
@@ -78,7 +78,7 @@ static u_int schedclk2;
 void
 interrupt(u_int64_t vector, struct trapframe *framep)
 {
-	atomic_add_int(&PCPU_GET(intr_nesting_level), 1);
+	atomic_add_int(PCPU_PTR(intr_nesting_level), 1);
 
 	switch (vector) {
 	case 240:	/* clock interrupt */
@@ -104,7 +104,7 @@ interrupt(u_int64_t vector, struct trapframe *framep)
 		panic("unexpected interrupt: vec %ld\n", vector);
 		/* NOTREACHED */
 	}
-	atomic_subtract_int(&PCPU_GET(intr_nesting_level), 1);
+	atomic_subtract_int(PCPU_PTR(intr_nesting_level), 1);
 }
 
 
diff --git a/sys/ia64/ia64/mp_machdep.c b/sys/ia64/ia64/mp_machdep.c
index 371b499..486d0aa 100644
--- a/sys/ia64/ia64/mp_machdep.c
+++ b/sys/ia64/ia64/mp_machdep.c
@@ -337,7 +337,7 @@ forward_statclock(int pscnt)
 	
 	map = 0;
 	for (id = 0; id < mp_ncpus; id++) {
-		if (id == cpuid)
+		if (id == PCPU_GET(cpuid))
 			continue;
 		if (((1 << id) & checkstate_probed_cpus) == 0)
 			continue;
@@ -415,7 +415,7 @@ forward_hardclock(int pscnt)
 	
 	map = 0;
 	for (id = 0; id < mp_ncpus; id++) {
-		if (id == cpuid)
+		if (id == PCPU_GET(cpuid))
 			continue;
 		if (((1 << id) & checkstate_probed_cpus) == 0)
 			continue;
@@ -741,7 +741,7 @@ smp_handle_ipi(struct trapframe *frame)
 
 	do {
 		ipis = PCPU_GET(pending_ipis);
-	} while (atomic_cmpset_64(&PCPU_GET(pending_ipis), ipis, 0));
+	} while (atomic_cmpset_64(PCPU_PTR(pending_ipis), ipis, 0));
 
 	CTR1(KTR_SMP, "smp_handle_ipi(), ipis=%lx", ipis);
 	while (ipis) {
diff --git a/sys/kern/init_main.c b/sys/kern/init_main.c
index a147eae..24bf964 100644
--- a/sys/kern/init_main.c
+++ b/sys/kern/init_main.c
@@ -415,7 +415,7 @@ proc0_post(void *dummy __unused)
 		p->p_runtime = 0;
 	}
 	ALLPROC_LOCK(AP_RELEASE);
-	microuptime(&switchtime);
+	microuptime(PCPU_PTR(switchtime));
 	PCPU_SET(switchticks, ticks);
 
 	/*
diff --git a/sys/kern/kern_clock.c b/sys/kern/kern_clock.c
index 223be91..5a8ade6 100644
--- a/sys/kern/kern_clock.c
+++ b/sys/kern/kern_clock.c
@@ -160,7 +160,7 @@ hardclock(frame)
 	int need_softclock = 0;
 
 	p = curproc;
-	if (p != idleproc) {
+	if (p != PCPU_GET(idleproc)) {
 		register struct pstats *pstats;
 
 		/*
@@ -400,7 +400,7 @@ statclock(frame)
 			cp_time[CP_INTR]++;
 		} else {
 			p->p_sticks++;
-			if (p != idleproc)
+			if (p != PCPU_GET(idleproc))
 				cp_time[CP_SYS]++;
 			else
 				cp_time[CP_IDLE]++;
diff --git a/sys/kern/kern_exit.c b/sys/kern/kern_exit.c
index e6a640b..41b540e 100644
--- a/sys/kern/kern_exit.c
+++ b/sys/kern/kern_exit.c
@@ -305,8 +305,8 @@ exit1(p, rv)
 	 * directly.  Set it now so that the rest of the exit time gets
 	 * counted somewhere if possible.
 	 */
-	microuptime(&switchtime);
-	switchticks = ticks;
+	microuptime(PCPU_PTR(switchtime));
+	PCPU_SET(switchticks, ticks);
 
 	/*
 	 * notify interested parties of our demise.
diff --git a/sys/kern/kern_ktr.c b/sys/kern/kern_ktr.c
index 1704d5d..457eb02 100644
--- a/sys/kern/kern_ktr.c
+++ b/sys/kern/kern_ktr.c
@@ -61,7 +61,7 @@
 #endif
 
 #ifdef SMP
-#define KTR_CPU		cpuid
+#define KTR_CPU		PCPU_GET(cpuid)
 #else
 #define KTR_CPU		0
 #endif
diff --git a/sys/kern/kern_malloc.c b/sys/kern/kern_malloc.c
index f268be6..d6275b8 100644
--- a/sys/kern/kern_malloc.c
+++ b/sys/kern/kern_malloc.c
@@ -147,7 +147,7 @@ malloc(size, type, flags)
 
 #if defined(INVARIANTS) && defined(__i386__)
 	if (flags == M_WAITOK)
-		KASSERT(intr_nesting_level == 0,
+		KASSERT(PCPU_GET(intr_nesting_level) == 0,
 		   ("malloc(M_WAITOK) in interrupt context"));
 #endif
 	indx = BUCKETINDX(size);
diff --git a/sys/kern/kern_mutex.c b/sys/kern/kern_mutex.c
index 77fbdc1..a79299c 100644
--- a/sys/kern/kern_mutex.c
+++ b/sys/kern/kern_mutex.c
@@ -844,7 +844,7 @@ witness_enter(struct mtx *m, int flags, const char *file, int line)
 		if (m->mtx_recurse != 0)
 			return;
 		mtx_enter(&w_mtx, MTX_SPIN | MTX_QUIET);
-		i = witness_spin_check;
+		i = PCPU_GET(witness_spin_check);
 		if (i != 0 && w->w_level < i) {
 			mtx_exit(&w_mtx, MTX_SPIN | MTX_QUIET);
 			panic("mutex_enter(%s:%x, MTX_SPIN) out of order @"
@@ -977,7 +977,8 @@ witness_exit(struct mtx *m, int flags, const char *file, int line)
 		if (m->mtx_recurse != 0)
 			return;
 		mtx_enter(&w_mtx, MTX_SPIN | MTX_QUIET);
-		PCPU_SET(witness_spin_check, witness_spin_check & ~w->w_level);
+		PCPU_SET(witness_spin_check,
+		    PCPU_GET(witness_spin_check) & ~w->w_level);
 		mtx_exit(&w_mtx, MTX_SPIN | MTX_QUIET);
 		return;
 	}
@@ -1011,7 +1012,8 @@ witness_try_enter(struct mtx *m, int flags, const char *file, int line)
 		if (m->mtx_recurse != 0)
 			return;
 		mtx_enter(&w_mtx, MTX_SPIN | MTX_QUIET);
-		PCPU_SET(witness_spin_check, witness_spin_check | w->w_level);
+		PCPU_SET(witness_spin_check,
+		    PCPU_GET(witness_spin_check) | w->w_level);
 		mtx_exit(&w_mtx, MTX_SPIN | MTX_QUIET);
 		w->w_file = file;
 		w->w_line = line;
diff --git a/sys/kern/kern_resource.c b/sys/kern/kern_resource.c
index b852cf6..e67ee79 100644
--- a/sys/kern/kern_resource.c
+++ b/sys/kern/kern_resource.c
@@ -534,13 +534,14 @@ calcru(p, up, sp, ip)
 		 * quantum, which is much greater than the sampling error.
 		 */
 		microuptime(&tv);
-		if (timevalcmp(&tv, &switchtime, <))
+		if (timevalcmp(&tv, PCPU_PTR(switchtime), <))
 			printf("microuptime() went backwards (%ld.%06ld -> %ld.%06ld)\n",
-			    switchtime.tv_sec, switchtime.tv_usec,
+			    PCPU_GET(switchtime.tv_sec), PCPU_GET(switchtime.tv_usec),
 			    tv.tv_sec, tv.tv_usec);
 		else
-			tu += (tv.tv_usec - switchtime.tv_usec) +
-			    (tv.tv_sec - switchtime.tv_sec) * (int64_t)1000000;
+			tu += (tv.tv_usec - PCPU_GET(switchtime.tv_usec)) +
+			    (tv.tv_sec - PCPU_GET(switchtime.tv_sec)) *
+			    (int64_t)1000000;
 	}
 	ptu = p->p_uu + p->p_su + p->p_iu;
 	if (tu < ptu || (int64_t)tu < 0) {
diff --git a/sys/kern/kern_shutdown.c b/sys/kern/kern_shutdown.c
index d1e23f8..eaf8b94 100644
--- a/sys/kern/kern_shutdown.c
+++ b/sys/kern/kern_shutdown.c
@@ -209,7 +209,7 @@ boot(int howto)
 
 #ifdef SMP
 	if (smp_active)
-		printf("boot() called on cpu#%d\n", cpuid);
+		printf("boot() called on cpu#%d\n", PCPU_GET(cpuid));
 #endif
 	/*
 	 * Do any callouts that should be done BEFORE syncing the filesystems.
@@ -557,7 +557,7 @@ panic(const char *fmt, ...)
 	printf("panic: %s\n", buf);
 #ifdef SMP
 	/* two seperate prints in case of an unmapped page and trap */
-	printf("cpuid = %d; ", cpuid);
+	printf("cpuid = %d; ", PCPU_GET(cpuid));
 #ifdef APIC_IO
 	printf("lapic.id = %08x\n", lapic.id);
 #endif
diff --git a/sys/kern/kern_subr.c b/sys/kern/kern_subr.c
index c379f79..c3d7849 100644
--- a/sys/kern/kern_subr.c
+++ b/sys/kern/kern_subr.c
@@ -92,7 +92,7 @@ uiomove(cp, n, uio)
 
 		case UIO_USERSPACE:
 		case UIO_USERISPACE:
-			if (ticks - switchticks >= hogticks)
+			if (ticks - PCPU_GET(switchticks) >= hogticks)
 				uio_yield();
 			if (uio->uio_rw == UIO_READ)
 				error = copyout(cp, iov->iov_base, cnt);
@@ -154,7 +154,7 @@ uiomoveco(cp, n, uio, obj)
 
 		case UIO_USERSPACE:
 		case UIO_USERISPACE:
-			if (ticks - switchticks >= hogticks)
+			if (ticks - PCPU_GET(switchticks) >= hogticks)
 				uio_yield();
 			if (uio->uio_rw == UIO_READ) {
 #ifdef ENABLE_VFS_IOOPT
@@ -236,7 +236,7 @@ uioread(n, uio, obj, nread)
 
 			cnt &= ~PAGE_MASK;
 
-			if (ticks - switchticks >= hogticks)
+			if (ticks - PCPU_GET(switchticks) >= hogticks)
 				uio_yield();
 			error = vm_uiomove(&curproc->p_vmspace->vm_map, obj,
 						uio->uio_offset, cnt,
diff --git a/sys/kern/kern_switch.c b/sys/kern/kern_switch.c
index 60024b0..7515ea8 100644
--- a/sys/kern/kern_switch.c
+++ b/sys/kern/kern_switch.c
@@ -232,12 +232,12 @@ chooseproc(void)
 	} else {
 		CTR1(KTR_PROC, "chooseproc: idleproc, schedlock %lx",
 			(long)sched_lock.mtx_lock);
-		return idleproc;
+		return PCPU_GET(idleproc);
 	}
 	p = TAILQ_FIRST(q);
 #ifdef SMP
 	/* wander down the current run queue for this pri level for a match */
-	id = cpuid;
+	id = PCPU_GET(cpuid);
 	while (p->p_lastcpu != id) {
 		p = TAILQ_NEXT(p, p_procq);
 		if (p == NULL) {
diff --git a/sys/kern/kern_synch.c b/sys/kern/kern_synch.c
index 90b9fa3..ff240db 100644
--- a/sys/kern/kern_synch.c
+++ b/sys/kern/kern_synch.c
@@ -142,7 +142,7 @@ maybe_resched(chk)
 	 * standard process becomes runaway cpu-bound, the system can lockup
 	 * due to idle-scheduler processes in wakeup never getting any cpu.
 	 */
-	if (p == idleproc) {
+	if (p == PCPU_GET(idleproc)) {
 #if 0
 		need_resched();
 #endif
@@ -176,7 +176,7 @@ roundrobin(arg)
 	need_resched();
 	forward_roundrobin();
 #else 
- 	if (p == idleproc || RTP_PRIO_NEED_RR(p->p_rtprio.type))
+ 	if (p == PCPU_GET(idleproc) || RTP_PRIO_NEED_RR(p->p_rtprio.type))
  		need_resched();
 #endif
 
@@ -925,14 +925,15 @@ mi_switch()
 	 * process was running, and add that to its total so far.
 	 */
 	microuptime(&new_switchtime);
-	if (timevalcmp(&new_switchtime, &switchtime, <)) {
+	if (timevalcmp(&new_switchtime, PCPU_PTR(switchtime), <)) {
 		printf("microuptime() went backwards (%ld.%06ld -> %ld.%06ld)\n",
-		    switchtime.tv_sec, switchtime.tv_usec,
+		    PCPU_GET(switchtime.tv_sec), PCPU_GET(switchtime.tv_usec),
 		    new_switchtime.tv_sec, new_switchtime.tv_usec);
-		new_switchtime = switchtime;
+		new_switchtime = PCPU_GET(switchtime);
 	} else {
-		p->p_runtime += (new_switchtime.tv_usec - switchtime.tv_usec) +
-		    (new_switchtime.tv_sec - switchtime.tv_sec) * (int64_t)1000000;
+		p->p_runtime += (new_switchtime.tv_usec - PCPU_GET(switchtime.tv_usec)) +
+		    (new_switchtime.tv_sec - PCPU_GET(switchtime.tv_sec)) *
+		    (int64_t)1000000;
 	}
 
 	/*
@@ -959,15 +960,15 @@ mi_switch()
 	 * Pick a new current process and record its start time.
 	 */
 	cnt.v_swtch++;
-	switchtime = new_switchtime;
+	PCPU_SET(switchtime, new_switchtime);
 	CTR4(KTR_PROC, "mi_switch: old proc %p (pid %d, %s), schedlock %p",
 		p, p->p_pid, p->p_comm, (void *) sched_lock.mtx_lock);
 	cpu_switch();
 	CTR4(KTR_PROC, "mi_switch: new proc %p (pid %d, %s), schedlock %p",
 		p, p->p_pid, p->p_comm, (void *) sched_lock.mtx_lock);
-	if (switchtime.tv_sec == 0)
-		microuptime(&switchtime);
-	switchticks = ticks;
+	if (PCPU_GET(switchtime.tv_sec) == 0)
+		microuptime(PCPU_PTR(switchtime));
+	PCPU_SET(switchticks, ticks);
 	splx(x);
 }
 
diff --git a/sys/kern/subr_prf.c b/sys/kern/subr_prf.c
index efff7f8..86a53766 100644
--- a/sys/kern/subr_prf.c
+++ b/sys/kern/subr_prf.c
@@ -113,7 +113,7 @@ uprintf(const char *fmt, ...)
 	struct putchar_arg pca;
 	int retval = 0;
 
-	if (p && p != idleproc && p->p_flag & P_CONTROLT &&
+	if (p && p != PCPU_GET(idleproc) && p->p_flag & P_CONTROLT &&
 	    p->p_session->s_ttyvp) {
 		va_start(ap, fmt);
 		pca.tty = p->p_session->s_ttyp;
diff --git a/sys/kern/subr_smp.c b/sys/kern/subr_smp.c
index 6c3abeb..79fc383 100644
--- a/sys/kern/subr_smp.c
+++ b/sys/kern/subr_smp.c
@@ -484,11 +484,11 @@ init_secondary(void)
 
 	gsel_tss = GSEL(GPROC0_SEL, SEL_KPL);
 	gdt[myid * NGDT + GPROC0_SEL].sd.sd_type = SDT_SYS386TSS;
-	common_tss.tss_esp0 = 0;	/* not used until after switch */
-	common_tss.tss_ss0 = GSEL(GDATA_SEL, SEL_KPL);
-	common_tss.tss_ioopt = (sizeof common_tss) << 16;
-	tss_gdt = &gdt[myid * NGDT + GPROC0_SEL].sd;
-	common_tssd = *tss_gdt;
+	PCPU_SET(common_tss.tss_esp0, 0); /* not used until after switch */
+	PCPU_SET(common_tss.tss_ss0, GSEL(GDATA_SEL, SEL_KPL));
+	PCPU_SET(common_tss.tss_ioopt, (sizeof (struct i386tss)) << 16);
+	PCPU_SET(tss_gdt, &gdt[myid * NGDT + GPROC0_SEL].sd);
+	PCPU_SET(common_tssd, *PCPU_GET(tss_gdt));
 	ltr(gsel_tss);
 
 	pmap_set_opt();
@@ -2045,7 +2045,7 @@ start_all_aps(u_int boot_addr)
 	}
 
 	/* build our map of 'other' CPUs */
-	other_cpus = all_cpus & ~(1 << cpuid);
+	PCPU_SET(other_cpus, all_cpus & ~(1 << PCPU_GET(cpuid)));
 
 	/* fill in our (BSP) APIC version */
 	cpu_apic_versions[0] = lapic.version;
@@ -2398,9 +2398,9 @@ ap_init(void)
 #endif
 
 	/* Build our map of 'other' CPUs. */
-	other_cpus = all_cpus & ~(1 << cpuid);
+	PCPU_SET(other_cpus, all_cpus & ~(1 << PCPU_GET(cpuid)));
 
-	printf("SMP: AP CPU #%d Launched!\n", cpuid);
+	printf("SMP: AP CPU #%d Launched!\n", PCPU_GET(cpuid));
 
 	/* set up CPU registers and state */
 	cpu_setregs();
@@ -2410,8 +2410,8 @@ ap_init(void)
 
 	/* A quick check from sanity claus */
 	apic_id = (apic_id_to_logical[(lapic.id & 0x0f000000) >> 24]);
-	if (cpuid != apic_id) {
-		printf("SMP: cpuid = %d\n", cpuid);
+	if (PCPU_GET(cpuid) != apic_id) {
+		printf("SMP: cpuid = %d\n", PCPU_GET(cpuid));
 		printf("SMP: apic_id = %d\n", apic_id);
 		printf("PTD[MPPTDI] = %p\n", (void *)PTD[MPPTDI]);
 		panic("cpuid mismatch! boom!!");
@@ -2445,10 +2445,10 @@ ap_init(void)
 	 * Set curproc to our per-cpu idleproc so that mutexes have
 	 * something unique to lock with.
 	 */
-	PCPU_SET(curproc,idleproc);
+	PCPU_SET(curproc, PCPU_GET(idleproc));
 
-	microuptime(&switchtime);
-	switchticks = ticks;
+	microuptime(PCPU_PTR(switchtime));
+	PCPU_SET(switchticks, ticks);
 
 	/* ok, now grab sched_lock and enter the scheduler */
 	enable_intr();
@@ -2610,7 +2610,7 @@ forward_statclock(int pscnt)
 
 	/* Step 1: Probe state   (user, cpu, interrupt, spinlock, idle ) */
 	
-	map = other_cpus & ~stopped_cpus ;
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 	checkstate_probed_cpus = 0;
 	if (map != 0)
 		selected_apic_ipi(map,
@@ -2636,7 +2636,7 @@ forward_statclock(int pscnt)
 	
 	map = 0;
 	for (id = 0; id < mp_ncpus; id++) {
-		if (id == cpuid)
+		if (id == PCPU_GET(cpuid))
 			continue;
 		if (((1 << id) & checkstate_probed_cpus) == 0)
 			continue;
@@ -2685,7 +2685,7 @@ forward_hardclock(int pscnt)
 
 	/* Step 1: Probe state   (user, cpu, interrupt, spinlock, idle) */
 	
-	map = other_cpus & ~stopped_cpus ;
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 	checkstate_probed_cpus = 0;
 	if (map != 0)
 		selected_apic_ipi(map,
@@ -2712,7 +2712,7 @@ forward_hardclock(int pscnt)
 	
 	map = 0;
 	for (id = 0; id < mp_ncpus; id++) {
-		if (id == cpuid)
+		if (id == PCPU_GET(cpuid))
 			continue;
 		if (((1 << id) & checkstate_probed_cpus) == 0)
 			continue;
@@ -2813,8 +2813,8 @@ forward_roundrobin(void)
 		return;
 	if (!forward_roundrobin_enabled)
 		return;
-	resched_cpus |= other_cpus;
-	map = other_cpus & ~stopped_cpus ;
+	resched_cpus |= PCPU_GET(other_cpus);
+	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
 #if 1
 	selected_apic_ipi(map, XCPUAST_OFFSET, APIC_DELMODE_FIXED);
 #else
diff --git a/sys/kern/subr_trap.c b/sys/kern/subr_trap.c
index db02d51..959c64f 100644
--- a/sys/kern/subr_trap.c
+++ b/sys/kern/subr_trap.c
@@ -480,7 +480,7 @@ restart:
 			if (in_vm86call)
 				break;
 
-			if (intr_nesting_level != 0)
+			if (PCPU_GET(intr_nesting_level) != 0)
 				break;
 
 			/*
@@ -493,7 +493,7 @@ restart:
 			 * a signal.
 			 */
 			if (frame.tf_eip == (int)cpu_switch_load_gs) {
-				curpcb->pcb_gs = 0;
+				PCPU_GET(curpcb)->pcb_gs = 0;
 				psignal(p, SIGBUS);
 				goto out;
 			}
@@ -519,13 +519,15 @@ restart:
 			if (frame.tf_eip == (int)doreti_popl_es) {
 				frame.tf_eip = (int)doreti_popl_es_fault;
 				goto out;
-				}
+			}
 			if (frame.tf_eip == (int)doreti_popl_fs) {
 				frame.tf_eip = (int)doreti_popl_fs_fault;
 				goto out;
 			}
-			if (curpcb && curpcb->pcb_onfault) {
-				frame.tf_eip = (int)curpcb->pcb_onfault;
+			if (PCPU_GET(curpcb) != NULL &&
+			    PCPU_GET(curpcb)->pcb_onfault != NULL) {
+				frame.tf_eip =
+				    (int)PCPU_GET(curpcb)->pcb_onfault;
 				goto out;
 			}
 			break;
@@ -685,8 +687,9 @@ trap_pfault(frame, usermode, eva)
 
 		if (p == NULL ||
 		    (!usermode && va < VM_MAXUSER_ADDRESS &&
-		     (intr_nesting_level != 0 || curpcb == NULL ||
-		      curpcb->pcb_onfault == NULL))) {
+		     (PCPU_GET(intr_nesting_level) != 0 ||
+		      PCPU_GET(curpcb) == NULL ||
+		      PCPU_GET(curpcb)->pcb_onfault == NULL))) {
 			trap_fatal(frame, eva);
 			return (-1);
 		}
@@ -748,8 +751,10 @@ trap_pfault(frame, usermode, eva)
 		return (0);
 nogo:
 	if (!usermode) {
-		if (intr_nesting_level == 0 && curpcb && curpcb->pcb_onfault) {
-			frame->tf_eip = (int)curpcb->pcb_onfault;
+		if (PCPU_GET(intr_nesting_level) == 0 &&
+		    PCPU_GET(curpcb) != NULL &&
+		    PCPU_GET(curpcb)->pcb_onfault != NULL) {
+			frame->tf_eip = (int)PCPU_GET(curpcb)->pcb_onfault;
 			return (0);
 		}
 		trap_fatal(frame, eva);
@@ -853,8 +858,10 @@ trap_pfault(frame, usermode, eva)
 		return (0);
 nogo:
 	if (!usermode) {
-		if (intr_nesting_level == 0 && curpcb && curpcb->pcb_onfault) {
-			frame->tf_eip = (int)curpcb->pcb_onfault;
+		if (PCPU_GET(intr_nesting_level) == 0 &&
+		    PCPU_GET(curpcb) != NULL &&
+		    PCPU_GET(curpcb)->pcb_onfault != NULL) {
+			frame->tf_eip = (int)PCPU_GET(curpcb)->pcb_onfault;
 			return (0);
 		}
 		trap_fatal(frame, eva);
@@ -886,7 +893,7 @@ trap_fatal(frame, eva)
 			ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel");
 #ifdef SMP
 	/* two seperate prints in case of a trap on an unmapped page */
-	printf("cpuid = %d; ", cpuid);
+	printf("cpuid = %d; ", PCPU_GET(cpuid));
 	printf("lapic.id = %08x\n", lapic.id);
 #endif
 	if (type == T_PAGEFLT) {
@@ -964,12 +971,12 @@ void
 dblfault_handler()
 {
 	printf("\nFatal double fault:\n");
-	printf("eip = 0x%x\n", common_tss.tss_eip);
-	printf("esp = 0x%x\n", common_tss.tss_esp);
-	printf("ebp = 0x%x\n", common_tss.tss_ebp);
+	printf("eip = 0x%x\n", PCPU_GET(common_tss.tss_eip));
+	printf("esp = 0x%x\n", PCPU_GET(common_tss.tss_esp));
+	printf("ebp = 0x%x\n", PCPU_GET(common_tss.tss_ebp));
 #ifdef SMP
 	/* two seperate prints in case of a trap on an unmapped page */
-	printf("cpuid = %d; ", cpuid);
+	printf("cpuid = %d; ", PCPU_GET(cpuid));
 	printf("lapic.id = %08x\n", lapic.id);
 #endif
 	panic("double fault");
diff --git a/sys/kern/subr_turnstile.c b/sys/kern/subr_turnstile.c
index 77fbdc1..a79299c 100644
--- a/sys/kern/subr_turnstile.c
+++ b/sys/kern/subr_turnstile.c
@@ -844,7 +844,7 @@ witness_enter(struct mtx *m, int flags, const char *file, int line)
 		if (m->mtx_recurse != 0)
 			return;
 		mtx_enter(&w_mtx, MTX_SPIN | MTX_QUIET);
-		i = witness_spin_check;
+		i = PCPU_GET(witness_spin_check);
 		if (i != 0 && w->w_level < i) {
 			mtx_exit(&w_mtx, MTX_SPIN | MTX_QUIET);
 			panic("mutex_enter(%s:%x, MTX_SPIN) out of order @"
@@ -977,7 +977,8 @@ witness_exit(struct mtx *m, int flags, const char *file, int line)
 		if (m->mtx_recurse != 0)
 			return;
 		mtx_enter(&w_mtx, MTX_SPIN | MTX_QUIET);
-		PCPU_SET(witness_spin_check, witness_spin_check & ~w->w_level);
+		PCPU_SET(witness_spin_check,
+		    PCPU_GET(witness_spin_check) & ~w->w_level);
 		mtx_exit(&w_mtx, MTX_SPIN | MTX_QUIET);
 		return;
 	}
@@ -1011,7 +1012,8 @@ witness_try_enter(struct mtx *m, int flags, const char *file, int line)
 		if (m->mtx_recurse != 0)
 			return;
 		mtx_enter(&w_mtx, MTX_SPIN | MTX_QUIET);
-		PCPU_SET(witness_spin_check, witness_spin_check | w->w_level);
+		PCPU_SET(witness_spin_check,
+		    PCPU_GET(witness_spin_check) | w->w_level);
 		mtx_exit(&w_mtx, MTX_SPIN | MTX_QUIET);
 		w->w_file = file;
 		w->w_line = line;
diff --git a/sys/kern/subr_witness.c b/sys/kern/subr_witness.c
index 77fbdc1..a79299c 100644
--- a/sys/kern/subr_witness.c
+++ b/sys/kern/subr_witness.c
@@ -844,7 +844,7 @@ witness_enter(struct mtx *m, int flags, const char *file, int line)
 		if (m->mtx_recurse != 0)
 			return;
 		mtx_enter(&w_mtx, MTX_SPIN | MTX_QUIET);
-		i = witness_spin_check;
+		i = PCPU_GET(witness_spin_check);
 		if (i != 0 && w->w_level < i) {
 			mtx_exit(&w_mtx, MTX_SPIN | MTX_QUIET);
 			panic("mutex_enter(%s:%x, MTX_SPIN) out of order @"
@@ -977,7 +977,8 @@ witness_exit(struct mtx *m, int flags, const char *file, int line)
 		if (m->mtx_recurse != 0)
 			return;
 		mtx_enter(&w_mtx, MTX_SPIN | MTX_QUIET);
-		PCPU_SET(witness_spin_check, witness_spin_check & ~w->w_level);
+		PCPU_SET(witness_spin_check,
+		    PCPU_GET(witness_spin_check) & ~w->w_level);
 		mtx_exit(&w_mtx, MTX_SPIN | MTX_QUIET);
 		return;
 	}
@@ -1011,7 +1012,8 @@ witness_try_enter(struct mtx *m, int flags, const char *file, int line)
 		if (m->mtx_recurse != 0)
 			return;
 		mtx_enter(&w_mtx, MTX_SPIN | MTX_QUIET);
-		PCPU_SET(witness_spin_check, witness_spin_check | w->w_level);
+		PCPU_SET(witness_spin_check,
+		    PCPU_GET(witness_spin_check) | w->w_level);
 		mtx_exit(&w_mtx, MTX_SPIN | MTX_QUIET);
 		w->w_file = file;
 		w->w_line = line;
diff --git a/sys/kern/vfs_bio.c b/sys/kern/vfs_bio.c
index 9949813..14ae4a8 100644
--- a/sys/kern/vfs_bio.c
+++ b/sys/kern/vfs_bio.c
@@ -510,7 +510,7 @@ bread(struct vnode * vp, daddr_t blkno, int size, struct ucred * cred,
 
 	/* if not found in cache, do some I/O */
 	if ((bp->b_flags & B_CACHE) == 0) {
-		if (curproc != idleproc)
+		if (curproc != PCPU_GET(idleproc))
 			curproc->p_stats->p_ru.ru_inblock++;
 		KASSERT(!(bp->b_flags & B_ASYNC), ("bread: illegal async bp %p", bp));
 		bp->b_iocmd = BIO_READ;
@@ -547,7 +547,7 @@ breadn(struct vnode * vp, daddr_t blkno, int size,
 
 	/* if not found in cache, do some I/O */
 	if ((bp->b_flags & B_CACHE) == 0) {
-		if (curproc != idleproc)
+		if (curproc != PCPU_GET(idleproc))
 			curproc->p_stats->p_ru.ru_inblock++;
 		bp->b_iocmd = BIO_READ;
 		bp->b_flags &= ~B_INVAL;
@@ -568,7 +568,7 @@ breadn(struct vnode * vp, daddr_t blkno, int size,
 		rabp = getblk(vp, *rablkno, *rabsize, 0, 0);
 
 		if ((rabp->b_flags & B_CACHE) == 0) {
-			if (curproc != idleproc)
+			if (curproc != PCPU_GET(idleproc))
 				curproc->p_stats->p_ru.ru_inblock++;
 			rabp->b_flags |= B_ASYNC;
 			rabp->b_flags &= ~B_INVAL;
@@ -695,7 +695,7 @@ bwrite(struct buf * bp)
 
 	bp->b_vp->v_numoutput++;
 	vfs_busy_pages(bp, 1);
-	if (curproc != idleproc)
+	if (curproc != PCPU_GET(idleproc))
 		curproc->p_stats->p_ru.ru_oublock++;
 	splx(s);
 	if (oldflags & B_ASYNC)
@@ -2107,7 +2107,7 @@ loop:
 	 * XXX remove if 0 sections (clean this up after its proven)
          */
 	if (numfreebuffers == 0) {
-		if (curproc == idleproc)
+		if (curproc == PCPU_GET(idleproc))
 			return NULL;
 		needsbuffer |= VFS_BIO_NEED_ANY;
 	}
diff --git a/sys/sys/buf.h b/sys/sys/buf.h
index 223c036..4cd8257 100644
--- a/sys/sys/buf.h
+++ b/sys/sys/buf.h
@@ -318,7 +318,7 @@ BUF_KERNPROC(struct buf *bp)
 {
 	struct proc *p = curproc;
 
-	if (p != idleproc && bp->b_lock.lk_lockholder == p->p_pid)
+	if (p != PCPU_GET(idleproc) && bp->b_lock.lk_lockholder == p->p_pid)
 		p->p_locks--;
 	bp->b_lock.lk_lockholder = LK_KERNPROC;
 }