Add sparc64 support.

Compiled (and helped) by:	pluknet

7 files changed, 105 insertions, 72 deletions

diff --git a/sys/sparc64/include/_types.h b/sys/sparc64/include/_types.h
index f810c15..7e993c4 100644
--- a/sys/sparc64/include/_types.h
+++ b/sys/sparc64/include/_types.h
@@ -55,7 +55,6 @@ typedef	unsigned long		__uint64_t;
  * Standard type definitions.
  */
 typedef	__int32_t	__clock_t;		/* clock()... */
-typedef	unsigned int	__cpumask_t;
 typedef	__int64_t	__critical_t;
 typedef	double		__double_t;
 typedef	float		__float_t;
diff --git a/sys/sparc64/include/pmap.h b/sys/sparc64/include/pmap.h
index e16ea97..adad257 100644
--- a/sys/sparc64/include/pmap.h
+++ b/sys/sparc64/include/pmap.h
@@ -40,6 +40,7 @@
 #define	_MACHINE_PMAP_H_
 
 #include <sys/queue.h>
+#include <sys/_cpuset.h>
 #include <sys/_lock.h>
 #include <sys/_mutex.h>
 #include <machine/cache.h>
@@ -61,7 +62,7 @@ struct pmap {
 	struct	mtx pm_mtx;
 	struct	tte *pm_tsb;
 	vm_object_t pm_tsb_obj;
-	cpumask_t pm_active;
+	cpuset_t pm_active;
 	u_int	pm_context[MAXCPU];
 	struct	pmap_statistics pm_stats;
 };
diff --git a/sys/sparc64/include/smp.h b/sys/sparc64/include/smp.h
index 3ca8e03..f8f7d1c 100644
--- a/sys/sparc64/include/smp.h
+++ b/sys/sparc64/include/smp.h
@@ -38,6 +38,7 @@
 
 #ifndef	LOCORE
 
+#include <sys/cpuset.h>
 #include <sys/proc.h>
 #include <sys/sched.h>
 
@@ -76,17 +77,17 @@ struct cpu_start_args {
 };
 
 struct ipi_cache_args {
-	cpumask_t ica_mask;
+	cpuset_t ica_mask;
 	vm_paddr_t ica_pa;
 };
 
 struct ipi_rd_args {
-	cpumask_t ira_mask;
+	cpuset_t ira_mask;
 	register_t *ira_val;
 };
 
 struct ipi_tlb_args {
-	cpumask_t ita_mask;
+	cpuset_t ita_mask;
 	struct	pmap *ita_pmap;
 	u_long	ita_start;
 	u_long	ita_end;
@@ -100,7 +101,7 @@ extern struct pcb stoppcbs[];
 void	cpu_mp_bootstrap(struct pcpu *pc);
 void	cpu_mp_shutdown(void);
 
-typedef	void cpu_ipi_selected_t(u_int, u_long, u_long, u_long);
+typedef	void cpu_ipi_selected_t(cpuset_t, u_long, u_long, u_long);
 extern	cpu_ipi_selected_t *cpu_ipi_selected;
 typedef	void cpu_ipi_single_t(u_int, u_long, u_long, u_long);
 extern	cpu_ipi_single_t *cpu_ipi_single;
@@ -140,7 +141,7 @@ ipi_all_but_self(u_int ipi)
 }
 
 static __inline void
-ipi_selected(u_int cpus, u_int ipi)
+ipi_selected(cpuset_t cpus, u_int ipi)
 {
 
 	cpu_ipi_selected(cpus, 0, (u_long)tl_ipi_level, ipi);
@@ -197,7 +198,8 @@ ipi_rd(u_int cpu, void *func, u_long *val)
 	sched_pin();
 	ira = &ipi_rd_args;
 	mtx_lock_spin(&ipi_mtx);
-	ira->ira_mask = 1 << cpu | PCPU_GET(cpumask);
+	ira->ira_mask = PCPU_GET(cpumask);
+	CPU_SET(cpu, &ira->ira_mask);
 	ira->ira_val = val;
 	cpu_ipi_single(cpu, 0, (u_long)func, (u_long)ira);
 	return (&ira->ira_mask);
@@ -207,18 +209,21 @@ static __inline void *
 ipi_tlb_context_demap(struct pmap *pm)
 {
 	struct ipi_tlb_args *ita;
-	cpumask_t cpus;
+	cpuset_t cpus;
 
 	if (smp_cpus == 1)
 		return (NULL);
 	sched_pin();
-	if ((cpus = (pm->pm_active & PCPU_GET(other_cpus))) == 0) {
+	cpus = pm->pm_active;
+	CPU_AND(&cpus, PCPU_PTR(other_cpus));
+	if (CPU_EMPTY(&cpus)) {
 		sched_unpin();
 		return (NULL);
 	}
 	ita = &ipi_tlb_args;
 	mtx_lock_spin(&ipi_mtx);
-	ita->ita_mask = cpus | PCPU_GET(cpumask);
+	CPU_OR(&cpus, PCPU_PTR(cpumask));
+	ita->ita_mask = cpus;
 	ita->ita_pmap = pm;
 	cpu_ipi_selected(cpus, 0, (u_long)tl_ipi_tlb_context_demap,
 	    (u_long)ita);
@@ -229,18 +234,21 @@ static __inline void *
 ipi_tlb_page_demap(struct pmap *pm, vm_offset_t va)
 {
 	struct ipi_tlb_args *ita;
-	cpumask_t cpus;
+	cpuset_t cpus;
 
 	if (smp_cpus == 1)
 		return (NULL);
 	sched_pin();
-	if ((cpus = (pm->pm_active & PCPU_GET(other_cpus))) == 0) {
+	cpus = pm->pm_active;
+	CPU_AND(&cpus, PCPU_PTR(other_cpus));
+	if (CPU_EMPTY(&cpus)) {
 		sched_unpin();
 		return (NULL);
 	}
 	ita = &ipi_tlb_args;
 	mtx_lock_spin(&ipi_mtx);
-	ita->ita_mask = cpus | PCPU_GET(cpumask);
+	CPU_OR(&cpus, PCPU_PTR(cpumask));
+	ita->ita_mask = cpus;
 	ita->ita_pmap = pm;
 	ita->ita_va = va;
 	cpu_ipi_selected(cpus, 0, (u_long)tl_ipi_tlb_page_demap, (u_long)ita);
@@ -251,18 +259,21 @@ static __inline void *
 ipi_tlb_range_demap(struct pmap *pm, vm_offset_t start, vm_offset_t end)
 {
 	struct ipi_tlb_args *ita;
-	cpumask_t cpus;
+	cpuset_t cpus;
 
 	if (smp_cpus == 1)
 		return (NULL);
 	sched_pin();
-	if ((cpus = (pm->pm_active & PCPU_GET(other_cpus))) == 0) {
+	cpus = pm->pm_active;
+	CPU_AND(&cpus, PCPU_PTR(other_cpus));
+	if (CPU_EMPTY(&cpus)) {
 		sched_unpin();
 		return (NULL);
 	}
 	ita = &ipi_tlb_args;
 	mtx_lock_spin(&ipi_mtx);
-	ita->ita_mask = cpus | PCPU_GET(cpumask);
+	CPU_OR(&cpus, PCPU_PTR(cpumask));
+	ita->ita_mask = cpus;
 	ita->ita_pmap = pm;
 	ita->ita_start = start;
 	ita->ita_end = end;
@@ -274,11 +285,11 @@ ipi_tlb_range_demap(struct pmap *pm, vm_offset_t start, vm_offset_t end)
 static __inline void
 ipi_wait(void *cookie)
 {
-	volatile cpumask_t *mask;
+	volatile cpuset_t *mask;
 
 	if ((mask = cookie) != NULL) {
-		atomic_clear_int(mask, PCPU_GET(cpumask));
-		while (*mask != 0)
+		CPU_NAND_ATOMIC(mask, PCPU_PTR(cpumask));
+		while (!CPU_EMPTY(mask))
 			;
 		mtx_unlock_spin(&ipi_mtx);
 		sched_unpin();
diff --git a/sys/sparc64/sparc64/intr_machdep.c b/sys/sparc64/sparc64/intr_machdep.c
index f6ef9a7..ed30182 100644
--- a/sys/sparc64/sparc64/intr_machdep.c
+++ b/sys/sparc64/sparc64/intr_machdep.c
@@ -445,8 +445,7 @@ intr_describe(int vec, void *ih, const char *descr)
  * allocate CPUs round-robin.
  */
 
-/* The BSP is always a valid target. */
-static cpumask_t intr_cpus = (1 << 0);
+static cpuset_t intr_cpus;
 static int current_cpu;
 
 static void
@@ -468,7 +467,7 @@ intr_assign_next_cpu(struct intr_vector *iv)
 		current_cpu++;
 		if (current_cpu > mp_maxid)
 			current_cpu = 0;
-	} while (!(intr_cpus & (1 << current_cpu)));
+	} while (!CPU_ISSET(current_cpu, &intr_cpus));
 }
 
 /* Attempt to bind the specified IRQ to the specified CPU. */
@@ -504,7 +503,7 @@ intr_add_cpu(u_int cpu)
 	if (bootverbose)
 		printf("INTR: Adding CPU %d as a target\n", cpu);
 
-	intr_cpus |= (1 << cpu);
+	CPU_SET(cpu, &intr_cpus);
 }
 
 /*
@@ -518,6 +517,9 @@ intr_shuffle_irqs(void *arg __unused)
 	struct intr_vector *iv;
 	int i;
 
+	/* The BSP is always a valid target. */
+	CPU_SETOF(0, &intr_cpus);
+
 	/* Don't bother on UP. */
 	if (mp_ncpus == 1)
 		return;
diff --git a/sys/sparc64/sparc64/mp_machdep.c b/sys/sparc64/sparc64/mp_machdep.c
index 8700f89..9bee247 100644
--- a/sys/sparc64/sparc64/mp_machdep.c
+++ b/sys/sparc64/sparc64/mp_machdep.c
@@ -121,7 +121,7 @@ cpu_ipi_single_t *cpu_ipi_single;
 static vm_offset_t mp_tramp;
 static u_int cpuid_to_mid[MAXCPU];
 static int isjbus;
-static volatile cpumask_t shutdown_cpus;
+static volatile cpuset_t shutdown_cpus;
 
 static void ap_count(phandle_t node, u_int mid, u_int cpu_impl);
 static void ap_start(phandle_t node, u_int mid, u_int cpu_impl);
@@ -228,7 +228,7 @@ void
 cpu_mp_setmaxid()
 {
 
-	all_cpus = 1 << curcpu;
+	CPU_SETOF(curcpu, &all_cpus);
 	mp_ncpus = 1;
 	mp_maxid = 0;
 
@@ -283,6 +283,7 @@ sun4u_startcpu(phandle_t cpu, void *func, u_long arg)
 void
 cpu_mp_start(void)
 {
+	cpuset_t ocpus;
 
 	mtx_init(&ipi_mtx, "ipi", NULL, MTX_SPIN);
 
@@ -299,7 +300,9 @@ cpu_mp_start(void)
 	KASSERT(!isjbus || mp_ncpus <= IDR_JALAPENO_MAX_BN_PAIRS,
 	    ("%s: can only IPI a maximum of %d JBus-CPUs",
 	    __func__, IDR_JALAPENO_MAX_BN_PAIRS));
-	PCPU_SET(other_cpus, all_cpus & ~(1 << curcpu));
+	ocpus = all_cpus;
+	CPU_CLR(curcpu, &ocpus);
+	PCPU_SET(other_cpus, ocpus);
 	smp_active = 1;
 }
 
@@ -357,7 +360,7 @@ ap_start(phandle_t node, u_int mid, u_int cpu_impl)
 
 	cache_init(pc);
 
-	all_cpus |= 1 << cpuid;
+	CPU_SET(cpuid, &all_cpus);
 	intr_add_cpu(cpuid);
 }
 
@@ -421,6 +424,7 @@ cpu_mp_unleash(void *v)
 void
 cpu_mp_bootstrap(struct pcpu *pc)
 {
+	cpuset_t ocpus;
 	volatile struct cpu_start_args *csa;
 
 	csa = &cpu_start_args;
@@ -465,7 +469,9 @@ cpu_mp_bootstrap(struct pcpu *pc)
 
 	smp_cpus++;
 	KASSERT(curthread != NULL, ("%s: curthread", __func__));
-	PCPU_SET(other_cpus, all_cpus & ~(1 << curcpu));
+	ocpus = all_cpus;
+	CPU_CLR(curcpu, &all_cpus);
+	PCPU_SET(other_cpus, ocpus);
 	printf("SMP: AP CPU #%d Launched!\n", curcpu);
 
 	csa->csa_count--;
@@ -484,14 +490,22 @@ cpu_mp_bootstrap(struct pcpu *pc)
 void
 cpu_mp_shutdown(void)
 {
+	cpuset_t cpus;
 	int i;
 
 	critical_enter();
 	shutdown_cpus = PCPU_GET(other_cpus);
-	if (stopped_cpus != PCPU_GET(other_cpus))	/* XXX */
-		stop_cpus(stopped_cpus ^ PCPU_GET(other_cpus));
+	cpus = shutdown_cpus;
+
+	/* XXX: Stopp all the CPUs which aren't already. */
+	if (CPU_CMP(&stopped_cpus, &cpus)) {
+
+		/* pc_other_cpus is just a flat "on" mask without curcpu. */
+		CPU_NAND(&cpus, &stopped_cpus);
+		stop_cpus(cpus);
+	}
 	i = 0;
-	while (shutdown_cpus != 0) {
+	while (!CPU_EMPTY(&shutdown_cpus)) {
 		if (i++ > 100000) {
 			printf("timeout shutting down CPUs.\n");
 			break;
@@ -509,20 +523,24 @@ cpu_ipi_ast(struct trapframe *tf)
 static void
 cpu_ipi_stop(struct trapframe *tf)
 {
+	cpuset_t tcmask;
 
 	CTR2(KTR_SMP, "%s: stopped %d", __func__, curcpu);
+	sched_pin();
 	savectx(&stoppcbs[curcpu]);
-	atomic_set_acq_int(&stopped_cpus, PCPU_GET(cpumask));
-	while ((started_cpus & PCPU_GET(cpumask)) == 0) {
-		if ((shutdown_cpus & PCPU_GET(cpumask)) != 0) {
-			atomic_clear_int(&shutdown_cpus, PCPU_GET(cpumask));
+	tcmask = PCPU_GET(cpumask);
+	CPU_OR_ATOMIC(&stopped_cpus, &tcmask);
+	while (!CPU_OVERLAP(&started_cpus, &tcmask)) {
+		if (CPU_OVERLAP(&shutdown_cpus, &tcmask)) {
+			CPU_OR_ATOMIC(&shutdown_cpus, &tcmask);
 			(void)intr_disable();
 			for (;;)
 				;
 		}
 	}
-	atomic_clear_rel_int(&started_cpus, PCPU_GET(cpumask));
-	atomic_clear_rel_int(&stopped_cpus, PCPU_GET(cpumask));
+	CPU_NAND_ATOMIC(&started_cpus, &tcmask);
+	CPU_NAND_ATOMIC(&stopped_cpus, &tcmask);
+	sched_unpin();
 	CTR2(KTR_SMP, "%s: restarted %d", __func__, curcpu);
 }
 
@@ -551,13 +569,13 @@ cpu_ipi_hardclock(struct trapframe *tf)
 }
 
 static void
-spitfire_ipi_selected(u_int cpus, u_long d0, u_long d1, u_long d2)
+spitfire_ipi_selected(cpuset_t cpus, u_long d0, u_long d1, u_long d2)
 {
 	u_int cpu;
 
-	while (cpus) {
-		cpu = ffs(cpus) - 1;
-		cpus &= ~(1 << cpu);
+	while (CPU_EMPTY(&cpus)) {
+		cpu = cpusetobj_ffs(&cpus) - 1;
+		CPU_CLR(cpu, &cpus);
 		spitfire_ipi_single(cpu, d0, d1, d2);
 	}
 }
@@ -657,20 +675,21 @@ cheetah_ipi_single(u_int cpu, u_long d0, u_long d1, u_long d2)
 }
 
 static void
-cheetah_ipi_selected(u_int cpus, u_long d0, u_long d1, u_long d2)
+cheetah_ipi_selected(cpuset_t cpus, u_long d0, u_long d1, u_long d2)
 {
+	char pbuf[CPUSETBUFSIZ];
 	register_t s;
 	u_long ids;
 	u_int bnp;
 	u_int cpu;
 	int i;
 
-	KASSERT((cpus & (1 << curcpu)) == 0,
-	    ("%s: CPU can't IPI itself", __func__));
+	KASSERT(!CPU_ISSET(curcpu, &cpus), ("%s: CPU can't IPI itself",
+	    __func__));
 	KASSERT((ldxa(0, ASI_INTR_DISPATCH_STATUS) &
 	    IDR_CHEETAH_ALL_BUSY) == 0,
 	    ("%s: outstanding dispatch", __func__));
-	if (cpus == 0)
+	if (CPU_EMPTY(&cpus))
 		return;
 	ids = 0;
 	for (i = 0; i < IPI_RETRIES * mp_ncpus; i++) {
@@ -681,7 +700,7 @@ cheetah_ipi_selected(u_int cpus, u_long d0, u_long d1, u_long d2)
 		membar(Sync);
 		bnp = 0;
 		for (cpu = 0; cpu < mp_ncpus; cpu++) {
-			if ((cpus & (1 << cpu)) != 0) {
+			if (CPU_ISSET(cpu, &cpus)) {
 				stxa(AA_INTR_SEND | (cpuid_to_mid[cpu] <<
 				    IDC_ITID_SHIFT) | bnp << IDC_BN_SHIFT,
 				    ASI_SDB_INTR_W, 0);
@@ -698,9 +717,9 @@ cheetah_ipi_selected(u_int cpus, u_long d0, u_long d1, u_long d2)
 			return;
 		bnp = 0;
 		for (cpu = 0; cpu < mp_ncpus; cpu++) {
-			if ((cpus & (1 << cpu)) != 0) {
+			if (CPU_ISSET(cpu, &cpus)) {
 				if ((ids & (IDR_NACK << (2 * bnp))) == 0)
-					cpus &= ~(1 << cpu);
+					CPU_CLR(cpu, &cpus);
 				bnp++;
 			}
 		}
@@ -709,7 +728,7 @@ cheetah_ipi_selected(u_int cpus, u_long d0, u_long d1, u_long d2)
 		 * CPUs we actually haven't tried to send an IPI to,
 		 * but which apparently can be safely ignored.
 		 */
-		if (cpus == 0)
+		if (CPU_EMPTY(&cpus))
 			return;
 		/*
 		 * Leave interrupts enabled for a bit before retrying
@@ -719,11 +738,11 @@ cheetah_ipi_selected(u_int cpus, u_long d0, u_long d1, u_long d2)
 		DELAY(2 * mp_ncpus);
 	}
 	if (kdb_active != 0 || panicstr != NULL)
-		printf("%s: couldn't send IPI (cpus=0x%u ids=0x%lu)\n",
-		    __func__, cpus, ids);
+		printf("%s: couldn't send IPI (cpus=%s ids=0x%lu)\n",
+		    __func__, cpusetobj_strprint(pbuf, &cpus), ids);
 	else
-		panic("%s: couldn't send IPI (cpus=0x%u ids=0x%lu)",
-		    __func__, cpus, ids);
+		panic("%s: couldn't send IPI (cpus=%s ids=0x%lu)",
+		    __func__, cpusetobj_strprint(pbuf, &cpus), ids);
 }
 
 static void
@@ -772,19 +791,20 @@ jalapeno_ipi_single(u_int cpu, u_long d0, u_long d1, u_long d2)
 }
 
 static void
-jalapeno_ipi_selected(u_int cpus, u_long d0, u_long d1, u_long d2)
+jalapeno_ipi_selected(cpuset_t cpus, u_long d0, u_long d1, u_long d2)
 {
+	char pbuf[CPUSETBUFSIZ];
 	register_t s;
 	u_long ids;
 	u_int cpu;
 	int i;
 
-	KASSERT((cpus & (1 << curcpu)) == 0,
-	    ("%s: CPU can't IPI itself", __func__));
+	KASSERT(!CPU_ISSET(curcpu, &cpus), ("%s: CPU can't IPI itself",
+	    __func__));
 	KASSERT((ldxa(0, ASI_INTR_DISPATCH_STATUS) &
 	    IDR_CHEETAH_ALL_BUSY) == 0,
 	    ("%s: outstanding dispatch", __func__));
-	if (cpus == 0)
+	if (CPU_EMPTY(&cpus))
 		return;
 	ids = 0;
 	for (i = 0; i < IPI_RETRIES * mp_ncpus; i++) {
@@ -794,7 +814,7 @@ jalapeno_ipi_selected(u_int cpus, u_long d0, u_long d1, u_long d2)
 		stxa(AA_SDB_INTR_D2, ASI_SDB_INTR_W, d2);
 		membar(Sync);
 		for (cpu = 0; cpu < mp_ncpus; cpu++) {
-			if ((cpus & (1 << cpu)) != 0) {
+			if (CPU_ISSET(cpu, &cpus)) {
 				stxa(AA_INTR_SEND | (cpuid_to_mid[cpu] <<
 				    IDC_ITID_SHIFT), ASI_SDB_INTR_W, 0);
 				membar(Sync);
@@ -808,10 +828,10 @@ jalapeno_ipi_selected(u_int cpus, u_long d0, u_long d1, u_long d2)
 		    (IDR_CHEETAH_ALL_BUSY | IDR_CHEETAH_ALL_NACK)) == 0)
 			return;
 		for (cpu = 0; cpu < mp_ncpus; cpu++)
-			if ((cpus & (1 << cpu)) != 0)
+			if (CPU_ISSET(cpu, &cpus))
 				if ((ids & (IDR_NACK <<
 				    (2 * cpuid_to_mid[cpu]))) == 0)
-					cpus &= ~(1 << cpu);
+					CPU_CLR(cpu, &cpus);
 		/*
 		 * Leave interrupts enabled for a bit before retrying
 		 * in order to avoid deadlocks if the other CPUs are
@@ -820,9 +840,9 @@ jalapeno_ipi_selected(u_int cpus, u_long d0, u_long d1, u_long d2)
 		DELAY(2 * mp_ncpus);
 	}
 	if (kdb_active != 0 || panicstr != NULL)
-		printf("%s: couldn't send IPI (cpus=0x%u ids=0x%lu)\n",
-		    __func__, cpus, ids);
+		printf("%s: couldn't send IPI (cpus=%s ids=0x%lu)\n",
+		    __func__, cpusetobj_strprint(pbuf, &cpus), ids);
 	else
-		panic("%s: couldn't send IPI (cpus=0x%u ids=0x%lu)",
-		    __func__, cpus, ids);
+		panic("%s: couldn't send IPI (cpus=%s ids=0x%lu)",
+		    __func__, cpusetobj_strprint(pbuf, &cpus), ids);
 }
diff --git a/sys/sparc64/sparc64/pmap.c b/sys/sparc64/sparc64/pmap.c
index 09482b9..22393a0 100644
--- a/sys/sparc64/sparc64/pmap.c
+++ b/sys/sparc64/sparc64/pmap.c
@@ -664,7 +664,7 @@ pmap_bootstrap(u_int cpu_impl)
 	pm = kernel_pmap;
 	for (i = 0; i < MAXCPU; i++)
 		pm->pm_context[i] = TLB_CTX_KERNEL;
-	pm->pm_active = ~0;
+	CPU_ZERO(&pm->pm_active);
 
 	/*
 	 * Flush all non-locked TLB entries possibly left over by the
@@ -1189,7 +1189,7 @@ pmap_pinit0(pmap_t pm)
 	PMAP_LOCK_INIT(pm);
 	for (i = 0; i < MAXCPU; i++)
 		pm->pm_context[i] = TLB_CTX_KERNEL;
-	pm->pm_active = 0;
+	CPU_ZERO(&pm->pm_active);
 	pm->pm_tsb = NULL;
 	pm->pm_tsb_obj = NULL;
 	bzero(&pm->pm_stats, sizeof(pm->pm_stats));
@@ -1229,7 +1229,7 @@ pmap_pinit(pmap_t pm)
 	mtx_lock_spin(&sched_lock);
 	for (i = 0; i < MAXCPU; i++)
 		pm->pm_context[i] = -1;
-	pm->pm_active = 0;
+	CPU_ZERO(&pm->pm_active);
 	mtx_unlock_spin(&sched_lock);
 
 	VM_OBJECT_LOCK(pm->pm_tsb_obj);
@@ -2230,7 +2230,7 @@ pmap_activate(struct thread *td)
 	PCPU_SET(tlb_ctx, context + 1);
 
 	pm->pm_context[curcpu] = context;
-	pm->pm_active |= PCPU_GET(cpumask);
+	CPU_OR(&pm->pm_active, PCPU_PTR(cpumask));
 	PCPU_SET(pmap, pm);
 
 	stxa(AA_DMMU_TSB, ASI_DMMU, pm->pm_tsb);
diff --git a/sys/sparc64/sparc64/tlb.c b/sys/sparc64/sparc64/tlb.c
index 990c777..9fcece6 100644
--- a/sys/sparc64/sparc64/tlb.c
+++ b/sys/sparc64/sparc64/tlb.c
@@ -80,7 +80,7 @@ tlb_context_demap(struct pmap *pm)
 	PMAP_STATS_INC(tlb_ncontext_demap);
 	cookie = ipi_tlb_context_demap(pm);
 	s = intr_disable();
-	if (pm->pm_active & PCPU_GET(cpumask)) {
+	if (CPU_OVERLAP(&pm->pm_active, PCPU_PTR(cpumask))) {
 		KASSERT(pm->pm_context[curcpu] != -1,
 		    ("tlb_context_demap: inactive pmap?"));
 		stxa(TLB_DEMAP_PRIMARY | TLB_DEMAP_CONTEXT, ASI_DMMU_DEMAP, 0);
@@ -101,7 +101,7 @@ tlb_page_demap(struct pmap *pm, vm_offset_t va)
 	PMAP_STATS_INC(tlb_npage_demap);
 	cookie = ipi_tlb_page_demap(pm, va);
 	s = intr_disable();
-	if (pm->pm_active & PCPU_GET(cpumask)) {
+	if (CPU_OVERLAP(&pm->pm_active, PCPU_PTR(cpumask))) {
 		KASSERT(pm->pm_context[curcpu] != -1,
 		    ("tlb_page_demap: inactive pmap?"));
 		if (pm == kernel_pmap)
@@ -128,7 +128,7 @@ tlb_range_demap(struct pmap *pm, vm_offset_t start, vm_offset_t end)
 	PMAP_STATS_INC(tlb_nrange_demap);
 	cookie = ipi_tlb_range_demap(pm, start, end);
 	s = intr_disable();
-	if (pm->pm_active & PCPU_GET(cpumask)) {
+	if (CPU_OVERLAP(&pm->pm_active, PCPU_PTR(cpumask))) {
 		KASSERT(pm->pm_context[curcpu] != -1,
 		    ("tlb_range_demap: inactive pmap?"));
 		if (pm == kernel_pmap)