Add a new 'pmap_invalidate_cache()' to flush the CPU caches via the

wbinvd() instruction.  This includes a new IPI so that all CPU caches on
all CPUs are flushed for the SMP case.

MFC after:	1 month

6 files changed, 64 insertions, 2 deletions

diff --git a/sys/amd64/amd64/apic_vector.S b/sys/amd64/amd64/apic_vector.S
index 96018f3..ab781ca 100644
--- a/sys/amd64/amd64/apic_vector.S
+++ b/sys/amd64/amd64/apic_vector.S
@@ -171,6 +171,25 @@ IDTVEC(invlrng)
 	iretq
 
 /*
+ * Invalidate cache.
+ */
+	.text
+	SUPERALIGN_TEXT
+IDTVEC(invlcache)
+	pushq	%rax
+
+	wbinvd
+
+	movq	lapic, %rax
+	movl	$0, LA_EOI(%rax)	/* End Of Interrupt to APIC */
+
+	lock
+	incl	smp_tlb_wait
+
+	popq	%rax
+	iretq
+
+/*
  * Handler for IPIs sent via the per-cpu IPI bitmap.
  */
 	.text
diff --git a/sys/amd64/amd64/mp_machdep.c b/sys/amd64/amd64/mp_machdep.c
index 226a95c..b2485c6 100644
--- a/sys/amd64/amd64/mp_machdep.c
+++ b/sys/amd64/amd64/mp_machdep.c
@@ -875,6 +875,14 @@ smp_targeted_tlb_shootdown(u_int mask, u_int vector, vm_offset_t addr1, vm_offse
 }
 
 void
+smp_cache_flush(void)
+{
+
+	if (smp_started)
+		smp_tlb_shootdown(IPI_INVLCACHE, 0, 0);
+}
+
+void
 smp_invltlb(void)
 {
 
diff --git a/sys/amd64/amd64/pmap.c b/sys/amd64/amd64/pmap.c
index fd5e21d..238f0b4 100644
--- a/sys/amd64/amd64/pmap.c
+++ b/sys/amd64/amd64/pmap.c
@@ -732,6 +732,30 @@ pmap_invalidate_all(pmap_t pmap)
 	else
 		critical_exit();
 }
+
+void
+pmap_invalidate_cache(void)
+{
+
+	if (smp_started) {
+		if (!(read_rflags() & PSL_I))
+			panic("%s: interrupts disabled", __func__);
+		mtx_lock_spin(&smp_ipi_mtx);
+	} else
+		critical_enter();
+	/*
+	 * We need to disable interrupt preemption but MUST NOT have
+	 * interrupts disabled here.
+	 * XXX we may need to hold schedlock to get a coherent pm_active
+	 * XXX critical sections disable interrupts again
+	 */
+	wbinvd();
+	smp_cache_flush();
+	if (smp_started)
+		mtx_unlock_spin(&smp_ipi_mtx);
+	else
+		critical_exit();
+}
 #else /* !SMP */
 /*
  * Normal, non-SMP, invalidation functions.
@@ -762,6 +786,13 @@ pmap_invalidate_all(pmap_t pmap)
 	if (pmap == kernel_pmap || pmap->pm_active)
 		invltlb();
 }
+
+PMAP_INLINE void
+pmap_invalidate_cache(void)
+{
+
+	wbinvd();
+}
 #endif /* !SMP */
 
 /*
diff --git a/sys/amd64/include/apicvar.h b/sys/amd64/include/apicvar.h
index c87dc7e..98cf3a3 100644
--- a/sys/amd64/include/apicvar.h
+++ b/sys/amd64/include/apicvar.h
@@ -118,8 +118,9 @@
 #define	IPI_INVLTLB	(APIC_IPI_INTS + 1)	/* TLB Shootdown IPIs */
 #define	IPI_INVLPG	(APIC_IPI_INTS + 2)
 #define	IPI_INVLRNG	(APIC_IPI_INTS + 3)
+#define	IPI_INVLCACHE	(APIC_IPI_INTS + 4)
 /* Vector to handle bitmap based IPIs */
-#define	IPI_BITMAP_VECTOR	(APIC_IPI_INTS + 5) 
+#define	IPI_BITMAP_VECTOR	(APIC_IPI_INTS + 6) 
 
 /* IPIs handled by IPI_BITMAPED_VECTOR  (XXX ups is there a better place?) */
 #define	IPI_AST		0 	/* Generate software trap. */
@@ -127,7 +128,7 @@
 #define IPI_BITMAP_LAST IPI_PREEMPT
 #define IPI_IS_BITMAPED(x) ((x) <= IPI_BITMAP_LAST)
 
-#define	IPI_STOP	(APIC_IPI_INTS + 6)	/* Stop CPU until restarted. */
+#define	IPI_STOP	(APIC_IPI_INTS + 7)	/* Stop CPU until restarted. */
 
 /*
  * The spurious interrupt can share the priority class with the IPIs since
diff --git a/sys/amd64/include/pmap.h b/sys/amd64/include/pmap.h
index 49e3139..0a774c7 100644
--- a/sys/amd64/include/pmap.h
+++ b/sys/amd64/include/pmap.h
@@ -309,6 +309,7 @@ void	pmap_unmapdev(vm_offset_t, vm_size_t);
 void	pmap_invalidate_page(pmap_t, vm_offset_t);
 void	pmap_invalidate_range(pmap_t, vm_offset_t, vm_offset_t);
 void	pmap_invalidate_all(pmap_t);
+void	pmap_invalidate_cache(void);
 
 #endif /* _KERNEL */
 
diff --git a/sys/amd64/include/smp.h b/sys/amd64/include/smp.h
index 94a7022..5a2d3aa 100644
--- a/sys/amd64/include/smp.h
+++ b/sys/amd64/include/smp.h
@@ -40,6 +40,7 @@ inthand_t
 	IDTVEC(invltlb),	/* TLB shootdowns - global */
 	IDTVEC(invlpg),		/* TLB shootdowns - 1 page */
 	IDTVEC(invlrng),	/* TLB shootdowns - page range */
+	IDTVEC(invlcache),	/* Write back and invalidate cache */
 	IDTVEC(ipi_intr_bitmap_handler), /* Bitmap based IPIs */ 
 	IDTVEC(cpustop),	/* CPU stops & waits to be restarted */
 	IDTVEC(rendezvous);	/* handle CPU rendezvous */
@@ -56,6 +57,7 @@ void 	ipi_bitmap_handler(struct trapframe frame);
 u_int	mp_bootaddress(u_int);
 int	mp_grab_cpu_hlt(void);
 void	mp_topology(void);
+void	smp_cache_flush(void);
 void	smp_invlpg(vm_offset_t addr);
 void	smp_masked_invlpg(u_int mask, vm_offset_t addr);
 void	smp_invlpg_range(vm_offset_t startva, vm_offset_t endva);