1 files changed, 103 insertions, 36 deletions

diff --git a/arch/s390/include/asm/tlbflush.h b/arch/s390/include/asm/tlbflush.h
index 29d5d6d4bec..16c9c88658c 100644
--- a/arch/s390/include/asm/tlbflush.h
+++ b/arch/s390/include/asm/tlbflush.h
@@ -7,19 +7,41 @@
 #include <asm/pgalloc.h>
 
 /*
- * Flush all tlb entries on the local cpu.
+ * Flush all TLB entries on the local CPU.
  */
 static inline void __tlb_flush_local(void)
 {
 	asm volatile("ptlb" : : : "memory");
 }
 
-#ifdef CONFIG_SMP
 /*
- * Flush all tlb entries on all cpus.
+ * Flush TLB entries for a specific ASCE on all CPUs
+ */
+static inline void __tlb_flush_idte(unsigned long asce)
+{
+	/* Global TLB flush for the mm */
+	asm volatile(
+		"	.insn	rrf,0xb98e0000,0,%0,%1,0"
+		: : "a" (2048), "a" (asce) : "cc");
+}
+
+/*
+ * Flush TLB entries for a specific ASCE on the local CPU
  */
+static inline void __tlb_flush_idte_local(unsigned long asce)
+{
+	/* Local TLB flush for the mm */
+	asm volatile(
+		"	.insn	rrf,0xb98e0000,0,%0,%1,1"
+		: : "a" (2048), "a" (asce) : "cc");
+}
+
+#ifdef CONFIG_SMP
 void smp_ptlb_all(void);
 
+/*
+ * Flush all TLB entries on all CPUs.
+ */
 static inline void __tlb_flush_global(void)
 {
 	register unsigned long reg2 asm("2");
@@ -27,12 +49,12 @@ static inline void __tlb_flush_global(void)
 	register unsigned long reg4 asm("4");
 	long dummy;
 
-#ifndef __s390x__
+#ifndef CONFIG_64BIT
 	if (!MACHINE_HAS_CSP) {
 		smp_ptlb_all();
 		return;
 	}
-#endif /* __s390x__ */
+#endif /* CONFIG_64BIT */
 
 	dummy = 0;
 	reg2 = reg3 = 0;
@@ -42,64 +64,109 @@ static inline void __tlb_flush_global(void)
 		: : "d" (reg2), "d" (reg3), "d" (reg4), "m" (dummy) : "cc" );
 }
 
+/*
+ * Flush TLB entries for a specific mm on all CPUs (in case gmap is used
+ * this implicates multiple ASCEs!).
+ */
 static inline void __tlb_flush_full(struct mm_struct *mm)
 {
-	cpumask_t local_cpumask;
-
 	preempt_disable();
-	/*
-	 * If the process only ran on the local cpu, do a local flush.
-	 */
-	local_cpumask = cpumask_of_cpu(smp_processor_id());
-	if (cpumask_equal(mm_cpumask(mm), &local_cpumask))
+	atomic_add(0x10000, &mm->context.attach_count);
+	if (cpumask_equal(mm_cpumask(mm), cpumask_of(smp_processor_id()))) {
+		/* Local TLB flush */
 		__tlb_flush_local();
-	else
+	} else {
+		/* Global TLB flush */
 		__tlb_flush_global();
+		/* Reset TLB flush mask */
+		if (MACHINE_HAS_TLB_LC)
+			cpumask_copy(mm_cpumask(mm),
+				     &mm->context.cpu_attach_mask);
+	}
+	atomic_sub(0x10000, &mm->context.attach_count);
 	preempt_enable();
 }
+
+/*
+ * Flush TLB entries for a specific ASCE on all CPUs.
+ */
+static inline void __tlb_flush_asce(struct mm_struct *mm, unsigned long asce)
+{
+	int active, count;
+
+	preempt_disable();
+	active = (mm == current->active_mm) ? 1 : 0;
+	count = atomic_add_return(0x10000, &mm->context.attach_count);
+	if (MACHINE_HAS_TLB_LC && (count & 0xffff) <= active &&
+	    cpumask_equal(mm_cpumask(mm), cpumask_of(smp_processor_id()))) {
+		__tlb_flush_idte_local(asce);
+	} else {
+		if (MACHINE_HAS_IDTE)
+			__tlb_flush_idte(asce);
+		else
+			__tlb_flush_global();
+		/* Reset TLB flush mask */
+		if (MACHINE_HAS_TLB_LC)
+			cpumask_copy(mm_cpumask(mm),
+				     &mm->context.cpu_attach_mask);
+	}
+	atomic_sub(0x10000, &mm->context.attach_count);
+	preempt_enable();
+}
+
+static inline void __tlb_flush_kernel(void)
+{
+	if (MACHINE_HAS_IDTE)
+		__tlb_flush_idte((unsigned long) init_mm.pgd |
+				 init_mm.context.asce_bits);
+	else
+		__tlb_flush_global();
+}
 #else
+#define __tlb_flush_global()	__tlb_flush_local()
 #define __tlb_flush_full(mm)	__tlb_flush_local()
-#endif
 
 /*
- * Flush all tlb entries of a page table on all cpus.
+ * Flush TLB entries for a specific ASCE on all CPUs.
  */
-static inline void __tlb_flush_idte(unsigned long asce)
+static inline void __tlb_flush_asce(struct mm_struct *mm, unsigned long asce)
 {
-	asm volatile(
-		"	.insn	rrf,0xb98e0000,0,%0,%1,0"
-		: : "a" (2048), "a" (asce) : "cc" );
+	if (MACHINE_HAS_TLB_LC)
+		__tlb_flush_idte_local(asce);
+	else
+		__tlb_flush_local();
 }
 
+static inline void __tlb_flush_kernel(void)
+{
+	if (MACHINE_HAS_TLB_LC)
+		__tlb_flush_idte_local((unsigned long) init_mm.pgd |
+				       init_mm.context.asce_bits);
+	else
+		__tlb_flush_local();
+}
+#endif
+
 static inline void __tlb_flush_mm(struct mm_struct * mm)
 {
-	if (unlikely(cpumask_empty(mm_cpumask(mm))))
-		return;
 	/*
 	 * If the machine has IDTE we prefer to do a per mm flush
 	 * on all cpus instead of doing a local flush if the mm
 	 * only ran on the local cpu.
 	 */
-	if (MACHINE_HAS_IDTE) {
-		if (mm->context.noexec)
-			__tlb_flush_idte((unsigned long)
-					 get_shadow_table(mm->pgd) |
-					 mm->context.asce_bits);
-		__tlb_flush_idte((unsigned long) mm->pgd |
+	if (MACHINE_HAS_IDTE && list_empty(&mm->context.gmap_list))
+		__tlb_flush_asce(mm, (unsigned long) mm->pgd |
 				 mm->context.asce_bits);
-		return;
-	}
-	__tlb_flush_full(mm);
+	else
+		__tlb_flush_full(mm);
 }
 
-static inline void __tlb_flush_mm_cond(struct mm_struct * mm)
+static inline void __tlb_flush_mm_lazy(struct mm_struct * mm)
 {
-	spin_lock(&mm->page_table_lock);
 	if (mm->context.flush_mm) {
 		__tlb_flush_mm(mm);
 		mm->context.flush_mm = 0;
 	}
-	spin_unlock(&mm->page_table_lock);
 }
 
 /*
@@ -126,19 +193,19 @@ static inline void __tlb_flush_mm_cond(struct mm_struct * mm)
 
 static inline void flush_tlb_mm(struct mm_struct *mm)
 {
-	__tlb_flush_mm_cond(mm);
+	__tlb_flush_mm_lazy(mm);
 }
 
 static inline void flush_tlb_range(struct vm_area_struct *vma,
 				   unsigned long start, unsigned long end)
 {
-	__tlb_flush_mm_cond(vma->vm_mm);
+	__tlb_flush_mm_lazy(vma->vm_mm);
 }
 
 static inline void flush_tlb_kernel_range(unsigned long start,
 					  unsigned long end)
 {
-	__tlb_flush_mm(&init_mm);
+	__tlb_flush_kernel();
 }
 
 #endif /* _S390_TLBFLUSH_H */