Pull fix-slow-tlb-purge into release branch

1 files changed, 9 insertions, 7 deletions

diff --git a/arch/ia64/mm/tlb.c b/arch/ia64/mm/tlb.c
index 987fb754d6ad..c93e0f2b5fea 100644
--- a/arch/ia64/mm/tlb.c
+++ b/arch/ia64/mm/tlb.c
@@ -87,10 +87,15 @@ wrap_mmu_context (struct mm_struct *mm)
 }
 
 void
-ia64_global_tlb_purge (unsigned long start, unsigned long end, unsigned long nbits)
+ia64_global_tlb_purge (struct mm_struct *mm, unsigned long start, unsigned long end, unsigned long nbits)
 {
         static DEFINE_SPINLOCK(ptcg_lock);
 
+        if (mm != current->active_mm) {
+                flush_tlb_all();
+                return;
+        }
+
         /* HW requires global serialization of ptc.ga.  */
         spin_lock(&ptcg_lock);
         {
@@ -136,15 +141,12 @@ flush_tlb_range (struct vm_area_struct *vma, unsigned long start, unsigned long
         unsigned long size = end - start;
         unsigned long nbits;
 
+#ifndef CONFIG_SMP
         if (mm != current->active_mm) {
-                /* this does happen, but perhaps it's not worth optimizing for? */
-#ifdef CONFIG_SMP
-                flush_tlb_all();
-#else
                 mm->context = 0;
-#endif
                 return;
         }
+#endif
 
         nbits = ia64_fls(size + 0xfff);
         while (unlikely (((1UL << nbits) & purge.mask) == 0) && (nbits < purge.max_bits))
@@ -154,7 +156,7 @@ flush_tlb_range (struct vm_area_struct *vma, unsigned long start, unsigned long
         start &= ~((1UL << nbits) - 1);
 
 # ifdef CONFIG_SMP
-        platform_global_tlb_purge(start, end, nbits);
+        platform_global_tlb_purge(mm, start, end, nbits);
 # else
         do {
                 ia64_ptcl(start, (nbits<<2));