[IA64] - Avoid slow TLB purges on SGI Altix systems

flush_tlb_all() can be a scaling issue on large SGI Altix systems
since it uses the global call_lock and always executes on all cpus.
When a process enters flush_tlb_range() to purge TLBs for another
process, it is possible to avoid flush_tlb_all() and instead allow
sn2_global_tlb_purge() to purge TLBs only where necessary.

This patch modifies flush_tlb_range() so that this case can be handled
by platform TLB purge functions and updates ia64_global_tlb_purge()
accordingly.  sn2_global_tlb_purge() now calculates the region register
value from the mm argument introduced with this patch.

Signed-off-by: Dean Roe <roe@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>

1 files changed, 9 insertions, 7 deletions

diff --git a/arch/ia64/mm/tlb.c b/arch/ia64/mm/tlb.c
index 464557e4ed8..99ea8c70f40 100644
--- a/arch/ia64/mm/tlb.c
+++ b/arch/ia64/mm/tlb.c
@@ -86,10 +86,15 @@ 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);
 	{
@@ -135,15 +140,12 @@ 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))
@@ -153,7 +155,7 @@ 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));