tlb flush: further bugs? (2.3.12-9)

PATCH New Manfred Spraul 1 participant 1 message Jul 29, 1999 Memory management

Share & tools

Manfred Spraul

author

1999-07-29 11:59 UTC #0 lore raw

PATCH 1/1 tlb flush: further bugs? (2.3.12-9)

I think I found 2 minor bugs in the current tlb flush code:
- during crossing TLB flushes, mm->cpu_mask is not set.
- flush_tlb_current_task() contains a race between flush_tlb_other_cpus(the IPI could set mm->cpu_mask) & the next line.
effectively, my patch means that flush_tlb_current_task() is identical to flush_tlb_mm(current->mm), perhaps this should be replaced with a #define.

I've attached a patch (untested) against 2.3.12-9,

Manfred

SMP.C renamed from smp.c.old +8 -4

@@ -1617,6 +1617,8 @@
 			if (test_bit(cpu, &smp_invalidate_needed)) {
 				clear_bit(cpu, &smp_invalidate_needed);
 				local_flush_tlb();
+                                if(current->mm)
+                                        atomic_set_mask(1 << cpu, &current->mm->cpu_vm_mask);
 			}
 			--stuck;
 			if (!stuck) {

@@ -1637,14 +1639,16 @@
  */	
 void flush_tlb_current_task(void)
 {
-	unsigned long vm_mask = 1 << current->processor;
 	struct mm_struct *mm = current->mm;
+	unsigned long vm_mask = 1 << current->processor;
+	unsigned long cpu_mask = mm->cpu_vm_mask & ~vm_mask;
 
-	if (mm->cpu_vm_mask != vm_mask) {
-		flush_tlb_others(mm->cpu_vm_mask & ~vm_mask);
+	mm->cpu_vm_mask = 0;
+	if (current->active_mm == mm) {
 		mm->cpu_vm_mask = vm_mask;
+		local_flush_tlb();
 	}
-	local_flush_tlb();
+	flush_tlb_others(cpu_mask);
 }
 
 void flush_tlb_mm(struct mm_struct * mm)