SMP - fixed usage of stale TLB entries

- when kernel copies from userspace, it must be sure that the TLB
  entries are not stale and thus the referenced memory is correct

- everytime we change a process' address space we set p_stale_tlb
  bits for all CPUs.

- Whenever a cpu finds its bit set when it wants to access the
  process' memory, it refreshes the TLB

- it is more conservative than it needs to be but it has low
  overhead than checking precisely

kernel/system/do_update.c

@@ -127,6 +127,11 @@ PUBLIC int do_update(struct proc * caller, message * m_ptr)
   printf("do_update: curr ptproc %d\n", get_cpulocal_var(ptproc)->p_endpoint);
 #endif
 
+#ifdef CONFIG_SMP
+  bits_fill(src_rp->p_stale_tlb, CONFIG_MAX_CPUS);
+  bits_fill(dst_rp->p_stale_tlb, CONFIG_MAX_CPUS);
+#endif
+
   return OK;
 }
 

kernel/system/do_vmctl.c

@@ -159,6 +159,16 @@ PUBLIC int do_vmctl(struct proc * caller, message * m_ptr)
 		 * cpu
 		 */
 		RTS_UNSET(p, RTS_VMINHIBIT);
+#ifdef CONFIG_SMP
+		/*
+		 * We don't know whether kernel has the changed mapping
+		 * installed to access userspace memory. And if so, on what CPU.
+		 * More over we don't know what mapping has changed and how and
+		 * therefore we must invalidate all mappings we have anywhere.
+		 * Next time we map memory, we map it fresh.
+		 */
+		bits_fill(p->p_stale_tlb, CONFIG_MAX_CPUS);
+#endif
 		return OK;
   }