separate tlbflush(u64 myreq) function to flush given an older start time

570214f8 · Nickolai Zeldovich · c841ea25 · 570214f8 · 570214f8
--- a/include/hwvm.hh
+++ b/include/hwvm.hh
@@ -13,6 +13,7 @@ pgmap*          setupkvm(void);
 int             mapkva(pgmap *pml4, char* kva, uptr uva, size_t size);
 std::atomic<pme_t>* walkpgdir(pgmap *pml4, u64, int);
 void            tlbflush(void);
+void            tlbflush(u64 req);

 template<class CB>
 void

--- a/kernel/hwvm.cc
+++ b/kernel/hwvm.cc
@@ -151,10 +151,13 @@ switchvm(struct proc *p)
  mycpu()->ts.rsp[0] = (u64) myproc()->kstack + KSTACKSIZE;
  mycpu()->ts.iomba = (u16)__offsetof(struct taskstate, iopb);
  ltr(TSSSEG);
+
+  u64 nreq = tlbflush_req.load();
  if (p->vmap != 0 && p->vmap->pml4 != 0)
    lcr3(v2p(p->vmap->pml4));  // switch to new address space
  else
    switchkvm();
+  mycpu()->tlbflush_done = nreq;

  writefs(UDSEG);
  writemsr(MSR_FS_BASE, p->user_fs_);
@@ -224,14 +227,20 @@ void
 tlbflush()
 {
  u64 myreq = tlbflush_req++;
+  tlbflush(myreq);
+}

+void
+tlbflush(u64 myreq)
+{
  // the caller may not hold any spinlock, because other CPUs might
  // be spinning waiting for that spinlock, with interrupts disabled,
  // so we will deadlock waiting for their TLB flush..
  assert(mycpu()->ncli == 0);

  for (int i = 0; i < ncpu; i++)
-    lapic_tlbflush(i);
+    if (cpus[i].tlbflush_done < myreq)
+      lapic_tlbflush(i);

  for (int i = 0; i < ncpu; i++)
    while (cpus[i].tlbflush_done < myreq)