User mode page faults for page zero require special treatment to prevent

a recursive DAE.

Scientists say this is proof that there's a ``strange attractor'' in the
m88k processors. What is certain, though, is that the processors'
documentation contains black holes.

1 files changed, 19 insertions, 1 deletions

diff --git a/sys/arch/mvme88k/mvme88k/trap.c b/sys/arch/mvme88k/mvme88k/trap.c
index 593ae28e294..b5cfa62e988 100644
--- a/sys/arch/mvme88k/mvme88k/trap.c
+++ b/sys/arch/mvme88k/mvme88k/trap.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: trap.c,v 1.43 2003/09/02 10:35:53 miod Exp $	*/
+/*	$OpenBSD: trap.c,v 1.44 2003/09/02 20:57:21 miod Exp $	*/
 /*
  * Copyright (c) 1998 Steve Murphree, Jr.
  * Copyright (c) 1996 Nivas Madhur
@@ -451,6 +451,24 @@ user_fault:
 
 		va = trunc_page((vm_offset_t)fault_addr);
 
+		/*
+		 * XXX
+		 * Page faults for addresses in page zero _SEEM TO BE_ special.
+		 * For some reason, we will keep faulting with the same
+		 * cause when the interrupts get enabled, later in the
+		 * exception process, and eventually eat all the interrupt
+		 * stack in a recursive DAE.
+		 * Since accesses to this page will always result in a
+		 * SIGSEGV, we can safely clear the DAE condition here.
+		 *
+		 * (behaviour experienced on MVME187C)
+		 */
+		if (va == 0) {
+			frame->dpfsr = 0;
+			frame->dmt0 = 0;
+			result = EFAULT;
+		}
+
 		vm = p->p_vmspace;
 		map = &vm->vm_map;