Correctly detect xmem operations faulting on missing pages on 88110.

These must be handled as write faults rather than read faults, since xmem
performs both a read and a write, and unlike on 88100, we don't have an easy
bit to check.

This solves libcrypto spinning on its locks on 88110.

diff --git a/sys/arch/m88k/m88k/trap.c b/sys/arch/m88k/m88k/trap.c
index 28cdc6b..ee542f3 100644 (file)
--- a/sys/arch/m88k/m88k/trap.c
+++ b/sys/arch/m88k/m88k/trap.c
@@ -1,4 +1,4 @@
-/*     $OpenBSD: trap.c,v 1.120 2021/12/09 00:26:11 guenther Exp $     */
+/*     $OpenBSD: trap.c,v 1.121 2022/08/02 20:15:28 miod Exp $ */
 /*
  * Copyright (c) 2004, Miodrag Vallat.
  * Copyright (c) 1998 Steve Murphree, Jr.
@@ -868,7 +868,17 @@ m88110_user_fault:
 #endif
                } else {
                        fault_addr = frame->tf_dlar;
-                       if (frame->tf_dsr & CMMU_DSR_RW) {
+                       /*
+                        * Unlike the 88100, there is no specific bit telling
+                        * us this is the read part of an xmem operation.
+                        * However, if the WE (Write Exception) bit is set,
+                        * then obviously this is not a read fault.
+                        * But the value of this bit can not be relied upon
+                        * if either PI or SI are set...
+                        */
+                       if ((frame->tf_dsr & CMMU_DSR_RW) != 0 &&
+                           ((frame->tf_dsr & (CMMU_DSR_PI|CMMU_DSR_SI)) != 0 ||
+                            (frame->tf_dsr & CMMU_DSR_WE) == 0)) {
                                access_type = PROT_READ;
                                fault_code = PROT_READ;
                        } else {