Change the behavior of isa_nmi to log an error message instead of

panicing and return a status so that we can decide whether to drop
into DDB or panic.  If the status from isa_nmi is true, panic the
kernel based on machdep.panic_on_nmi, otherwise if DDB is
enabled, drop to DDB based on machdep.ddb_on_nmi.

Reviewed by:	peter, phk

3 files changed, 93 insertions, 45 deletions

diff --git a/sys/amd64/amd64/trap.c b/sys/amd64/amd64/trap.c
index 5a435fe..51de1ac 100644
--- a/sys/amd64/amd64/trap.c
+++ b/sys/amd64/amd64/trap.c
@@ -56,6 +56,7 @@
 #include <sys/resourcevar.h>
 #include <sys/signalvar.h>
 #include <sys/syscall.h>
+#include <sys/sysctl.h>
 #include <sys/sysent.h>
 #include <sys/uio.h>
 #include <sys/vmmeter.h>
@@ -147,6 +148,15 @@ static __inline int userret __P((struct proc *p, struct trapframe *frame,
 extern int has_f00f_bug;
 #endif
 
+#ifdef DDB
+static int ddb_on_nmi = 1;
+SYSCTL_INT(_machdep, OID_AUTO, ddb_on_nmi, CTLFLAG_RW,
+	&ddb_on_nmi, 0, "Go to DDB on NMI");
+#endif
+static int panic_on_nmi = 1;
+SYSCTL_INT(_machdep, OID_AUTO, panic_on_nmi, CTLFLAG_RW,
+	&panic_on_nmi, 0, "Panic on NMI");
+
 static __inline int
 userret(p, frame, oticks, have_mplock)
 	struct proc *p;
@@ -372,13 +382,19 @@ restart:
 			/* machine/parity/power fail/"kitchen sink" faults */
 			if (isa_nmi(code) == 0) {
 #ifdef DDB
-				/* NMI can be hooked up to a pushbutton for debugging */
-				printf ("NMI ... going to debugger\n");
-				kdb_trap (type, 0, &frame);
+				/*
+				 * NMI can be hooked up to a pushbutton
+				 * for debugging.
+				 */
+				if (ddb_on_nmi) {
+					printf ("NMI ... going to debugger\n");
+					kdb_trap (type, 0, &frame);
+				}
 #endif /* DDB */
 				return;
-			}
-			panic("NMI indicates hardware failure");
+			} else if (panic_on_nmi)
+				panic("NMI indicates hardware failure");
+			break;
 #endif /* POWERFAIL_NMI */
 #endif /* NISA > 0 */
 
@@ -577,12 +593,18 @@ kernel_trap:
 			/* machine/parity/power fail/"kitchen sink" faults */
 			if (isa_nmi(code) == 0) {
 #ifdef DDB
-				/* NMI can be hooked up to a pushbutton for debugging */
-				printf ("NMI ... going to debugger\n");
-				kdb_trap (type, 0, &frame);
+				/*
+				 * NMI can be hooked up to a pushbutton
+				 * for debugging.
+				 */
+				if (ddb_on_nmi) {
+					printf ("NMI ... going to debugger\n");
+					kdb_trap (type, 0, &frame);
+				}
 #endif /* DDB */
 				return;
-			}
+			} else if (panic_on_nmi == 0)
+				return;
 			/* FALL THROUGH */
 #endif /* POWERFAIL_NMI */
 #endif /* NISA > 0 */
diff --git a/sys/amd64/isa/intr_machdep.c b/sys/amd64/isa/intr_machdep.c
index 576b054..34a8c22 100644
--- a/sys/amd64/isa/intr_machdep.c
+++ b/sys/amd64/isa/intr_machdep.c
@@ -229,32 +229,41 @@ int
 isa_nmi(cd)
 	int cd;
 {
+	int retval = 0;
 #ifdef PC98
  	int port = inb(0x33);
+
+	log(LOG_CRIT, "NMI PC98 port = %x\n", port);
 	if (epson_machine_id == 0x20)
 		epson_outb(0xc16, epson_inb(0xc16) | 0x1);
 	if (port & NMI_PARITY) {
-		panic("BASE RAM parity error, likely hardware failure.");
+		log(LOG_CRIT, "BASE RAM parity error, likely hardware failure.");
+		retval = 1;
 	} else if (port & NMI_EPARITY) {
-		panic("EXTENDED RAM parity error, likely hardware failure.");
+		log(LOG_CRIT, "EXTENDED RAM parity error, likely hardware failure.");
+		retval = 1;
 	} else {
-		printf("\nNMI Resume ??\n");
-		return(0);
+		log(LOG_CRIT, "\nNMI Resume ??\n");
 	}
 #else /* IBM-PC */
 	int isa_port = inb(0x61);
 	int eisa_port = inb(0x461);
 
+	log(LOG_CRIT, "NMI ISA %x, EISA %x\n", isa_port, eisa_port);
 #if NMCA > 0
 	if (MCA_system && mca_bus_nmi())
 		return(0);
 #endif
 	
-	if (isa_port & NMI_PARITY)
-		panic("RAM parity error, likely hardware failure.");
+	if (isa_port & NMI_PARITY) {
+		log(LOG_CRIT, "RAM parity error, likely hardware failure.");
+		retval = 1;
+	}
 
-	if (isa_port & NMI_IOCHAN)
-		panic("I/O channel check, likely hardware failure.");
+	if (isa_port & NMI_IOCHAN) {
+		log(LOG_CRIT, "I/O channel check, likely hardware failure.");
+		retval = 1;
+	}
 
 	/*
 	 * On a real EISA machine, this will never happen.  However it can
@@ -262,20 +271,24 @@ isa_nmi(cd)
 	 * error handling (very rare).  Save them from a meaningless panic.
 	 */
 	if (eisa_port == 0xff)
-		return(0);
+		return(retval);
 
-	if (eisa_port & ENMI_WATCHDOG)
-		panic("EISA watchdog timer expired, likely hardware failure.");
-
-	if (eisa_port & ENMI_BUSTIMER)
-		panic("EISA bus timeout, likely hardware failure.");
+	if (eisa_port & ENMI_WATCHDOG) {
+		log(LOG_CRIT, "EISA watchdog timer expired, likely hardware failure.");
+		retval = 1;
+	}
 
-	if (eisa_port & ENMI_IOSTATUS)
-		panic("EISA I/O port status error.");
+	if (eisa_port & ENMI_BUSTIMER) {
+		log(LOG_CRIT, "EISA bus timeout, likely hardware failure.");
+		retval = 1;
+	}
 
-	printf("\nNMI ISA %x, EISA %x\n", isa_port, eisa_port);
-	return(0);
+	if (eisa_port & ENMI_IOSTATUS) {
+		log(LOG_CRIT, "EISA I/O port status error.");
+		retval = 1;
+	}
 #endif
+	return(retval);
 }
 
 /*
diff --git a/sys/amd64/isa/nmi.c b/sys/amd64/isa/nmi.c
index 576b054..34a8c22 100644
--- a/sys/amd64/isa/nmi.c
+++ b/sys/amd64/isa/nmi.c
@@ -229,32 +229,41 @@ int
 isa_nmi(cd)
 	int cd;
 {
+	int retval = 0;
 #ifdef PC98
  	int port = inb(0x33);
+
+	log(LOG_CRIT, "NMI PC98 port = %x\n", port);
 	if (epson_machine_id == 0x20)
 		epson_outb(0xc16, epson_inb(0xc16) | 0x1);
 	if (port & NMI_PARITY) {
-		panic("BASE RAM parity error, likely hardware failure.");
+		log(LOG_CRIT, "BASE RAM parity error, likely hardware failure.");
+		retval = 1;
 	} else if (port & NMI_EPARITY) {
-		panic("EXTENDED RAM parity error, likely hardware failure.");
+		log(LOG_CRIT, "EXTENDED RAM parity error, likely hardware failure.");
+		retval = 1;
 	} else {
-		printf("\nNMI Resume ??\n");
-		return(0);
+		log(LOG_CRIT, "\nNMI Resume ??\n");
 	}
 #else /* IBM-PC */
 	int isa_port = inb(0x61);
 	int eisa_port = inb(0x461);
 
+	log(LOG_CRIT, "NMI ISA %x, EISA %x\n", isa_port, eisa_port);
 #if NMCA > 0
 	if (MCA_system && mca_bus_nmi())
 		return(0);
 #endif
 	
-	if (isa_port & NMI_PARITY)
-		panic("RAM parity error, likely hardware failure.");
+	if (isa_port & NMI_PARITY) {
+		log(LOG_CRIT, "RAM parity error, likely hardware failure.");
+		retval = 1;
+	}
 
-	if (isa_port & NMI_IOCHAN)
-		panic("I/O channel check, likely hardware failure.");
+	if (isa_port & NMI_IOCHAN) {
+		log(LOG_CRIT, "I/O channel check, likely hardware failure.");
+		retval = 1;
+	}
 
 	/*
 	 * On a real EISA machine, this will never happen.  However it can
@@ -262,20 +271,24 @@ isa_nmi(cd)
 	 * error handling (very rare).  Save them from a meaningless panic.
 	 */
 	if (eisa_port == 0xff)
-		return(0);
+		return(retval);
 
-	if (eisa_port & ENMI_WATCHDOG)
-		panic("EISA watchdog timer expired, likely hardware failure.");
-
-	if (eisa_port & ENMI_BUSTIMER)
-		panic("EISA bus timeout, likely hardware failure.");
+	if (eisa_port & ENMI_WATCHDOG) {
+		log(LOG_CRIT, "EISA watchdog timer expired, likely hardware failure.");
+		retval = 1;
+	}
 
-	if (eisa_port & ENMI_IOSTATUS)
-		panic("EISA I/O port status error.");
+	if (eisa_port & ENMI_BUSTIMER) {
+		log(LOG_CRIT, "EISA bus timeout, likely hardware failure.");
+		retval = 1;
+	}
 
-	printf("\nNMI ISA %x, EISA %x\n", isa_port, eisa_port);
-	return(0);
+	if (eisa_port & ENMI_IOSTATUS) {
+		log(LOG_CRIT, "EISA I/O port status error.");
+		retval = 1;
+	}
 #endif
+	return(retval);
 }
 
 /*