[PATCH] fix ras irq handlers

Paul Mackerras paulus at samba.org
Fri Jul 2 21:22:24 EST 2004


Nathan,

I just dug out a patch that you sent a while ago that made some
changes to ras.c, mainly to cater for the possibility of epow-events
and internal-errors having multiple elements.  I have reworked it -
let me know what you think of the patch below.

One of the changes I have made is to check the #interrupt-cells
applying to the interrupts property.  Without this I think we will
incorrectly try to get interrupt 0 or 1, since when #interrupt-cells
is 2, the second cell is the edge/level indicator.

Paul.

diff -urN test25/arch/ppc64/kernel/prom.c ppc64-2.5-pseries/arch/ppc64/kernel/prom.c
--- test25/arch/ppc64/kernel/prom.c	2004-06-30 22:00:47.000000000 +1000
+++ ppc64-2.5-pseries/arch/ppc64/kernel/prom.c	2004-07-02 17:06:51.000000000 +1000
@@ -1881,8 +1881,7 @@
  * Find out the size of each entry of the interrupts property
  * for a node.
  */
-static int __devinit
-prom_n_intr_cells(struct device_node *np)
+int __devinit prom_n_intr_cells(struct device_node *np)
 {
 	struct device_node *p;
 	unsigned int *icp;
@@ -1896,7 +1895,7 @@
 		    || get_property(p, "interrupt-map", NULL) != NULL) {
 			printk("oops, node %s doesn't have #interrupt-cells\n",
 			       p->full_name);
-		return 1;
+			return 1;
 		}
 	}
 #ifdef DEBUG_IRQ
diff -urN ppc64-linux-2.5/arch/ppc64/kernel/ras.c ppc64-2.5-pseries/arch/ppc64/kernel/ras.c
--- ppc64-linux-2.5/arch/ppc64/kernel/ras.c	2004-04-13 14:04:32.000000000 +1000
+++ ppc64-2.5-pseries/arch/ppc64/kernel/ras.c	2004-07-02 21:07:28.932004888 +1000
@@ -52,6 +52,16 @@
 #include <asm/rtas.h>
 #include <asm/ppcdebug.h>

+static unsigned char log_buf[RTAS_ERROR_LOG_MAX];
+static spinlock_t log_lock = SPIN_LOCK_UNLOCKED;
+
+static int ras_get_sensor_state_token;
+static int ras_check_exception_token;
+
+#define EPOW_SENSOR_TOKEN	9
+#define EPOW_SENSOR_INDEX	0
+#define RAS_VECTOR_OFFSET	0x500
+
 static irqreturn_t ras_epow_interrupt(int irq, void *dev_id,
 					struct pt_regs * regs);
 static irqreturn_t ras_error_interrupt(int irq, void *dev_id,
@@ -59,6 +69,35 @@

 /* #define DEBUG */

+static void request_ras_irqs(struct device_node *np, char *propname,
+			irqreturn_t (*handler)(int, void *, struct pt_regs *),
+			const char *name)
+{
+	unsigned int *ireg, len, i;
+	int virq, n_intr;
+
+	ireg = (unsigned int *)get_property(np, propname, &len);
+	if (ireg == NULL)
+		return;
+	n_intr = prom_n_intr_cells(np);
+	len /= n_intr * sizeof(*ireg);
+
+	for (i = 0; i < len; i++) {
+		virq = virt_irq_create_mapping(*ireg);
+		if (virq == NO_IRQ) {
+			printk(KERN_ERR "Unable to allocate interrupt "
+			       "number for %s\n", np->full_name);
+			return;
+		}
+		if (request_irq(irq_offset_up(virq), handler, 0, name, NULL)) {
+			printk(KERN_ERR "Unable to request interrupt %d for "
+			       "%s\n", irq_offset_up(virq), np->full_name);
+			return;
+		}
+		ireg += n_intr;
+	}
+}
+
 /*
  * Initialize handlers for the set of interrupts caused by hardware errors
  * and power system events.
@@ -66,52 +105,33 @@
 static int __init init_ras_IRQ(void)
 {
 	struct device_node *np;
-	unsigned int *ireg, len, i;
-	int virq;

-	if ((np = of_find_node_by_path("/event-sources/internal-errors")) &&
-	    (ireg = (unsigned int *)get_property(np, "open-pic-interrupt",
-						 &len))) {
-		for (i=0; i<(len / sizeof(*ireg)); i++) {
-			virq = virt_irq_create_mapping(*(ireg));
-			if (virq == NO_IRQ) {
-				printk(KERN_ERR "Unable to allocate interrupt "
-				       "number for %s\n", np->full_name);
-				break;
-			}
-			request_irq(irq_offset_up(virq),
-				    ras_error_interrupt, 0,
-				    "RAS_ERROR", NULL);
-			ireg++;
-		}
+	ras_get_sensor_state_token = rtas_token("get-sensor-state");
+	ras_check_exception_token = rtas_token("check-exception");
+
+	/* Internal Errors */
+	np = of_find_node_by_path("/event-sources/internal-errors");
+	if (np != NULL) {
+		request_ras_irqs(np, "open-pic-interrupt", ras_error_interrupt,
+				 "RAS_ERROR");
+		request_ras_irqs(np, "interrupts", ras_error_interrupt,
+				 "RAS_ERROR");
+		of_node_put(np);
 	}
-	of_node_put(np);

-	if ((np = of_find_node_by_path("/event-sources/epow-events")) &&
-	    (ireg = (unsigned int *)get_property(np, "open-pic-interrupt",
-						 &len))) {
-		for (i=0; i<(len / sizeof(*ireg)); i++) {
-			virq = virt_irq_create_mapping(*(ireg));
-			if (virq == NO_IRQ) {
-				printk(KERN_ERR "Unable to allocate interrupt "
-				       " number for %s\n", np->full_name);
-				break;
-			}
-			request_irq(irq_offset_up(virq),
-				    ras_epow_interrupt, 0,
-				    "RAS_EPOW", NULL);
-			ireg++;
-		}
+	np = of_find_node_by_path("/event-sources/epow-events");
+	if (np != NULL) {
+		request_ras_irqs(np, "open-pic-interrupt", ras_epow_interrupt,
+				"RAS_EPOW");
+		request_ras_irqs(np, "interrupts", ras_epow_interrupt,
+				"RAS_EPOW");
+		of_node_put(np);
 	}
-	of_node_put(np);

 	return 1;
 }
 __initcall(init_ras_IRQ);

-static struct rtas_error_log log_buf;
-static spinlock_t log_lock = SPIN_LOCK_UNLOCKED;
-
 /*
  * Handle power subsystem events (EPOW).
  *
@@ -122,30 +142,35 @@
 static irqreturn_t
 ras_epow_interrupt(int irq, void *dev_id, struct pt_regs * regs)
 {
-	struct rtas_error_log log_entry;
-	unsigned int size = sizeof(log_entry);
-	long status = 0xdeadbeef;
+	int status = 0xdeadbeef;
+	int state = 0;
+	int virq = irq_offset_down(irq);
+	int critical;

 	spin_lock(&log_lock);

-	status = rtas_call(rtas_token("check-exception"), 6, 1, NULL,
-			   0x500, irq,
-			   RTAS_EPOW_WARNING | RTAS_POWERMGM_EVENTS,
-			   1,  /* Time Critical */
-			   __pa(&log_buf), size);
+	status = rtas_call(ras_get_sensor_state_token, 2, 2, &state,
+			   EPOW_SENSOR_TOKEN, EPOW_SENSOR_INDEX);

-	log_entry = log_buf;
-
-	spin_unlock(&log_lock);
+	if (state > 3)
+		critical = 1;  /* Time Critical */
+	else
+		critical = 0;

-	udbg_printf("EPOW <0x%lx 0x%lx>\n",
-		    *((unsigned long *)&log_entry), status);
-	printk(KERN_WARNING
-		"EPOW <0x%lx 0x%lx>\n",*((unsigned long *)&log_entry), status);
+	status = rtas_call(ras_check_exception_token, 6, 1, NULL,
+			   RAS_VECTOR_OFFSET, virt_irq_to_real(virq),
+			   RTAS_EPOW_WARNING | RTAS_POWERMGM_EVENTS,
+			   critical, __pa(&log_buf), RTAS_ERROR_LOG_MAX);
+
+	udbg_printf("EPOW <0x%lx 0x%x 0x%x>\n",
+		    *((unsigned long *)&log_buf), status, state);
+	printk(KERN_WARNING "EPOW <0x%lx 0x%x 0x%x>\n",
+	       *((unsigned long *)&log_buf), status, state);

 	/* format and print the extended information */
-	log_error((char *)&log_entry, ERR_TYPE_RTAS_LOG, 0);
-
+	log_error(log_buf, ERR_TYPE_RTAS_LOG, 0);
+
+	spin_unlock(&log_lock);
 	return IRQ_HANDLED;
 }

@@ -160,37 +185,34 @@
 static irqreturn_t
 ras_error_interrupt(int irq, void *dev_id, struct pt_regs * regs)
 {
-	struct rtas_error_log log_entry;
-	unsigned int size = sizeof(log_entry);
-	long status = 0xdeadbeef;
+	struct rtas_error_log *rtas_elog;
+	int status = 0xdeadbeef;
 	int fatal;

 	spin_lock(&log_lock);

-	status = rtas_call(rtas_token("check-exception"), 6, 1, NULL,
-			   0x500, irq,
-			   RTAS_INTERNAL_ERROR,
-			   1, /* Time Critical */
-			   __pa(&log_buf), size);
-
-	log_entry = log_buf;
+	status = rtas_call(ras_check_exception_token, 6, 1, NULL,
+			   RAS_VECTOR_OFFSET,
+			   virt_irq_to_real(irq_offset_down(irq)),
+			   RTAS_INTERNAL_ERROR,
+			   1 /* Time Critical */,
+			   __pa(&log_buf), RTAS_ERROR_LOG_MAX);

-	spin_unlock(&log_lock);
+	rtas_elog = (struct rtas_error_log *)log_buf;

-	if ((status == 0) && (log_entry.severity >= SEVERITY_ERROR_SYNC))
+	if ((status == 0) && (rtas_elog->severity >= SEVERITY_ERROR_SYNC))
 		fatal = 1;
 	else
 		fatal = 0;

 	/* format and print the extended information */
-	log_error((char *)&log_entry, ERR_TYPE_RTAS_LOG, fatal);
+	log_error(log_buf, ERR_TYPE_RTAS_LOG, fatal);

 	if (fatal) {
-		udbg_printf("HW Error <0x%lx 0x%lx>\n",
-			    *((unsigned long *)&log_entry), status);
-		printk(KERN_EMERG
-		       "Error: Fatal hardware error <0x%lx 0x%lx>\n",
-		       *((unsigned long *)&log_entry), status);
+		udbg_printf("HW Error <0x%lx 0x%x>\n",
+			    *((unsigned long *)&log_buf), status);
+		printk(KERN_EMERG "Error: Fatal hardware error <0x%lx 0x%x>\n",
+		       *((unsigned long *)&log_buf), status);

 #ifndef DEBUG
 		/* Don't actually power off when debugging so we can test
@@ -200,11 +222,13 @@
 		ppc_md.power_off();
 #endif
 	} else {
-		udbg_printf("Recoverable HW Error <0x%lx 0x%lx>\n",
-			    *((unsigned long *)&log_entry), status);
+		udbg_printf("Recoverable HW Error <0x%lx 0x%x>\n",
+			    *((unsigned long *)&log_buf), status);
 		printk(KERN_WARNING
-		       "Warning: Recoverable hardware error <0x%lx 0x%lx>\n",
-		       *((unsigned long *)&log_entry), status);
+		       "Warning: Recoverable hardware error <0x%lx 0x%x>\n",
+		       *((unsigned long *)&log_buf), status);
 	}
+
+	spin_unlock(&log_lock);
 	return IRQ_HANDLED;
 }
diff -urN test25/include/asm-ppc64/prom.h ppc64-2.5-pseries/include/asm-ppc64/prom.h
--- test25/include/asm-ppc64/prom.h	2004-06-24 21:46:45.000000000 +1000
+++ ppc64-2.5-pseries/include/asm-ppc64/prom.h	2004-07-02 17:06:22.000000000 +1000
@@ -269,6 +269,7 @@
 extern void print_properties(struct device_node *node);
 extern int prom_n_addr_cells(struct device_node* np);
 extern int prom_n_size_cells(struct device_node* np);
+extern int prom_n_intr_cells(struct device_node* np);
 extern void prom_get_irq_senses(unsigned char *senses, int off, int max);
 extern void prom_add_property(struct device_node* np, struct property* prop);


** Sent via the linuxppc64-dev mail list. See http://lists.linuxppc.org/





More information about the Linuxppc64-dev mailing list