powerpc/powernv: Improve kexec reliability
We add a machine_shutdown hook that frees the OPAL interrupts
(so they get masked at the source and don't fire while kexec'ing)
and which triggers an IODA reset on all the PCIe host bridges
which will have the effect of blocking all DMAs and subsequent
PCIs interrupts.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
diff --git a/arch/powerpc/platforms/powernv/opal.c b/arch/powerpc/platforms/powernv/opal.c
index 12d9846..27907cb 100644
--- a/arch/powerpc/platforms/powernv/opal.c
+++ b/arch/powerpc/platforms/powernv/opal.c
@@ -15,6 +15,7 @@
#include <linux/of.h>
#include <linux/of_platform.h>
#include <linux/interrupt.h>
+#include <linux/slab.h>
#include <asm/opal.h>
#include <asm/firmware.h>
@@ -28,6 +29,8 @@
static struct device_node *opal_node;
static DEFINE_SPINLOCK(opal_write_lock);
extern u64 opal_mc_secondary_handler[];
+static unsigned int *opal_irqs;
+static unsigned int opal_irq_count;
int __init early_init_dt_scan_opal(unsigned long node,
const char *uname, int depth, void *data)
@@ -323,6 +326,8 @@
irqs = of_get_property(opal_node, "opal-interrupts", &irqlen);
pr_debug("opal: Found %d interrupts reserved for OPAL\n",
irqs ? (irqlen / 4) : 0);
+ opal_irq_count = irqlen / 4;
+ opal_irqs = kzalloc(opal_irq_count * sizeof(unsigned int), GFP_KERNEL);
for (i = 0; irqs && i < (irqlen / 4); i++, irqs++) {
unsigned int hwirq = be32_to_cpup(irqs);
unsigned int irq = irq_create_mapping(NULL, hwirq);
@@ -334,7 +339,19 @@
if (rc)
pr_warning("opal: Error %d requesting irq %d"
" (0x%x)\n", rc, irq, hwirq);
+ opal_irqs[i] = irq;
}
return 0;
}
subsys_initcall(opal_init);
+
+void opal_shutdown(void)
+{
+ unsigned int i;
+
+ for (i = 0; i < opal_irq_count; i++) {
+ if (opal_irqs[i])
+ free_irq(opal_irqs[i], 0);
+ opal_irqs[i] = 0;
+ }
+}