[SPARC64]: Do not ACK an INO if it is disabled or inprogress.
This is also a partial workaround for a bug in the LDOM firmware which
double-transmits RX inos during high load. Without this, such an
event causes the kernel to loop forever in the interrupt call chain
ACK'ing but never actually running the IRQ handler (and thus clearing
the interrupt condition in the device).
There is still a bad potential effect when double INOs occur,
not covered by this changeset. Namely, if the INO is already on
the per-cpu INO vector list, we still blindly re-insert it and
thus we can end up losing interrupts already linked in after
it.
We could deal with that by traversing the list before insertion,
but that's too expensive for this edge case.
Signed-off-by: David S. Miller <davem@davemloft.net>
diff --git a/arch/sparc64/kernel/irq.c b/arch/sparc64/kernel/irq.c
index 6b6165d3..634194e 100644
--- a/arch/sparc64/kernel/irq.c
+++ b/arch/sparc64/kernel/irq.c
@@ -309,6 +309,10 @@
static void sun4u_irq_end(unsigned int virt_irq)
{
struct irq_handler_data *data = get_irq_chip_data(virt_irq);
+ struct irq_desc *desc = irq_desc + virt_irq;
+
+ if (unlikely(desc->status & (IRQ_DISABLED|IRQ_INPROGRESS)))
+ return;
if (likely(data))
upa_writeq(ICLR_IDLE, data->iclr);
@@ -373,6 +377,10 @@
{
struct ino_bucket *bucket = virt_irq_to_bucket(virt_irq);
unsigned int ino = bucket - &ivector_table[0];
+ struct irq_desc *desc = irq_desc + virt_irq;
+
+ if (unlikely(desc->status & (IRQ_DISABLED|IRQ_INPROGRESS)))
+ return;
if (likely(bucket)) {
int err;
@@ -443,6 +451,10 @@
{
struct ino_bucket *bucket = virt_irq_to_bucket(virt_irq);
unsigned int ino = bucket - &ivector_table[0];
+ struct irq_desc *desc = irq_desc + virt_irq;
+
+ if (unlikely(desc->status & (IRQ_DISABLED|IRQ_INPROGRESS)))
+ return;
if (likely(bucket)) {
unsigned long dev_handle, dev_ino;