[PATCH #upstream-fixes] ahci: give another shot at clearing all bitsin irq_stat

[Tejun Heo]

Commit ea0c62f7cf70f13a67830471b613337bd0c9a62e tried to clear all
bits in irq_stat but it didn't actually achieve that as irq_stat was
anded with port_map right after read.  This patch makes ahci driver
always use the unmasked value to clear irq_status.

While at it, add explanation on the peculiarities of ahci IRQ
clearing.

This was spotted by Linus Torvalds.

Signed-off-by: Tejun Heo <tj@xxxxxxxxxx>
Cc: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
---
The original report requires further investigation regarding why the
bogus change fixed the problem but this change is the correct thing to
do regardless.

Thanks.

 drivers/ata/ahci.c |   16 +++++++++++++---
 1 file changed, 13 insertions(+), 3 deletions(-)

diff --git a/drivers/ata/ahci.c b/drivers/ata/ahci.c
index 061817a..5e6468a 100644
--- a/drivers/ata/ahci.c
+++ b/drivers/ata/ahci.c
@@ -1777,7 +1777,7 @@ static irqreturn_t ahci_interrupt(int irq, void *dev_instance)
 	struct ahci_host_priv *hpriv;
 	unsigned int i, handled = 0;
 	void __iomem *mmio;
-	u32 irq_stat;
+	u32 irq_stat, irq_masked;
 
 	VPRINTK("ENTER\n");
 
@@ -1786,16 +1786,17 @@ static irqreturn_t ahci_interrupt(int irq, void *dev_instance)
 
 	/* sigh.  0xffffffff is a valid return from h/w */
 	irq_stat = readl(mmio + HOST_IRQ_STAT);
-	irq_stat &= hpriv->port_map;
 	if (!irq_stat)
 		return IRQ_NONE;
 
+	irq_masked = irq_stat & hpriv->port_map;
+
 	spin_lock(&host->lock);
 
 	for (i = 0; i < host->n_ports; i++) {
 		struct ata_port *ap;
 
-		if (!(irq_stat & (1 << i)))
+		if (!(irq_masked & (1 << i)))
 			continue;
 
 		ap = host->ports[i];
@@ -1812,6 +1813,15 @@ static irqreturn_t ahci_interrupt(int irq, void *dev_instance)
 		handled = 1;
 	}
 
+	/* HOST_IRQ_STAT behaves as level triggered latch meaning that
+	 * it should be cleared after all the port events are cleared;
+	 * otherwise, it will raise a spurious interrupt after each
+	 * valid one.  Please read section 10.6.2 of ahci 1.1 for more
+	 * information.
+	 *
+	 * Also, use the unmasked value to clear interrupt as spurious
+	 * pending event on a dummy port might cause screaming IRQ.
+	 */
 	writel(irq_stat, mmio + HOST_IRQ_STAT);
 
 	spin_unlock(&host->lock);
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/