[libata] sata_mv: fix irq port status usage

Interrupt handler did not properly initialize a variable on a per-port
basis, leading to incorrect behavior on ports other than port 0.

Bug caught and fixed by Mark Lord.

Signed-off-by: Jeff Garzik <jeff@garzik.org>

diff --git a/drivers/scsi/sata_mv.c b/drivers/scsi/sata_mv.c
index aceaf56..e561281 100644 (file)
--- a/drivers/scsi/sata_mv.c
+++ b/drivers/scsi/sata_mv.c
@@ -1353,7 +1353,6 @@ static void mv_host_intr(struct ata_host_set *host_set, u32 relevant,
        u32 hc_irq_cause;
        int shift, port, port0, hard_port, handled;
        unsigned int err_mask;
-       u8 ata_status = 0;
 
        if (hc == 0) {
                port0 = 0;
@@ -1371,6 +1370,7 @@ static void mv_host_intr(struct ata_host_set *host_set, u32 relevant,
                hc,relevant,hc_irq_cause);
 
        for (port = port0; port < port0 + MV_PORTS_PER_HC; port++) {
+               u8 ata_status = 0;
                ap = host_set->ports[port];
                hard_port = port & MV_PORT_MASK;        /* range 0-3 */
                handled = 0;    /* ensure ata_status is set if handled++ */