On Monday 14 December 2009 11:06:34 pm Jeff Garzik wrote:

On 12/14/2009 04:40 PM, Ryan Mallon wrote:

quoted

I have added some of my own debugging. The problem appears to be that
__pata_ep93xx_write gets called from probe (via ata_host_activate), but
ap->private_data (ata_timing) is still null. The timing private_data is
set by pata_ep93xx_set_piomode, but that needs adev->pio_mode set, but I
don't know where this happens. I assume the ATA core handles this. Do I
need to call pata_ep93xx_set_piomode from pata_ep93xx_probe before
ata_host_activate, or should the private_data timing be set to some
default in the probe?

ap->private_data is traditionally initialized in the ->port_start() hook.

Ryan, this incremental patch should fix the OOPS (it adds ->port_start
implementation which sets the default 'port PIO mode' to PIO0 so we have
sane default setting before device probing takes place and ->set_piomode
takes over), please retest the driver once you have a minute..

[ Thanks for debugging the problem and sorry for the delay.. ]

---
 drivers/ata/pata_ep93xx.c |   11 +++++++++++
 1 file changed, 11 insertions(+)

Index: b/drivers/ata/pata_ep93xx.c
===================================================================

--- a/drivers/ata/pata_ep93xx.c
+++ b/drivers/ata/pata_ep93xx.c

@@ -613,6 +613,15 @@ static void pata_ep93xx_postreset(struct
 	}
 }
 
+static int pata_ep93xx_port_start(struct ata_port *ap)
+{
+	const struct ata_timing *cmd_t = ata_timing_find_mode(XFER_PIO_0);
+
+	ap->private_data = (void *)cmd_t;
+
+	return ata_sff_port_start(ap);
+}
+
 static struct ata_port_operations pata_ep93xx_port_ops = {
 	.inherits		= &ata_sff_port_ops,
 

@@ -631,6 +640,8 @@ static struct ata_port_operations pata_e
 	.freeze			= pata_ep93xx_freeze,
 	.softreset		= pata_ep93xx_softreset,
 	.postreset		= pata_ep93xx_postreset,
+
+	.port_start		= pata_ep93xx_port_start,
 };
 
 static int __init pata_ep93xx_probe(struct platform_device *pdev)