On Fri, May 09, 2014 at 06:47:22PM +0200, Boris BREZILLON wrote:

On 09/05/2014 18:03, Ezequiel Garcia wrote:

quoted

On 12 Mar 07:07 PM, Boris BREZILLON wrote:

quoted

--- /dev/null
+++ b/drivers/mtd/nand/sunxi_nand.c

@@ -0,0 +1,1276 @@

...

quoted

quoted

+static int sunxi_nand_ecc_init(struct mtd_info *mtd, struct nand_ecc_ctrl *ecc,
+			       struct device_node *np)
+{
+	struct nand_chip *nand = mtd->priv;
+	int ecc_step_size, ecc_strength;
+	int ret;
+
+	ecc_step_size = of_get_nand_ecc_step_size(np);
+	ecc_strength = of_get_nand_ecc_strength(np);
+	if (ecc_step_size > 0 && ecc_strength > 0) {
+		ecc->size = ecc_step_size;
+		ecc->strength = ecc_strength;
+	} else {
+		ecc->size = nand->ecc_step_ds;
+		ecc->strength = nand->ecc_strength_ds;
+	}
+

Shouldn't you check the devicetree value is not weaker than the ONFI-obtained?

I can definitely do that.

You can do that here, but take a look at the discussion Ezequiel and I
had about this:

  http://thread.gmane.org/gmane.linux.drivers.devicetree/67462

We probably don't want to be doing anything drastic like overriding the
device tree configuration. Instead, we might want to stick a warning
into the core nand_base code that does the proper comparison of the
'*_ds' values with the actual values chosen in
chip->ecc->{size,strength}. The comparison is kind of subtle, actually,
so it'd be good to do it exactly once for everyone.

Brian