[PATCH v3 05/11] memory: add Atmel EBI (External Bus Interface) driver

On Monday 15 December 2014 11:22:30, Boris Brezillon wrote:

Hi Alexander,

On Tue, 09 Dec 2014 21:53:12 +0100
Alexander Stein [off-list ref] wrote:

quoted

Hi,

On Monday 01 December 2014, 11:27:21 wrote Boris Brezillon:

quoted

+static int at91_ebi_probe(struct platform_device *pdev)
+{
+	const struct of_device_id *match;
+	struct device_node *child;
+	struct device_node *np;
+	struct at91_ebi *ebi;
+	struct clk *clk;
+	int ret;
+
+	match = of_match_device(at91_ebi_id_table, &pdev->dev);
+	if (!match || !match->data)
+		return -EINVAL;
+
+	ebi = devm_kzalloc(&pdev->dev, sizeof(*ebi), GFP_KERNEL);
+	if (!ebi)
+		return -ENOMEM;
+
+	ebi->caps = match->data;
+	ebi->dev = &pdev->dev;
+
+	clk = devm_clk_get(&pdev->dev, NULL);
+	if (IS_ERR(clk))
+		return PTR_ERR(clk);
+
+	ebi->clk = clk;
+
+	np = of_parse_phandle(pdev->dev.of_node, "atmel,smc", 0);
+	if (!np)
+		return -EINVAL;
+
+	ebi->smc = syscon_node_to_regmap(np);
+	if (IS_ERR(ebi->smc))
+		return PTR_ERR(ebi->smc);
+
+	/*
+	 * The sama5d3 does not provide an EBICSA register and thus does need
+	 * to access the matrix registers.
+	 */
+	if (ebi->caps->ebi_csa) {
+		np = of_parse_phandle(pdev->dev.of_node, "atmel,matrix", 0);
+		if (np)

         ^^^^^^^
Shouldn't this be if (!np)?

Absolutely, thanks for pointing this out.
I'll test on other at91 platforms (I currently only test on sama5d3
boards, which do not have this EBICSA register) before submitting a new
version.

Oh, well. I've noticed you posted a v4 of this patch which changed this specific code. I didn't noticed this.
I'll replace the old v3 and retry. I'm using this code on 9263 so far configuring EBI1 for a PSRAM. Works so far.
Once I exchanged patch 05/11 with your v4 I'll send my DT changes for 9263 to support this feature.

Best regards,
Alexander