On Thu, Dec 09, 2010 at 09:54:01AM +0530, Rupjyoti Sarmah wrote:

A few error path issues, plus one leak in the good case.  I'll wait to
see if there are any more comments for a couple of days.

+static int __init ppc460ex_canyonlands_fixup(void)
+{
+	u8 __iomem *bcsr ;
+	void __iomem *vaddr;
+	struct device_node *np;
+	int ret = 0;
+
+	np = of_find_compatible_node(NULL, NULL, "amcc,ppc460ex-bcsr");
+	if (!np) {
+		printk(KERN_ERR "failed did not find amcc, ppc460ex bcsr node\n");
+		return -ENODEV;
+	}
+
+	bcsr = of_iomap(np, 0);
+	of_node_put(np);
+
+	if (!bcsr) {
+		printk(KERN_CRIT "Could not remap bcsr\n");
+		ret = -ENODEV;
+	}
+
+	np = of_find_compatible_node(NULL, NULL, "ibm,ppc4xx-gpio");

This can fail if you boot a new kernel on a board with an old FDT that
lacks the gpio node.  Calling of_iomap with a NULL np seems like it
would work without crashing anything, so maybe that's not a huge issue.

+	vaddr = of_iomap(np, 0);
+	if (!vaddr) {
+		printk(KERN_CRIT "Could not get gpio node address\n");
+		ret = -ENODEV;
+		goto err_bcsr;
+	}

A reference count is leaked for the node in this error case because
of_node_put isn't called for the gpio node.  Also, iounmap is never
called on vaddr.

+	/* Disable USB, through the BCSR7 bits */
+	setbits8(&bcsr[7], BCSR_USB_EN);
+
+	/* Wait for a while after reset */
+	msleep(100);
+
+	/* Enable USB here */
+	clrbits8(&bcsr[7], BCSR_USB_EN);
+
+	/*
+	 * Configure multiplexed gpio16 and gpio19 as alternate1 output
+	 * source after USB reset. In this configuration gpio16 will be
+	 * USB2HStop and gpio19 will be USB2DStop. For more details refer to
+	 * table 34-7 of PPC460EX user manual.
+	 */
+	setbits32((vaddr + GPIO0_OSRH), 0x42000000);
+	setbits32((vaddr + GPIO0_TSRH), 0x42000000);
+	of_node_put(np);
+err_bcsr:
+	iounmap(bcsr);
+	return ret;
+}
+machine_device_initcall(canyonlands, ppc460ex_canyonlands_fixup);

josh