On 8 June 2015 at 10:24, Rafał Miłecki <zajec5@xxxxxxxxx> wrote:
> On 2 June 2015 at 21:35, Hauke Mehrtens <hauke@xxxxxxxxxx> wrote:
>> @@ -159,6 +160,55 @@ static void bcma_hcd_init_chip(struct bcma_device *dev)
>> }
>> }
>>
>> +static void bcma_hcd_init_chip_arm(struct bcma_device *dev)
>> +{
>> + struct bcma_device *arm_core;
>> + void __iomem *dmu;
>> + u32 val;
>> +
>> + bcma_core_enable(dev, 0);
>> +
>> + usleep_range(1000, 2000);
>> +
>> + /* Set packet buffer OUT threshold */
>> + val = bcma_read32(dev, 0x94);
>> + val &= 0xffff;
>> + val |= 0x80 << 16;
>> + bcma_write32(dev, 0x94, val);
>> +
>> + /* Enable break memory transfer */
>> + val = bcma_read32(dev, 0x9c);
>> + val |= 1;
>> + bcma_write32(dev, 0x9c, val);
>> +
>> + if (dev->bus->chipinfo.pkg != BCMA_PKG_ID_BCM4707 &&
>> + dev->bus->chipinfo.pkg != BCMA_PKG_ID_BCM4708)
>> + return;
>> +
>> + arm_core = bcma_find_core(dev->bus, BCMA_CORE_ARMCA9);
>> + if (!arm_core) {
>> + dev_err(&dev->dev, "can not find ARM Cortex A9 ihost core\n");
>> + return;
>> + }
>> +
>> + dmu = ioremap_nocache(arm_core->addr_s[0], 0x1000);
>> + if (!dmu) {
>> + dev_err(&dev->dev, "can not map ARM Cortex A9 ihost core\n");
>> + return;
>> + }
>> +
>> + /* Unlock DMU PLL settings */
>> + iowrite32(0x0000ea68, dmu + 0x180);
>> +
>> + /* Write USB 2.0 PLL control setting */
>> + iowrite32(0x00dd10c3, dmu + 0x164);
>> +
>> + /* Lock DMU PLL settings */
>> + iowrite32(0x00000000, dmu + 0x180);
>> +
>> + iounmap(dmu);
>> +}
>
> I started comparing above changes to Broadcom's reference code and I
> see some potential problems.
>
> 1) You check package only
> Broadcom keeps developing new SoCs so it may make sense to check both:
> id and pkg. I can imagine Broadcom re-using BCMA_CORE_NS_USB20 on some
> other SoC requiring different programming.
> So I guess you could check for BCMA_CHIP_ID_BCM4707 ||
> BCMA_CHIP_ID_BCM53018. just to be safe.
>
> 2) Init order
> I believe you should init PHY first (3 iowrite32 ops), then wait for
> it to setup itself with (usleep_range) and then finally init
> controller (bcma_[read|write]).
>
> 3) Controller init
> It should be done on every package, not just BCM4707 and BCM4708.
Oh, wait, I just noticed you check chip ID before caling
bcma_hcd_init_chip_arm. So I guess you could just reorder the
initialization and it should be fine by then.
--
Rafał
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
