[U-Boot] [PATCH v2] sunxi: A64: OHCI: prevent turning off shared USB clock

[Jagan Teki]

On Thu, Jul 5, 2018 at 5:27 AM, Andre Przywara <andre.przywara at arm.com> wrote:
> On the A64 the clock for the first USB controller is actually the parent
> of the clock for the second controller, so turning them off in that order
> makes the system hang.
> Fix this by only turning off *both* clocks when the *last* OHCI controller
> is brought down. This covers the case when only one controller is used.
>
> Signed-off-by: Andre Przywara <andre.przywara at arm.com>
> ---
> Hi,
>
> as requested by Marek, a second version to address the problem of only
> one controller instantiated. I tested all cases:
> - only [EO]HCI1 enabled (current U-Boot master DT)
> - both controllers enabled (mainline Linux DT)
> - only [EO]HCI0 enabled (DT hack)
> In all cases the system booted without hanging, plus I confirmed that
> the USB clocks were disabled in all cases (early in the kernel).
>
> Cheers,
> Andre.
>
> P.S. I found the MMC0, EMAC and USB-OTG AHB gates and resets still running,

Yes MUSB doesn't have shutdown support yet for UCLASS_USB_DEV_GENERIC
so .remove is not able to call. But I've verified this with some
change[1] and it's working as usual. will send it soon.

> but this is an unrelated issue and no regression. Just in case somebody
> feels bored ...
>
>  drivers/usb/host/ohci-sunxi.c | 19 ++++++++++++++++++-
>  1 file changed, 18 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/usb/host/ohci-sunxi.c b/drivers/usb/host/ohci-sunxi.c
> index 0ddbdbe460..bb3c2475df 100644
> --- a/drivers/usb/host/ohci-sunxi.c
> +++ b/drivers/usb/host/ohci-sunxi.c
> @@ -44,6 +44,8 @@ struct ohci_sunxi_priv {
>         const struct ohci_sunxi_cfg *cfg;
>  };
>
> +static fdt_addr_t last_ohci_addr = 0;
> +
>  static int ohci_usb_probe(struct udevice *dev)
>  {
>         struct usb_bus_priv *bus_priv = dev_get_uclass_priv(dev);
> @@ -53,6 +55,9 @@ static int ohci_usb_probe(struct udevice *dev)
>         u8 reg_mask = 0;
>         int phys, ret;
>
> +       if ((fdt_addr_t)regs > last_ohci_addr)
> +               last_ohci_addr = (fdt_addr_t)regs;
> +
>         priv->cfg = (const struct ohci_sunxi_cfg *)dev_get_driver_data(dev);
>         priv->ccm = (struct sunxi_ccm_reg *)SUNXI_CCM_BASE;
>         if (IS_ERR(priv->ccm))
> @@ -114,6 +119,7 @@ no_phy:
>  static int ohci_usb_remove(struct udevice *dev)
>  {
>         struct ohci_sunxi_priv *priv = dev_get_priv(dev);
> +       fdt_addr_t base_addr = devfdt_get_addr(dev);
>         int ret;
>
>         if (generic_phy_valid(&priv->phy)) {
> @@ -130,7 +136,18 @@ static int ohci_usb_remove(struct udevice *dev)
>
>         if (priv->cfg->has_reset)
>                 clrbits_le32(priv->reset0_cfg, priv->ahb_gate_mask);
> -       clrbits_le32(&priv->ccm->usb_clk_cfg, priv->usb_gate_mask);
> +       /*
> +        * On the A64 CLK_USB_OHCI0 is the parent of CLK_USB_OHCI1, so
> +        * we have to wait with bringing down any clock until the last
> +        * OHCI controller is removed.
> +        */
> +       if (!priv->cfg->extra_usb_gate_mask || base_addr == last_ohci_addr) {
> +               u32 usb_gate_mask = priv->usb_gate_mask;
> +
> +               usb_gate_mask |= priv->cfg->extra_usb_gate_mask;
> +               clrbits_le32(&priv->ccm->usb_clk_cfg, usb_gate_mask);
> +       }
> +
>         clrbits_le32(&priv->ccm->ahb_gate0, priv->ahb_gate_mask);
>
>         return 0;
> --

Reviewed-by: Jagan Teki <jagan at openedev.com>
Tested-by: Jagan Teki <jagan at openedev.com>

[1] https://github.com/amarula/u-boot-amarula/commit/401f74ebfb21bb92f793d50e2d3042c19327051c