On Tue, 2018-01-16 at 11:36 +0300, Sergei Shtylyov wrote:
> Hello!
>
> On 1/16/2018 12:53 AM, Benjamin Herrenschmidt wrote:
>
> > > > The Aspeed SoCs use uhci-platform. With the new dynamic clock
> > > > control framework, the corresponding IP block clock must be
> > > > properly enabled.
> > > >
> > > > This is a simplified variant of what ehci-platform does, it
> > > > looks for *one* clock attached to the device, and if it's
> > > > there, enables it.
> > > >
> > > > Signed-off-by: Benjamin Herrenschmidt <benh@xxxxxxxxxxxxxxxxxxx>
> > > > Acked-by: Alan Stern <stern@xxxxxxxxxxxxxxxxxxx>
> > > > ---
> > > >
> > > > v4. Remove unnecessary NULL checks
> > > >
> > > > v3. Fix tab vs. spaces and probe error path
> > > >
> > > > v2. Don't forget to git add latest changes :-) This adds the
> > > > part where we turn the clock off when removing the driver.
> > > > ---
> > > > drivers/usb/host/uhci-hcd.h | 3 +++
> > > > drivers/usb/host/uhci-platform.c | 20 ++++++++++++++++++--
> > > > 2 files changed, 21 insertions(+), 2 deletions(-)
> > > >
> > > > diff --git a/drivers/usb/host/uhci-hcd.h b/drivers/usb/host/uhci-hcd.h
> > > > index f1cc47292a59..7f9f33c8c232 100644
> > > > --- a/drivers/usb/host/uhci-hcd.h
> > > > +++ b/drivers/usb/host/uhci-hcd.h
> > > > @@ -4,6 +4,7 @@
> > > >
> > > > #include <linux/list.h>
> > > > #include <linux/usb.h>
> > > > +#include <linux/clk.h>
> > > >
> > > > #define usb_packetid(pipe) (usb_pipein(pipe) ? USB_PID_IN : USB_PID_OUT)
> > > > #define PIPE_DEVEP_MASK 0x0007ff00
> > > > @@ -447,6 +448,8 @@ struct uhci_hcd {
> > > > int total_load; /* Sum of array values */
> > > > short load[MAX_PHASE]; /* Periodic allocations */
> > > >
> > > > + struct clk *clk; /* (optional) clock source */
> > > > +
> > > > /* Reset host controller */
> > > > void (*reset_hc) (struct uhci_hcd *uhci);
> > > > int (*check_and_reset_hc) (struct uhci_hcd *uhci);
> > > > diff --git a/drivers/usb/host/uhci-platform.c b/drivers/usb/host/uhci-platform.c
> > > > index 6cb16d4b2257..55c2d99bf7ec 100644
> > > > --- a/drivers/usb/host/uhci-platform.c
> > > > +++ b/drivers/usb/host/uhci-platform.c
> > > > @@ -89,6 +89,8 @@ static int uhci_hcd_platform_probe(struct platform_device *pdev)
> > > > if (!hcd)
> > > > return -ENOMEM;
> > > >
> > > > + uhci = hcd_to_uhci(hcd);
> > > > +
> > > > res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> > > > hcd->regs = devm_ioremap_resource(&pdev->dev, res);
> > > > if (IS_ERR(hcd->regs)) {
> > > > @@ -98,8 +100,6 @@ static int uhci_hcd_platform_probe(struct platform_device *pdev)
> > > > hcd->rsrc_start = res->start;
> > > > hcd->rsrc_len = resource_size(res);
> > > >
> > > > - uhci = hcd_to_uhci(hcd);
> > > > -
> > > > uhci->regs = hcd->regs;
> > > >
> > > > /* Grab some things from the device-tree */
> > > > @@ -119,6 +119,19 @@ static int uhci_hcd_platform_probe(struct platform_device *pdev)
> > > > "Enabled Aspeed implementation workarounds\n");
> > > > }
> > > > }
> > > > +
> > > > + /* Get and enable clock if any specified */
> > > > + uhci->clk = devm_clk_get(&pdev->dev, NULL);
> > > > + if (IS_ERR(uhci->clk)) {
> > > > + ret = PTR_ERR(uhci->clk);
> > > > + goto err_rmr;
> > > > + }
> > > > + ret = clk_prepare_enable(uhci->clk);
> > > > + if (ret) {
> > > > + dev_err(&pdev->dev, "Error couldn't enable clock (%d)\n", ret);
> > > > + goto err_rmr;
> > > > + }
> > > > +
> > > > ret = usb_add_hcd(hcd, pdev->resource[1].start, IRQF_SHARED);
> > > > if (ret)
> > > > goto err_rmr;
> > > > @@ -127,6 +140,7 @@ static int uhci_hcd_platform_probe(struct platform_device *pdev)
> > > > return 0;
> > > >
> > > > err_rmr:
> > > > + clk_disable_unprepare(uhci->clk);
> > >
> > > Really? You haven't enabled the clock before either *goto* above, why
> > > disable it here?
> >
> > If clk is NULL or an error, disable will do nothing. The only case
>
> The rule of thumb was that you don't pass the error pointers to APIs but
> indeed I'm now seeing the NULL check extended to the error pointers too...
>
> > would be if enable at failed but then disable will do nothing in that
> > case.
>
> Are you sure there'll be no WARN_ON() triggered by clk_disable() with the
> enable count == 0?
There is, and while we won't hit that on Aspeed as it won't fail, we
should probably make it right.
> > The above change was done at the suggestion of Alan. If you really
>
> I'd probably missed his mail.
>
> > prefer I could add a clk_put and clearing of "clk" in the error path of
> > clk_prepare_enable, but that's about it.
I think the best is separate goto labels. I'll respin tomorrow.
Cheers,
Ben.
--
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
