lkml.org 
[lkml]   [2015]   [Apr]   [7]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 2/9][v2]usb:fsl:otg: Add support to add/remove usb host driver
On Tue, 7 Apr 2015, Ramneek Mehresh wrote:

> Add workqueue to add/remove host driver (outside interrupt context)
> upon each id change

This patch needs to be fixed. See below.

> Signed-off-by: Ramneek Mehresh <ramneek.mehresh@freescale.com>
> ---
> drivers/usb/host/ehci-fsl.c | 107 ++++++++++++++++++++++++++++++++++++--------
> drivers/usb/host/ehci.h | 1 -
> 2 files changed, 88 insertions(+), 20 deletions(-)
>
> diff --git a/drivers/usb/host/ehci-fsl.c b/drivers/usb/host/ehci-fsl.c
> index ab4eee3..9c91fbe 100644
> --- a/drivers/usb/host/ehci-fsl.c
> +++ b/drivers/usb/host/ehci-fsl.c
> @@ -33,6 +33,58 @@
>
> #include "ehci-fsl.h"
>
> +struct ehci_fsl {
> + struct ehci_hcd ehci;
> +
> +#ifdef CONFIG_PM
> + /* Saved USB PHY settings, need to restore after deep sleep. */
> + u32 usb_ctrl;
> +#if defined(CONFIG_FSL_USB2_OTG) || defined(CONFIG_FSL_USB2_OTG_MODULE)
> + struct work_struct change_hcd_work;
> +#endif
> +#endif
> +
> + /* store current hcd state for otg;
> + * have_hcd is true when host drv al already part of otg framework,
> + * otherwise false;
> + * hcd_add is true when otg framework wants to add host
> + * drv as part of otg;flase when it wants to remove it
> + */
> + unsigned have_hcd:1;
> + unsigned hcd_add:1;
> +};

This is the wrong way to connect platform-private data with the
ehci_hcd structure. The right way is to use an ehci_driver_overrides
structure, like in ehci-omap.c and quite a few other EHCI platform
drivers.

> +#if defined(CONFIG_FSL_USB2_OTG) || defined(CONFIG_FSL_USB2_OTG_MODULE)
> +static struct ehci_fsl *hcd_to_ehci_fsl(struct usb_hcd *hcd)
> +{
> + struct ehci_hcd *ehci = hcd_to_ehci(hcd);
> +
> + return container_of(ehci, struct ehci_fsl, ehci);
> +}

This is bound to fail no matter what, because the ehci_hcd structure is
already embedded in the usb_hcd structure. It can't be embedded in an
ehci_fsl structure too.

(Of course, this won't matter when you switch to the approved way of
connecting private data.)

> +static void do_change_hcd(struct work_struct *work)
> +{
> + struct ehci_fsl *ehci_fsl = container_of(work, struct ehci_fsl,
> + change_hcd_work);
> + struct ehci_hcd *ehci = &ehci_fsl->ehci;
> + struct usb_hcd *hcd = ehci_to_hcd(ehci);
> +
> + void __iomem *non_ehci = hcd->regs;
> + int retval;
> +
> + if (ehci_fsl->hcd_add && !ehci_fsl->have_hcd) {
> + writel(USBMODE_CM_HOST, non_ehci + FSL_SOC_USB_USBMODE);
> + /* host, gadget and otg share same int line */
> + retval = usb_add_hcd(hcd, hcd->irq, IRQF_SHARED);
> + if (retval == 0)
> + ehci_fsl->have_hcd = 1;
> + } else if (!ehci_fsl->hcd_add && ehci_fsl->have_hcd) {
> + usb_remove_hcd(hcd);
> + ehci_fsl->have_hcd = 0;
> + }
> +}

Does this really work right? What about things like
device_wakeup_enable()? Don't you need to disable the controller when
leaving host mode and enable it when entering host mode?

Alan Stern



\
 
 \ /
  Last update: 2015-04-07 18:21    [W:0.102 / U:0.052 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site