[PATCH V2] usb: dwc3: Initialize DMA ops/mask for xhci-hcd

[Adam Wallis]

On 8/28/2017 1:57 PM, Vivek Gautam wrote:
> Hi Adam,
> 
> 
> On Mon, Aug 28, 2017 at 10:05 PM, Adam Wallis <awallis at codeaurora.org> wrote:
>> The dma ops from the parent DWC device are not getting passed to the
>> child xhci-hcd device. This patch makes use of
>> platform_device_register_full to set the DMA ops. For the DT/OF case,
>> dma_ops were still null after the the device register, so
>> of_dma_configure is called in only the OF case.
>>
>> Signed-off-by: Adam Wallis <awallis at codeaurora.org>
>> ---
>>  drivers/usb/dwc3/host.c | 39 +++++++++++++++++----------------------
>>  1 file changed, 17 insertions(+), 22 deletions(-)
>>
>> diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c
>> index 76f0b0d..c289da1 100644
>> --- a/drivers/usb/dwc3/host.c
>> +++ b/drivers/usb/dwc3/host.c
>> @@ -16,6 +16,7 @@
>>   */
>>
>>  #include <linux/platform_device.h>
>> +#include <linux/of_platform.h>
>>
>>  #include "core.h"
>>
>> @@ -56,6 +57,7 @@ int dwc3_host_init(struct dwc3 *dwc)
>>  {
>>         struct property_entry   props[3];
>>         struct platform_device  *xhci;
>> +       struct platform_device_info dwc_plat_info = {};
>>         int                     ret, irq;
>>         struct resource         *res;
>>         struct platform_device  *dwc3_pdev = to_platform_device(dwc->dev);
>> @@ -79,22 +81,22 @@ int dwc3_host_init(struct dwc3 *dwc)
>>         dwc->xhci_resources[1].flags = res->flags;
>>         dwc->xhci_resources[1].name = res->name;
>>
>> -       xhci = platform_device_alloc("xhci-hcd", PLATFORM_DEVID_AUTO);
>> -       if (!xhci) {
>> -               dev_err(dwc->dev, "couldn't allocate xHCI device\n");
>> -               return -ENOMEM;
>> -       }
>> +       dwc_plat_info.name = "xhci-hcd";
>> +       dwc_plat_info.parent = dwc->sysdev;
>> +       dwc_plat_info.res = dwc->xhci_resources;
>> +       dwc_plat_info.num_res = DWC3_XHCI_RESOURCES_NUM;
>> +       dwc_plat_info.fwnode = dwc->sysdev->fwnode;
>> +       dwc_plat_info.dma_mask = *dwc->sysdev->dma_mask;
>>
>> -       xhci->dev.parent        = dwc->dev;
>> +       xhci = platform_device_register_full(&dwc_plat_info);
>> +       if (IS_ERR(xhci)) {
>> +               dev_err(dwc->dev, "failed to register xHCI device\n");
>> +               return PTR_ERR(xhci);
>> +       }
>>
>>         dwc->xhci = xhci;
>> -
>> -       ret = platform_device_add_resources(xhci, dwc->xhci_resources,
>> -                                               DWC3_XHCI_RESOURCES_NUM);
>> -       if (ret) {
>> -               dev_err(dwc->dev, "couldn't add resources to xHCI device\n");
>> -               goto err1;
>> -       }
>> +       if (dwc->sysdev->of_node)
>> +               of_dma_configure(&xhci->dev, dwc->sysdev->of_node);
> 
> Isn't this code doing the same thing what the xhci-plat also does right now
> since dwc->sysdev is either dwc->dev or dwc->dev->parent.
> 
> From xhci_plat_probe():
>         sysdev = &pdev->dev;
>         if (sysdev->parent && !sysdev->of_node && sysdev->parent->of_node)
>                 sysdev = sysdev->parent;
> 
> after which it tries to set the dma masks.
> 
> <snip>
> 
Yep, you are right - I was testing against 4.11 (which was broken). When I pull
in a8c06e407ef969461b7f51ec72839fe382dd3c29 and
4c39d4b949d36faffbc726525b469886cf311d1c, I am good to go. This patch can be
disregarded. Thanks for the help
> 
> regards
> Vivek
> 


-- 
Adam Wallis
Qualcomm Datacenter Technologies as an affiliate of Qualcomm Technologies, Inc.
Qualcomm Technologies, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project.