Re: linux-next: Tree for Dec 9 (usb/cdns3)

Re: linux-next: Tree for Dec 9 (usb/cdns3)

[Randy Dunlap]

[-- Attachment #1: Type: text/plain, Size: 929 bytes --]

On 12/9/20 2:44 AM, Stephen Rothwell wrote:
> Hi all,
> 
> Changes since 20201208:
> 

(I don't know what to do about this one -- seeking help.)


on x86_64:

ld: drivers/usb/cdns3/host.o: in function `xhci_cdns3_suspend_quirk':
host.c:(.text+0x9): undefined reference to `usb_hcd_is_primary_hcd'

This reference to 'usb_hdc_is_primary_hcd' is from hcd_to_xhci(),
which is being built as a loadable module:

int xhci_cdns3_suspend_quirk(struct usb_hcd *hcd)
{
	struct xhci_hcd	*xhci = hcd_to_xhci(hcd);




CONFIG_USB_GADGET=y
CONFIG_USB_SUPPORT=y
CONFIG_USB_COMMON=y
# CONFIG_USB_CONN_GPIO is not set
CONFIG_USB_ARCH_HAS_HCD=y
CONFIG_USB=m

CONFIG_USB_CDNS_SUPPORT=y
CONFIG_USB_CDNS_HOST=y
CONFIG_USB_CDNS3=m
CONFIG_USB_CDNS3_GADGET=y
CONFIG_USB_CDNS3_HOST=y

Problem is mostly that CONFIG_USB=m and CONFIG_USB_GADGET=y.


Full randconfig file is attached.


thanks.
-- 
~Randy
Reported-by: Randy Dunlap <rdunlap@infradead.org>

[-- Attachment #2: config-r6651.xz --]
[-- Type: application/x-xz, Size: 31248 bytes --]

Re: linux-next: Tree for Dec 9 (usb/cdns3)

[Peter Chen]

On 20-12-09 16:58:16, Randy Dunlap wrote:
> On 12/9/20 2:44 AM, Stephen Rothwell wrote:
> > Hi all,
> > 
> > Changes since 20201208:
> > 
> 
> (I don't know what to do about this one -- seeking help.)

Add Pawel.

Hi Pawel,

Add old cdns3 logic, when the CONFIG_USB=m
If CONFIG_USB_CDNS3 is M, the host will be built as module
If CONFIG_USB_CDNS3 is build-in, the host will not built due to
USB=m, so USB!= USB_CDNS3 at below dependency.

config USB_CDNS3_HOST
	bool "Cadence USB3 host controller"
	depends on USB=y || USB=USB_CDNS3

So, it has no such issue.

But after adding CDNSSP support, the configuration relationship is
much complicated, both CDNS3 and CDNSSP could choose host file,
would you have a check for this issue?

Peter

> 
> 
> on x86_64:
> 
> ld: drivers/usb/cdns3/host.o: in function `xhci_cdns3_suspend_quirk':
> host.c:(.text+0x9): undefined reference to `usb_hcd_is_primary_hcd'
> 
> This reference to 'usb_hdc_is_primary_hcd' is from hcd_to_xhci(),
> which is being built as a loadable module:
> 
> int xhci_cdns3_suspend_quirk(struct usb_hcd *hcd)
> {
> 	struct xhci_hcd	*xhci = hcd_to_xhci(hcd);
> 
> 
> 
> 
> CONFIG_USB_GADGET=y
> CONFIG_USB_SUPPORT=y
> CONFIG_USB_COMMON=y
> # CONFIG_USB_CONN_GPIO is not set
> CONFIG_USB_ARCH_HAS_HCD=y
> CONFIG_USB=m
> 
> CONFIG_USB_CDNS_SUPPORT=y
> CONFIG_USB_CDNS_HOST=y
> CONFIG_USB_CDNS3=m
> CONFIG_USB_CDNS3_GADGET=y
> CONFIG_USB_CDNS3_HOST=y
> 
> Problem is mostly that CONFIG_USB=m and CONFIG_USB_GADGET=y.
> 
> 
> Full randconfig file is attached.
> 
> 
> thanks.
> -- 
> ~Randy
> Reported-by: Randy Dunlap <rdunlap@infradead.org>



-- 

Thanks,
Peter Chen

RE: linux-next: Tree for Dec 9 (usb/cdns3)

[Pawel Laszczak]

>
>On 20-12-09 16:58:16, Randy Dunlap wrote:
>> On 12/9/20 2:44 AM, Stephen Rothwell wrote:
>> > Hi all,
>> >
>> > Changes since 20201208:
>> >
>>
>> (I don't know what to do about this one -- seeking help.)
>
>Add Pawel.
>
>Hi Pawel,
>
>Add old cdns3 logic, when the CONFIG_USB=m
>If CONFIG_USB_CDNS3 is M, the host will be built as module
>If CONFIG_USB_CDNS3 is build-in, the host will not built due to
>USB=m, so USB!= USB_CDNS3 at below dependency.
>
>config USB_CDNS3_HOST
>	bool "Cadence USB3 host controller"
>	depends on USB=y || USB=USB_CDNS3
>
>So, it has no such issue.
>
>But after adding CDNSSP support, the configuration relationship is
>much complicated, both CDNS3 and CDNSSP could choose host file,
>would you have a check for this issue?

I can recreate this issue. I will try to resolve it. 

>
>Peter
>
>>
>>
>> on x86_64:
>>
>> ld: drivers/usb/cdns3/host.o: in function `xhci_cdns3_suspend_quirk':
>> host.c:(.text+0x9): undefined reference to `usb_hcd_is_primary_hcd'
>>
>> This reference to 'usb_hdc_is_primary_hcd' is from hcd_to_xhci(),
>> which is being built as a loadable module:
>>
>> int xhci_cdns3_suspend_quirk(struct usb_hcd *hcd)
>> {
>> 	struct xhci_hcd	*xhci = hcd_to_xhci(hcd);
>>
>>
>>
>>
>> CONFIG_USB_GADGET=y
>> CONFIG_USB_SUPPORT=y
>> CONFIG_USB_COMMON=y
>> # CONFIG_USB_CONN_GPIO is not set
>> CONFIG_USB_ARCH_HAS_HCD=y
>> CONFIG_USB=m
>>
>> CONFIG_USB_CDNS_SUPPORT=y
>> CONFIG_USB_CDNS_HOST=y
>> CONFIG_USB_CDNS3=m
>> CONFIG_USB_CDNS3_GADGET=y
>> CONFIG_USB_CDNS3_HOST=y
>>
>> Problem is mostly that CONFIG_USB=m and CONFIG_USB_GADGET=y.
>>
>>
>> Full randconfig file is attached.
>>
>>
>> thanks.
>> --
>> ~Randy
>> Reported-by: Randy Dunlap <rdunlap@infradead.org>
>
>
>
--

Thanks
Pawel Laszczak

RE: linux-next: Tree for Dec 9 (usb/cdns3)

[Pawel Laszczak]

Peter,

>
>>
>>On 20-12-09 16:58:16, Randy Dunlap wrote:
>>> On 12/9/20 2:44 AM, Stephen Rothwell wrote:
>>> > Hi all,
>>> >
>>> > Changes since 20201208:
>>> >
>>>
>>> (I don't know what to do about this one -- seeking help.)
>>
>>Add Pawel.
>>
>>Hi Pawel,
>>
>>Add old cdns3 logic, when the CONFIG_USB=m
>>If CONFIG_USB_CDNS3 is M, the host will be built as module
>>If CONFIG_USB_CDNS3 is build-in, the host will not built due to
>>USB=m, so USB!= USB_CDNS3 at below dependency.
>>
>>config USB_CDNS3_HOST
>>	bool "Cadence USB3 host controller"
>>	depends on USB=y || USB=USB_CDNS3
>>
>>So, it has no such issue.
>>
>>But after adding CDNSSP support, the configuration relationship is
>>much complicated, both CDNS3 and CDNSSP could choose host file,
>>would you have a check for this issue?
>
>I can recreate this issue. I will try to resolve it.

config USB_CDNS3_HOST
        bool "Cadence USB3 host controller"
-       depends on USB=y || USB=USB_CDNS3
+       depends on USB=USB_CDNS3
        select USB_CDNS_HOST
        help
          Say Y here to enable host controller functionality of the
@@ -110,7 +110,7 @@ config USB_CDNSP_GADGET

 config USB_CDNSP_HOST
        bool "Cadence CDNSP host controller"
-       depends on USB=y || USB=USB_CDNSP_PCI
+       depends on USB=USB_CDNSP_PCI
        select USB_CDNS_HOST
        help
          Say Y here to enable host controller functionality of the

Peter, what about such change. It fix this issue but I need to check
It with some other kernel configurations. 
At this moment it's the only solution which I've found but
it's introduces some limitation in CDNS3/CDNSP configuration.

>
>>
>>Peter
>>
>>>
>>>
>>> on x86_64:
>>>
>>> ld: drivers/usb/cdns3/host.o: in function `xhci_cdns3_suspend_quirk':
>>> host.c:(.text+0x9): undefined reference to `usb_hcd_is_primary_hcd'
>>>
>>> This reference to 'usb_hdc_is_primary_hcd' is from hcd_to_xhci(),
>>> which is being built as a loadable module:
>>>
>>> int xhci_cdns3_suspend_quirk(struct usb_hcd *hcd)
>>> {
>>> 	struct xhci_hcd	*xhci = hcd_to_xhci(hcd);
>>>
>>>
>>>
>>>
>>> CONFIG_USB_GADGET=y
>>> CONFIG_USB_SUPPORT=y
>>> CONFIG_USB_COMMON=y
>>> # CONFIG_USB_CONN_GPIO is not set
>>> CONFIG_USB_ARCH_HAS_HCD=y
>>> CONFIG_USB=m
>>>
>>> CONFIG_USB_CDNS_SUPPORT=y
>>> CONFIG_USB_CDNS_HOST=y
>>> CONFIG_USB_CDNS3=m
>>> CONFIG_USB_CDNS3_GADGET=y
>>> CONFIG_USB_CDNS3_HOST=y
>>>
>>> Problem is mostly that CONFIG_USB=m and CONFIG_USB_GADGET=y.
>>>
>>>
>>> Full randconfig file is attached.
>>>
>>>
>>> thanks.
>>> --
>>> ~Randy
>>> Reported-by: Randy Dunlap <rdunlap@infradead.org>
>>
>>
>>
--

Thanks
Pawel Laszczak

RE: linux-next: Tree for Dec 9 (usb/cdns3)

[Pawel Laszczak]

Peter,

>
>>
>>>
>>>On 20-12-09 16:58:16, Randy Dunlap wrote:
>>>> On 12/9/20 2:44 AM, Stephen Rothwell wrote:
>>>> > Hi all,
>>>> >
>>>> > Changes since 20201208:
>>>> >
>>>>
>>>> (I don't know what to do about this one -- seeking help.)
>>>
>>>Add Pawel.
>>>
>>>Hi Pawel,
>>>
>>>Add old cdns3 logic, when the CONFIG_USB=m
>>>If CONFIG_USB_CDNS3 is M, the host will be built as module
>>>If CONFIG_USB_CDNS3 is build-in, the host will not built due to
>>>USB=m, so USB!= USB_CDNS3 at below dependency.
>>>
>>>config USB_CDNS3_HOST
>>>	bool "Cadence USB3 host controller"
>>>	depends on USB=y || USB=USB_CDNS3
>>>
>>>So, it has no such issue.
>>>
>>>But after adding CDNSSP support, the configuration relationship is
>>>much complicated, both CDNS3 and CDNSSP could choose host file,
>>>would you have a check for this issue?
>>
>>I can recreate this issue. I will try to resolve it.
>
>config USB_CDNS3_HOST
>        bool "Cadence USB3 host controller"
>-       depends on USB=y || USB=USB_CDNS3
>+       depends on USB=USB_CDNS3
>        select USB_CDNS_HOST
>        help
>          Say Y here to enable host controller functionality of the
>@@ -110,7 +110,7 @@ config USB_CDNSP_GADGET
>
> config USB_CDNSP_HOST
>        bool "Cadence CDNSP host controller"
>-       depends on USB=y || USB=USB_CDNSP_PCI
>+       depends on USB=USB_CDNSP_PCI
>        select USB_CDNS_HOST
>        help
>          Say Y here to enable host controller functionality of the
>
>Peter, what about such change. It fix this issue but I need to check
>It with some other kernel configurations.
>At this moment it's the only solution which I've found but
>it's introduces some limitation in CDNS3/CDNSP configuration.

It doesn't work correct.
I posted the patch with other solution. 

>
>>
>>>
>>>Peter
>>>
>>>>
>>>>
>>>> on x86_64:
>>>>
>>>> ld: drivers/usb/cdns3/host.o: in function `xhci_cdns3_suspend_quirk':
>>>> host.c:(.text+0x9): undefined reference to `usb_hcd_is_primary_hcd'
>>>>
>>>> This reference to 'usb_hdc_is_primary_hcd' is from hcd_to_xhci(),
>>>> which is being built as a loadable module:
>>>>
>>>> int xhci_cdns3_suspend_quirk(struct usb_hcd *hcd)
>>>> {
>>>> 	struct xhci_hcd	*xhci = hcd_to_xhci(hcd);
>>>>
>>>>
>>>>
>>>>
>>>> CONFIG_USB_GADGET=y
>>>> CONFIG_USB_SUPPORT=y
>>>> CONFIG_USB_COMMON=y
>>>> # CONFIG_USB_CONN_GPIO is not set
>>>> CONFIG_USB_ARCH_HAS_HCD=y
>>>> CONFIG_USB=m
>>>>
>>>> CONFIG_USB_CDNS_SUPPORT=y
>>>> CONFIG_USB_CDNS_HOST=y
>>>> CONFIG_USB_CDNS3=m
>>>> CONFIG_USB_CDNS3_GADGET=y
>>>> CONFIG_USB_CDNS3_HOST=y
>>>>
>>>> Problem is mostly that CONFIG_USB=m and CONFIG_USB_GADGET=y.
>>>>
>>>>
>>>> Full randconfig file is attached.
>>>>
>>>>
>>>> thanks.
>>>> --
>>>> ~Randy
>>>> Reported-by: Randy Dunlap <rdunlap@infradead.org>
>>>
>>>
>>>
--

Thanks
Pawel Laszczak