Re: [PATCH v4 1/2] xen-pciback: prepare for the split for stub and PV

[Oleksandr Andrushchenko <Oleksandr_Andrushchenko@xxxxxxxx>]

On 28.09.21 10:20, Juergen Gross wrote:
> On 28.09.21 09:17, Oleksandr Andrushchenko wrote:
>>
>> On 28.09.21 09:59, Juergen Gross wrote:
>>> On 28.09.21 08:56, Oleksandr Andrushchenko wrote:
>>>>
>>>> On 28.09.21 09:42, Jan Beulich wrote:
>>>>> On 28.09.2021 06:18, Stefano Stabellini wrote:
>>>>>> On Mon, 27 Sep 2021, Juergen Gross wrote:
>>>>>>> On 27.09.21 09:35, Oleksandr Andrushchenko wrote:
>>>>>>>> On 27.09.21 10:26, Jan Beulich wrote:
>>>>>>>>> On 27.09.2021 08:58, Oleksandr Andrushchenko wrote:
>>>>>>>>>> From: Oleksandr Andrushchenko <oleksandr_andrushchenko@xxxxxxxx>
>>>>>>>>>>
>>>>>>>>>> Currently PCI backend implements multiple functionalities at a time.
>>>>>>>>>> To name a few:
>>>>>>>>>> 1. It is used as a database for assignable PCI devices, e.g. xl
>>>>>>>>>>         pci-assignable-{add|remove|list} manipulates that list. So,
>>>>>>>>>> whenever
>>>>>>>>>>         the toolstack needs to know which PCI devices can be passed 
>>>>>>>>>> through
>>>>>>>>>>         it reads that from the relevant sysfs entries of the pciback.
>>>>>>>>>> 2. It is used to hold the unbound PCI devices list, e.g. when passing
>>>>>>>>>>         through a PCI device it needs to be unbound from the relevant
>>>>>>>>>> device
>>>>>>>>>>         driver and bound to pciback (strictly speaking it is not 
>>>>>>>>>> required
>>>>>>>>>>         that the device is bound to pciback, but pciback is again 
>>>>>>>>>> used as a
>>>>>>>>>>         database of the passed through PCI devices, so we can 
>>>>>>>>>> re-bind the
>>>>>>>>>>         devices back to their original drivers when guest domain 
>>>>>>>>>> shuts
>>>>>>>>>> down)
>>>>>>>>>> 3. Device reset for the devices being passed through
>>>>>>>>>> 4. Para-virtualised use-cases support
>>>>>>>>>>
>>>>>>>>>> The para-virtualised part of the driver is not always needed as some
>>>>>>>>>> architectures, e.g. Arm or x86 PVH Dom0, are not using 
>>>>>>>>>> backend-frontend
>>>>>>>>>> model for PCI device passthrough. For such use-cases make the very
>>>>>>>>>> first step in splitting the xen-pciback driver into two parts: Xen
>>>>>>>>>> PCI stub and PCI PV backend drivers.
>>>>>>>>>>
>>>>>>>>>> Signed-off-by: Oleksandr Andrushchenko
>>>>>>>>>> <oleksandr_andrushchenko@xxxxxxxx>
>>>>>>>>>>
>>>>>>>>>> ---
>>>>>>>>>> Changes since v3:
>>>>>>>>>> - Move CONFIG_XEN_PCIDEV_STUB to the second patch
>>>>>>>>> I'm afraid this wasn't fully done:
>>>>>>>>>
>>>>>>>>>> --- a/drivers/xen/xen-pciback/Makefile
>>>>>>>>>> +++ b/drivers/xen/xen-pciback/Makefile
>>>>>>>>>> @@ -1,5 +1,6 @@
>>>>>>>>>>       # SPDX-License-Identifier: GPL-2.0
>>>>>>>>>>       obj-$(CONFIG_XEN_PCIDEV_BACKEND) += xen-pciback.o
>>>>>>>>>> +obj-$(CONFIG_XEN_PCIDEV_STUB) += xen-pciback.o
>>>>>>>>> While benign when added here, this addition still doesn't seem to
>>>>>>>>> belong here.
>>>>>>>> My bad. So, it seems without CONFIG_XEN_PCIDEV_STUB the change seems
>>>>>>>>
>>>>>>>> to be non-functional. With CONFIG_XEN_PCIDEV_STUB we fail to build on 
>>>>>>>> 32-bit
>>>>>>>>
>>>>>>>> architectures...
>>>>>>>>
>>>>>>>> What would be the preference here? Stefano suggested that we still 
>>>>>>>> define
>>>>>>>>
>>>>>>>> CONFIG_XEN_PCIDEV_STUB, but in disabled state, e.g. we add tristate to 
>>>>>>>> it
>>>>>>>>
>>>>>>>> in the second patch
>>>>>>>>
>>>>>>>> Another option is just to squash the two patches.
>>>>>>> Squashing would be fine for me.
>>>>>>     It is fine for me to squash the two patches.
>>>>>>
>>>>>> But in any case, wouldn't it be better to modify that specific change to:
>>>>>>
>>>>>> diff --git a/drivers/xen/xen-pciback/Makefile 
>>>>>> b/drivers/xen/xen-pciback/Makefile
>>>>>> index e2cb376444a6..e23c758b85ae 100644
>>>>>> --- a/drivers/xen/xen-pciback/Makefile
>>>>>> +++ b/drivers/xen/xen-pciback/Makefile
>>>>>> @@ -1,6 +1,5 @@
>>>>>>     # SPDX-License-Identifier: GPL-2.0
>>>>>> -obj-$(CONFIG_XEN_PCIDEV_BACKEND) += xen-pciback.o
>>>>>> -obj-$(CONFIG_XEN_PCIDEV_STUB) += xen-pciback.o
>>>>>> +obj-$(CONFIG_XEN_PCI_STUB) += xen-pciback.o
>>>>> But that wouldn't allow the driver to be a module anymore, would it?
>>>>
>>>> Exactly. I forgot that when playing with module/built-in I was not able
>>>>
>>>> to control that anymore because CONFIG_XEN_PCI_STUB will always be
>>>>
>>>> in "y" state, thus even if you have CONFIG_XEN_PCIDEV_BACKEND=m
>>>>
>>>> you won't be able to build it as module. So, I will probably put a comment
>>>>
>>>> about that in the Makefile explaining the need for
>>>>
>>>> obj-$(CONFIG_XEN_PCIDEV_BACKEND) += xen-pciback.o
>>>> obj-$(CONFIG_XEN_PCIDEV_STUB) += xen-pciback.o
>>>
>>> In case the real split between both parts of xen-pciback is done this
>>> will be needed anyway.
>>
>> Yes, it will
>>
>> So, I'll put a comment in the Makefile:
>>
>> # N.B. This cannot be expressed with a single line using CONFIG_XEN_PCI_STUB
>>
>> # as it always remains in "y" state, thus preventing the driver to be built 
>> as
>>
>> # a module.
>>
>> obj-$(CONFIG_XEN_PCIDEV_BACKEND) += xen-pciback.o
>> obj-$(CONFIG_XEN_PCIDEV_STUB) += xen-pciback.o
>>
>> Will this be ok or needs some re-wording?
>
> I'd add that CONFIG_XEN_PCIDEV_BACKEND and CONFIG_XEN_PCIDEV_STUB are
> mutually exclusive.
# N.B. The below cannot be expressed with a single line using
# CONFIG_XEN_PCI_STUB as it always remains in "y" state,
# thus preventing the driver to be built as a module.
# Please note, that CONFIG_XEN_PCIDEV_BACKEND and
# CONFIG_XEN_PCIDEV_STUB are mutually exclusive.
obj-$(CONFIG_XEN_PCIDEV_BACKEND) += xen-pciback.o
obj-$(CONFIG_XEN_PCIDEV_STUB) += xen-pciback.o

>
>
> Juergen

Thank you,

Oleksandr