Re: [dpdk-dev] [PATCH v5 3/3] PCI: don't use vfio ioctl call to access PIO resource

[Maxime Coquelin <maxime.coquelin@redhat.com>]

On 1/27/21 11:32 AM, Ferruh Yigit wrote:
> On 1/26/2021 10:44 AM, Maxime Coquelin wrote:
>>
>>
>> On 1/22/21 8:25 AM, 谢华伟(此时此刻） wrote:
>>>
>>> On 2021/1/21 23:38, Maxime Coquelin wrote:
>>>>> Do you mean we apply or abandon patch 3? I am both OK. The first
>>>>> priority to me is to enable MMIO bar support.
>>>> OK, so yes, I think we should abandon patch 2 and patch 3.
>>>> For patch 1, it looks valid to me, but I'll let Ferruh decide.
>>>>
>>>> For your device, if my understanding is correct, what we need to do is
>>>> to support MMIO for legacy devices. Correct?
>>> yes.
>>>> If so, the change should be in virtio_pci.c. In vtpci_init(), after
>>>> modern detection has failed, we should check the the BAR is PIO or MMIO
>>>> based on the flag. the result can be saved in struct virtio_pci_dev.
>>>>
>>>>
>>>> We would introduce new wrappers like vtpci_legacy_read,
>>>> vtpci_legacy_write that would either call rte_pci_ioport_read,
>>>> rte_pci_ioport_read in case of PIO, or rte_read32, rte_write32 in case
>>>> of MMIO.
>>>
>>> There are two choices.
>>>
>>> 1, apply patch 2.
>>>
>>>      IO/MMIO port are mapped and accessed using the same API. Kernel is
>>> doing in the same way like the following.
>>>
>>>              io_addr = pci_iomap
>>>
>>>                  get PIO directly or ioremap
>>>
>>>              iowrite16/32(val, io_addr + offset)
>>>
>>> I think applying patch 2 is a correct choice. It is a fix. Driver had
>>> better not know if bar is PIO or MMIO.  ioport in ioport_xx API means
>>> IO, not PIO.
>>>
>>> Btw, it only affects virtio PMD,  not that intrusive.
>>>
>>>   2, virtio specific change to enable MMIO support.
>>>
>>> Comparing with choice 1, i feels it is not that clean and pretty.
>>
>> OK, that makes sense. I am OK with keeping patch 2, but would like
>> Ferruh's ACK.
>>
> 
> I was waiting for clarification if this can be solved in virtio, which
> seems clarified and decided to go with this patch, I am OK to proceed
> with patch 1 & 2.
> 
> But first patch changes how PIO address get, it changes the Linux
> interface used to get the PIO.
> And as far as I can see second patch requires this new interface to be
> able to access the MEM resources.
> 
> I have a concern that this interface change may cause issues with
> various distros, kernel versions etc.. And prefer it goes through a full
> -rc1 validation cycle.

While I think the risk for patch 2 is close to zero, I understand your
concern on patch 1 (especially with the upcoming holidays in China,
which will have an impact on QE capacity).

Huawei, do you think patch 2 can be slightly modified to be applied
alone, without patch 1? If possible, we may be able to pick patch2 for
this release and postpone patch 1 to v21.05?

> Huawei, I am aware the patch is around for a while but to play safe, I
> suggest considering it for early next release, so it can be tested
> enough, instead of getting if for -rc2/3 in this release.
> 
> Thanks,
> ferruh
> 
> 
>> Could you please post v6?
>>
>> Thanks,
>> Maxime
>>
>>>>
>>>> It is not too late for this release, as the change will not be that
>>>> intrusive. But if you prepare such patch, please base it on top of my
>>>> virtio rework series; To make it easier to you, I added it to the dpdk-
>>>> next-virtio tree:
>>>> https://git.dpdk.org/next/dpdk-next-virtio/log/?h=virtio_pmd_rework_v2
>>>>
>>>> Thanks,
>>>> Maxime
>>>>
>>>
>>
>