Re: 答复: 答复: [PATCH v6 01/17] net/r8169: add PMD driver skeleton

[Ferruh Yigit]

On 11/11/2024 9:24 AM, 王颢 wrote:
> Dear 
> 
> Ahh, I see, matching is simple for 8125, device id and product number is same :), is it same for all supported devices?
> 
> ===> Yes, up until now, the Linux inbox driver has been doing it this way too.
> 

Got it, so OK to keep the code as it is, thanks for the clarification.

> I am not sure if there is a misunderstanding but change requests are mostly simple ones, I expect they can be fixed in one or two days, and I am hopping to merge the driver in -rc2 or -rc3 at worst.
> ===> I will do it as soon as possible. I have had a fever recently, so it might be a bit slow.
>

Great, thanks. (And I hope you get better soon ;)

> 
> Best Regards,
> Howard Wang
> 
> -----邮件原件-----
> 发件人: Ferruh Yigit <ferruh.yigit at amd.com> 
> 发送时间: 2024年11月11日 17:09
> 收件人: 王颢 <howard_wang at realsil.com.cn>; dev at dpdk.org
> 抄送: pro_nic_dpdk at realtek.com
> 主题: Re: 答复: [PATCH v6 01/17] net/r8169: add PMD driver skeleton
> 
> 
> External mail.
> 
> 
> 
> On 11/11/2024 3:46 AM, 王颢 wrote:
>> Dear Ferruh,
>>
>> I apologize for the issues that have arisen during the process of splitting a complete driver into several small patches.
>>
>> On 11/8/2024 12:11 PM, Howard Wang wrote:
>>> Meson build infrastructure, r8169_ethdev minimal skeleton, header 
>>> with Realtek NIC device and vendor IDs.
>>>
>>> Signed-off-by: Howard Wang <howard_wang at realsil.com.cn>
>>
>> <...>
>>
>> ===> I referred to Atlantic’s approach for this, and as a matter of fact, his first commit message was done in this manner.
>> I will think again about how it should be modified.
>>
> 
> No problem with the commit splits and orders, thanks for the update there.
> 
> Only comment in this patch is 'RTL_DEV_PRIVATE' macro is not used in the patch and it can be added when you use the macro, so this is a simple issue.
> 
> Also I have a comment that build fails after some patchs, this also does not require changing how patch split done, just fix build after each patch.
> 
> I am not sure if there is a misunderstanding but change requests are mostly simple ones, I expect they can be fixed in one or two days, and I am hopping to merge the driver in -rc2 or -rc3 at worst.
> 
>> net/atlantic: add PMD driver skeleton
>> Makefile/meson build infrastructure, atl_ethdev minimal skeleton, 
>> header with aquantia aQtion NIC device and vendor IDs.
>>
>> Signed-off-by: Igor Russkikh <igor.russkikh at aquantia.com>
>> Signed-off-by: Pavel Belous <pavel.belous at aquantia.com>
>>
>>
>>
>> What do you think to add macros for the PCI device IDs, naming them makes it easier to know which devices are supported.
>>
>> ===> Regarding the device ID, for 0x8125, the supported device is indeed called 8125. So, should we name a macro like DEVICE_ID_RTL8125 to replace it?
>>
> 
> 
> Ahh, I see, matching is simple for 8125, device id and product number is same :), is it same for all supported devices?
>         { RTE_PCI_DEVICE(PCI_VENDOR_ID_REALTEK, 0x8125) },
>         { RTE_PCI_DEVICE(PCI_VENDOR_ID_REALTEK, 0x8162) },
>         { RTE_PCI_DEVICE(PCI_VENDOR_ID_REALTEK, 0x8126) },
>         { RTE_PCI_DEVICE(PCI_VENDOR_ID_REALTEK, 0x5000) },
> 
>>
>> Best regards,
>> Howard Wang
>>
>> -----邮件原件-----
>> 发件人: Ferruh Yigit <ferruh.yigit at amd.com>
>> 发送时间: 2024年11月11日 8:15
>> 收件人: 王颢 <howard_wang at realsil.com.cn>; dev at dpdk.org
>> 抄送: pro_nic_dpdk at realtek.com
>> 主题: Re: [PATCH v6 01/17] net/r8169: add PMD driver skeleton
>>
>>
>> External mail.
>>
>>
>>
>> On 11/8/2024 12:11 PM, Howard Wang wrote:
>>> Meson build infrastructure, r8169_ethdev minimal skeleton, header 
>>> with Realtek NIC device and vendor IDs.
>>>
>>> Signed-off-by: Howard Wang <howard_wang at realsil.com.cn>
>>
>> <...>
>>
>>> +/*
>>> + * The set of PCI devices this driver supports  */ static const 
>>> +struct rte_pci_id pci_id_r8169_map[] = {
>>> +     { RTE_PCI_DEVICE(PCI_VENDOR_ID_REALTEK, 0x8125) },
>>> +     { RTE_PCI_DEVICE(PCI_VENDOR_ID_REALTEK, 0x8162) },
>>> +     { RTE_PCI_DEVICE(PCI_VENDOR_ID_REALTEK, 0x8126) },
>>> +     { RTE_PCI_DEVICE(PCI_VENDOR_ID_REALTEK, 0x5000) },
>>>
>>
>> What do you think to add macros for the PCI device IDs, naming them makes it easier to know which devices are supported.
>>
>> <...>
>>
>>> +
>>> +#define RTL_DEV_PRIVATE(eth_dev) \
>>> +     ((struct rtl_adapter *)((eth_dev)->data->dev_private))
>>> +
>>>
>>
>> This macro is only used a few patches later, what do you think to add this macro when used?
>>
>>
>