Re: [PATCH net-next] net: kbuild: Don't default net vendor configs to y

[Saeed Mahameed <saeedm@nvidia.com>]

On 02 Feb 06:30, Christophe Leroy wrote:
>
>
>Le 02/02/2022 à 06:16, Saeed Mahameed a écrit :
>> On 01 Feb 20:58, Jakub Kicinski wrote:
>>> On Tue, 1 Feb 2022 20:46:03 -0800 Saeed Mahameed wrote:
>>>> I am getting mixed messages here, on one hand we know that this patch
>>>> might break some old or def configs, but on the other hand people claim
>>>> that they have to manually fixup their own configs every time
>>>> "something in configs" changes and they are fine with that.
>>>>
>>>> Obviously I belong to the 2nd camp, hence this patch..
>>>>
>>>> I can sum it up with "it's fine to controllably break *some* .configs
>>>> for
>>>> the greater good" .. that's my .2cent.
>>>
>>> I think we agree that we don't care about oldconfigs IOW someone's
>>> random config.
>>>
>>> But we do care about defconfigs in the tree, if those indeed include
>>> ethernet drivers which would get masked out by vendor=n - they need
>>> fixin':
>>>
>>> $ find arch/ | grep defconfig
>>> arch/x86/configs/i386_defconfig
>>> arch/x86/configs/x86_64_defconfig
>>> arch/ia64/configs/generic_defconfig
>>> arch/ia64/configs/gensparse_defconfig
>>> ...
>>>
>>> First one from the top:
>>>
>>> $ make O=build_tmp/ i386_defconfig
>>> $ $EDITOR drivers/net/ethernet/intel/Kconfig
>>> $ git diff
>>> diff --git a/drivers/net/ethernet/intel/Kconfig
>>> b/drivers/net/ethernet/intel/Kconfig
>>> index 3facb55b7161..b9fdf2a835b0 100644
>>> --- a/drivers/net/ethernet/intel/Kconfig
>>> +++ b/drivers/net/ethernet/intel/Kconfig
>>> @@ -5,7 +5,6 @@
>>>
>>> config NET_VENDOR_INTEL
>>>        bool "Intel devices"
>>> -       default y
>>>        help
>>>          If you have a network (Ethernet) card belonging to this
>>> class, say Y.
>>>
>>> $ make O=build_tmp/ i386_defconfig
>>> $ diff -urpb build_tmp/.config.old build_tmp/.config
>>> --- build_tmp/.config.old    2022-02-01 20:55:37.087373905 -0800
>>> +++ build_tmp/.config    2022-02-01 20:56:32.126044628 -0800
>>> @@ -1784,22 +1784,7 @@ CONFIG_NET_VENDOR_GOOGLE=y
>>> # CONFIG_GVE is not set
>>> CONFIG_NET_VENDOR_HUAWEI=y
>>> # CONFIG_HINIC is not set
>>> -CONFIG_NET_VENDOR_I825XX=y
>>> -CONFIG_NET_VENDOR_INTEL=y
>>> -CONFIG_E100=y
>>> -CONFIG_E1000=y
>>> -CONFIG_E1000E=y
>>> -CONFIG_E1000E_HWTS=y
>>> -# CONFIG_IGB is not set
>>> -# CONFIG_IGBVF is not set
>>> -# CONFIG_IXGB is not set
>>> -# CONFIG_IXGBE is not set
>>> -# CONFIG_IXGBEVF is not set
>>> -# CONFIG_I40E is not set
>>> -# CONFIG_I40EVF is not set
>>> -# CONFIG_ICE is not set
>>> -# CONFIG_FM10K is not set
>>> -# CONFIG_IGC is not set
>>> +# CONFIG_NET_VENDOR_INTEL is not set
>>
>> We can introduce CONFIG_NET_LEGACY_VENDOR that selects all current vendors.
>> it will be off by default but will be added where needed in the defconfigs
>
>Why such a hack ?
>

Reduced chance of error.

>I think we can fix all defconfig with some scripting, all you have to do
>it to add the relevant CONFIG_NET_VENDOR_SOMEONE=y wherever you find one
>of its boards in the defconfig.
>

Such a script could easily mess up!
I can't think of a clever easily verifiable way to map boards to their VENDORS.
Add to that dispersing the VENDORS configs accurately.

I might be just tired though, i will give it a shot in the morning :) 

>Christophe