[Question] Does CONFIG_XFRM_OFFLOAD depends any other configs?

[Question] Does CONFIG_XFRM_OFFLOAD depends any other configs?

[Hangbin Liu]

Hi Herbert, Jakub,

Yesterday I tried to build the kernel with CONFIG_XFRM_OFFLOAD=y via `vng`[1],
but the result .config actually doesn't contain CONFIG_XFRM_OFFLOAD=y. I saw
XFRM_OFFLOAD in net/xfrm/Kconfig doesn't has any dependences. Do you know if
I missed something?

Here are my steps:

# cat tools/testing/selftests/drivers/net/bonding/config
CONFIG_BONDING=y
CONFIG_BRIDGE=y
CONFIG_DUMMY=y
CONFIG_IPV6=y
CONFIG_MACVLAN=y
CONFIG_NET_ACT_GACT=y
CONFIG_NET_CLS_FLOWER=y
CONFIG_NET_SCH_INGRESS=y
CONFIG_NLMON=y
CONFIG_VETH=y
CONFIG_XFRM_OFFLOAD=y
# vng --build --config tools/testing/selftests/drivers/net/bonding/config

[1] https://github.com/linux-netdev/nipa/wiki/How-to-run-netdev-selftests-CI-style

Thanks
Hangbin

Re: [Question] Does CONFIG_XFRM_OFFLOAD depends any other configs?

[Jakub Kicinski]

On Tue, 20 Aug 2024 07:39:08 +0800 Hangbin Liu wrote:
> Yesterday I tried to build the kernel with CONFIG_XFRM_OFFLOAD=y via `vng`[1],
> but the result .config actually doesn't contain CONFIG_XFRM_OFFLOAD=y. I saw
> XFRM_OFFLOAD in net/xfrm/Kconfig doesn't has any dependences. Do you know if
> I missed something?

It's a hidden config option, not directly controlled by the user.
You should enable INET_ESP_OFFLOAD and INET6_ESP_OFFLOAD instead
(which "select" it)

Re: [Question] Does CONFIG_XFRM_OFFLOAD depends any other configs?

[Hangbin Liu]

Hi Jakub,
On Mon, Aug 19, 2024 at 05:22:32PM -0700, Jakub Kicinski wrote:
> On Tue, 20 Aug 2024 07:39:08 +0800 Hangbin Liu wrote:
> > Yesterday I tried to build the kernel with CONFIG_XFRM_OFFLOAD=y via `vng`[1],
> > but the result .config actually doesn't contain CONFIG_XFRM_OFFLOAD=y. I saw
> > XFRM_OFFLOAD in net/xfrm/Kconfig doesn't has any dependences. Do you know if
> > I missed something?
> 
> It's a hidden config option, not directly controlled by the user.
> You should enable INET_ESP_OFFLOAD and INET6_ESP_OFFLOAD instead
> (which "select" it)

Thanks for your reply. How to know if an option is hide other than review all
`make menuconfig` result?

Should we add a "depends on" for XFRM_OFFLOAD?

Regards
Hangbin

Re: [Question] Does CONFIG_XFRM_OFFLOAD depends any other configs?

[Jakub Kicinski]

On Tue, 20 Aug 2024 08:58:51 +0800 Hangbin Liu wrote:
> > It's a hidden config option, not directly controlled by the user.
> > You should enable INET_ESP_OFFLOAD and INET6_ESP_OFFLOAD instead
> > (which "select" it)  
> 
> Thanks for your reply. How to know if an option is hide other than review all
> `make menuconfig` result?

If it has no description after bool or tristate -- it will be hidden

> Should we add a "depends on" for XFRM_OFFLOAD?

You need it in bonding? You should use select (but not that select
doesn't resolve dependencies, it only enables that single option).

Re: [Question] Does CONFIG_XFRM_OFFLOAD depends any other configs?

[Hangbin Liu]

On Mon, Aug 19, 2024 at 06:05:25PM -0700, Jakub Kicinski wrote:
> On Tue, 20 Aug 2024 08:58:51 +0800 Hangbin Liu wrote:
> > > It's a hidden config option, not directly controlled by the user.
> > > You should enable INET_ESP_OFFLOAD and INET6_ESP_OFFLOAD instead
> > > (which "select" it)  
> > 
> > Thanks for your reply. How to know if an option is hide other than review all
> > `make menuconfig` result?
> 
> If it has no description after bool or tristate -- it will be hidden

Appreciate, another tips learned.

> 
> > Should we add a "depends on" for XFRM_OFFLOAD?
> 
> You need it in bonding? You should use select (but not that select

Yes, I may write some xfrm offload tests for bonding.

> doesn't resolve dependencies, it only enables that single option).

I didn't get which "select" you mean here. Since INET_ESP_OFFLOAD will select
XFRM_OFFLOAD, Isn't adding

CONFIG_INET_ESP=y
CONFIG_INET_ESP_OFFLOAD=y

in tools/testing/selftests/drivers/net/bonding/config enough?

Thanks
Hangbin

Re: [Question] Does CONFIG_XFRM_OFFLOAD depends any other configs?

[Jakub Kicinski]

On Tue, 20 Aug 2024 09:26:49 +0800 Hangbin Liu wrote:
> > doesn't resolve dependencies, it only enables that single option).  
> 
> I didn't get which "select" you mean here. Since INET_ESP_OFFLOAD will select
> XFRM_OFFLOAD, Isn't adding
> 
> CONFIG_INET_ESP=y
> CONFIG_INET_ESP_OFFLOAD=y
> 
> in tools/testing/selftests/drivers/net/bonding/config enough?

Sorry, I misunderstood. Yes, for the test that should be enough.