[PATCH bpf-next] mptcp: Add struct mptcp_sock definition when CONFIG_MPTCP is disabled

[PATCH bpf-next] mptcp: Add struct mptcp_sock definition when CONFIG_MPTCP is disabled

[Jiri Olsa]

The btf_sock_ids array needs struct mptcp_sock BTF ID for
the bpf_skc_to_mptcp_sock helper.

When CONFIG_MPTCP is disabled, the 'struct mptcp_sock' is not
defined and resolve_btfids will complain with:

  BTFIDS  vmlinux
WARN: resolve_btfids: unresolved symbol mptcp_sock

Adding empty difinition for struct mptcp_sock when CONFIG_MPTCP
is disabled.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 include/net/mptcp.h | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/include/net/mptcp.h b/include/net/mptcp.h
index ac9cf7271d46..25741a52c666 100644
--- a/include/net/mptcp.h
+++ b/include/net/mptcp.h
@@ -59,6 +59,10 @@ struct mptcp_addr_info {
 	};
 };
 
+#if !IS_ENABLED(CONFIG_MPTCP)
+struct mptcp_sock { };
+#endif
+
 struct mptcp_out_options {
 #if IS_ENABLED(CONFIG_MPTCP)
 	u16 suboptions;
-- 
2.35.3

Re: mptcp: Add struct mptcp_sock definition when CONFIG_MPTCP is disabled: Tests Results

[MPTCP CI]

Hi Jiri,

Thank you for your modifications, that's great!

Our CI did some validations and here is its report:

- KVM Validation: normal:
  - Unstable: 2 failed test(s): selftest_mptcp_join selftest_simult_flows 🔴:
  - Task: https://cirrus-ci.com/task/5307521400307712
  - Summary: https://api.cirrus-ci.com/v1/artifact/task/5307521400307712/summary/summary.txt

- KVM Validation: debug:
  - Unstable: 1 failed test(s): selftest_mptcp_join 🔴:
  - Task: https://cirrus-ci.com/task/6433421307150336
  - Summary: https://api.cirrus-ci.com/v1/artifact/task/6433421307150336/summary/summary.txt

Initiator: Patchew Applier
Commits: https://github.com/multipath-tcp/mptcp_net-next/commits/1d9ed6cb0b33


If there are some issues, you can reproduce them using the same environment as
the one used by the CI thanks to a docker image, e.g.:

    $ cd [kernel source code]
    $ docker run -v "${PWD}:${PWD}:rw" -w "${PWD}" --privileged --rm -it \
        --pull always mptcp/mptcp-upstream-virtme-docker:latest \
        auto-debug

For more details:

    https://github.com/multipath-tcp/mptcp-upstream-virtme-docker


Please note that despite all the efforts that have been already done to have a
stable tests suite when executed on a public CI like here, it is possible some
reported issues are not due to your modifications. Still, do not hesitate to
help us improve that ;-)

Cheers,
MPTCP GH Action bot
Bot operated by Matthieu Baerts (Tessares)

Re: [PATCH bpf-next] mptcp: Add struct mptcp_sock definition when CONFIG_MPTCP is disabled

[Martin KaFai Lau]

On Mon, Jul 11, 2022 at 03:07:31PM +0200, Jiri Olsa wrote:
> The btf_sock_ids array needs struct mptcp_sock BTF ID for
> the bpf_skc_to_mptcp_sock helper.
> 
> When CONFIG_MPTCP is disabled, the 'struct mptcp_sock' is not
> defined and resolve_btfids will complain with:
> 
>   BTFIDS  vmlinux
> WARN: resolve_btfids: unresolved symbol mptcp_sock
BTF_TYPE_EMIT(struct mptcp_sock) did not help here because
'struct mptcp_sock' is not defined anywhere in the include/ ?  :(

> 
> Adding empty difinition for struct mptcp_sock when CONFIG_MPTCP
> is disabled.
> 
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
>  include/net/mptcp.h | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/include/net/mptcp.h b/include/net/mptcp.h
> index ac9cf7271d46..25741a52c666 100644
> --- a/include/net/mptcp.h
> +++ b/include/net/mptcp.h
> @@ -59,6 +59,10 @@ struct mptcp_addr_info {
>  	};
>  };
>  
> +#if !IS_ENABLED(CONFIG_MPTCP)
> +struct mptcp_sock { };
> +#endif
Considering a similar case in the existing 'struct mptcp_out_options',
this seems reasonable.

Acked-by: Martin KaFai Lau <kafai@fb.com>

> +
>  struct mptcp_out_options {
>  #if IS_ENABLED(CONFIG_MPTCP)
>  	u16 suboptions;
> -- 
> 2.35.3
> 

Re: [PATCH bpf-next] mptcp: Add struct mptcp_sock definition when CONFIG_MPTCP is disabled

[Mat Martineau]

On Mon, 11 Jul 2022, Jiri Olsa wrote:

> The btf_sock_ids array needs struct mptcp_sock BTF ID for
> the bpf_skc_to_mptcp_sock helper.
>
> When CONFIG_MPTCP is disabled, the 'struct mptcp_sock' is not
> defined and resolve_btfids will complain with:
>
>  BTFIDS  vmlinux
> WARN: resolve_btfids: unresolved symbol mptcp_sock
>
> Adding empty difinition for struct mptcp_sock when CONFIG_MPTCP
> is disabled.
>
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
> include/net/mptcp.h | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/include/net/mptcp.h b/include/net/mptcp.h
> index ac9cf7271d46..25741a52c666 100644
> --- a/include/net/mptcp.h
> +++ b/include/net/mptcp.h
> @@ -59,6 +59,10 @@ struct mptcp_addr_info {
> 	};
> };
>
> +#if !IS_ENABLED(CONFIG_MPTCP)
> +struct mptcp_sock { };
> +#endif

The only use of struct mptcp_sock I see with !CONFIG_MPTCP is from this 
stub at the end of mptcp.h:

static inline struct mptcp_sock *bpf_mptcp_sock_from_subflow(struct sock *sk) { return NULL; }

It's normally defined in net/mptcp/protocol.h for the MPTCP subsystem 
code.

The conditional could be added on the line before the stub to make it 
clear that the empty struct is associated with that inline stub.

> +
> struct mptcp_out_options {
> #if IS_ENABLED(CONFIG_MPTCP)
> 	u16 suboptions;
> -- 
> 2.35.3
>
>

--
Mat Martineau
Intel

Re: [PATCH bpf-next] mptcp: Add struct mptcp_sock definition when CONFIG_MPTCP is disabled

[Matthieu Baerts]

Hi Jiri, Mat,

On 11/07/2022 23:21, Mat Martineau wrote:
> On Mon, 11 Jul 2022, Jiri Olsa wrote:
> 
>> The btf_sock_ids array needs struct mptcp_sock BTF ID for
>> the bpf_skc_to_mptcp_sock helper.
>>
>> When CONFIG_MPTCP is disabled, the 'struct mptcp_sock' is not
>> defined and resolve_btfids will complain with:
>>
>>  BTFIDS  vmlinux
>> WARN: resolve_btfids: unresolved symbol mptcp_sock
>>
>> Adding empty difinition for struct mptcp_sock when CONFIG_MPTCP
>> is disabled.
>>
>> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
>> ---
>> include/net/mptcp.h | 4 ++++
>> 1 file changed, 4 insertions(+)
>>
>> diff --git a/include/net/mptcp.h b/include/net/mptcp.h
>> index ac9cf7271d46..25741a52c666 100644
>> --- a/include/net/mptcp.h
>> +++ b/include/net/mptcp.h
>> @@ -59,6 +59,10 @@ struct mptcp_addr_info {
>>     };
>> };
>>
>> +#if !IS_ENABLED(CONFIG_MPTCP)
>> +struct mptcp_sock { };
>> +#endif
> 
> The only use of struct mptcp_sock I see with !CONFIG_MPTCP is from this
> stub at the end of mptcp.h:
> 
> static inline struct mptcp_sock *bpf_mptcp_sock_from_subflow(struct sock
> *sk) { return NULL; }
> 
> It's normally defined in net/mptcp/protocol.h for the MPTCP subsystem code.
> 
> The conditional could be added on the line before the stub to make it
> clear that the empty struct is associated with that inline stub.

If this is required only for this specific BPF function, why not
modifying this stub (or add a define) to return "void *" instead of
"struct mptcp_sock *"?

Cheers,
Matt
-- 
Tessares | Belgium | Hybrid Access Solutions
www.tessares.net

Re: [PATCH bpf-next] mptcp: Add struct mptcp_sock definition when CONFIG_MPTCP is disabled

[Jiri Olsa]

On Tue, Jul 12, 2022 at 11:06:38AM +0200, Matthieu Baerts wrote:
> Hi Jiri, Mat,
> 
> On 11/07/2022 23:21, Mat Martineau wrote:
> > On Mon, 11 Jul 2022, Jiri Olsa wrote:
> > 
> >> The btf_sock_ids array needs struct mptcp_sock BTF ID for
> >> the bpf_skc_to_mptcp_sock helper.
> >>
> >> When CONFIG_MPTCP is disabled, the 'struct mptcp_sock' is not
> >> defined and resolve_btfids will complain with:
> >>
> >>  BTFIDS  vmlinux
> >> WARN: resolve_btfids: unresolved symbol mptcp_sock
> >>
> >> Adding empty difinition for struct mptcp_sock when CONFIG_MPTCP
> >> is disabled.
> >>
> >> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> >> ---
> >> include/net/mptcp.h | 4 ++++
> >> 1 file changed, 4 insertions(+)
> >>
> >> diff --git a/include/net/mptcp.h b/include/net/mptcp.h
> >> index ac9cf7271d46..25741a52c666 100644
> >> --- a/include/net/mptcp.h
> >> +++ b/include/net/mptcp.h
> >> @@ -59,6 +59,10 @@ struct mptcp_addr_info {
> >>     };
> >> };
> >>
> >> +#if !IS_ENABLED(CONFIG_MPTCP)
> >> +struct mptcp_sock { };
> >> +#endif
> > 
> > The only use of struct mptcp_sock I see with !CONFIG_MPTCP is from this
> > stub at the end of mptcp.h:
> > 
> > static inline struct mptcp_sock *bpf_mptcp_sock_from_subflow(struct sock
> > *sk) { return NULL; }
> > 
> > It's normally defined in net/mptcp/protocol.h for the MPTCP subsystem code.
> > 
> > The conditional could be added on the line before the stub to make it
> > clear that the empty struct is associated with that inline stub.
> 
> If this is required only for this specific BPF function, why not
> modifying this stub (or add a define) to return "void *" instead of
> "struct mptcp_sock *"?

so btf_sock_ids array needs BTF ID for 'struct mptcp_sock' and if CONFIG_MPTCP
is not enabled, then resolve_btfids (which resolves and populate all BTF IDs)
won't find it and will complain

btf_sock_ids keeps all socket IDs regardles the state of their CONFIG options,
and relies that sock structs are defined even if related CONFIG option is disabled

if that is false assumption then maybe we need to make btf_sock_ids values optional
somehow

jirka

Re: [PATCH bpf-next] mptcp: Add struct mptcp_sock definition when CONFIG_MPTCP is disabled

[Matthieu Baerts]

Hi Jiri,

On 12/07/2022 14:12, Jiri Olsa wrote:
> On Tue, Jul 12, 2022 at 11:06:38AM +0200, Matthieu Baerts wrote:
>> Hi Jiri, Mat,
>>
>> On 11/07/2022 23:21, Mat Martineau wrote:
>>> On Mon, 11 Jul 2022, Jiri Olsa wrote:
>>>
>>>> The btf_sock_ids array needs struct mptcp_sock BTF ID for
>>>> the bpf_skc_to_mptcp_sock helper.
>>>>
>>>> When CONFIG_MPTCP is disabled, the 'struct mptcp_sock' is not
>>>> defined and resolve_btfids will complain with:
>>>>
>>>>  BTFIDS  vmlinux
>>>> WARN: resolve_btfids: unresolved symbol mptcp_sock
>>>>
>>>> Adding empty difinition for struct mptcp_sock when CONFIG_MPTCP
>>>> is disabled.
>>>>
>>>> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
>>>> ---
>>>> include/net/mptcp.h | 4 ++++
>>>> 1 file changed, 4 insertions(+)
>>>>
>>>> diff --git a/include/net/mptcp.h b/include/net/mptcp.h
>>>> index ac9cf7271d46..25741a52c666 100644
>>>> --- a/include/net/mptcp.h
>>>> +++ b/include/net/mptcp.h
>>>> @@ -59,6 +59,10 @@ struct mptcp_addr_info {
>>>>     };
>>>> };
>>>>
>>>> +#if !IS_ENABLED(CONFIG_MPTCP)
>>>> +struct mptcp_sock { };
>>>> +#endif
>>>
>>> The only use of struct mptcp_sock I see with !CONFIG_MPTCP is from this
>>> stub at the end of mptcp.h:
>>>
>>> static inline struct mptcp_sock *bpf_mptcp_sock_from_subflow(struct sock
>>> *sk) { return NULL; }
>>>
>>> It's normally defined in net/mptcp/protocol.h for the MPTCP subsystem code.
>>>
>>> The conditional could be added on the line before the stub to make it
>>> clear that the empty struct is associated with that inline stub.
>>
>> If this is required only for this specific BPF function, why not
>> modifying this stub (or add a define) to return "void *" instead of
>> "struct mptcp_sock *"?
> 
> so btf_sock_ids array needs BTF ID for 'struct mptcp_sock' and if CONFIG_MPTCP
> is not enabled, then resolve_btfids (which resolves and populate all BTF IDs)
> won't find it and will complain
> 
> btf_sock_ids keeps all socket IDs regardles the state of their CONFIG options,
> and relies that sock structs are defined even if related CONFIG option is disabled

Thank you for the explanation. I didn't know about that.

Then it is fine for me to leave it in mptcp.h. If it is not directly
linked to bpf_mptcp_sock_from_subflow(), I guess it can stay there but
maybe better to wait for Mat's answer about that.

> if that is false assumption then maybe we need to make btf_sock_ids values optional
> somehow

Cheers,
Matt
-- 
Tessares | Belgium | Hybrid Access Solutions
www.tessares.net

Re: [PATCH bpf-next] mptcp: Add struct mptcp_sock definition when CONFIG_MPTCP is disabled

[Mat Martineau]

[-- Attachment #1: Type: text/plain, Size: 2973 bytes --]

On Tue, 12 Jul 2022, Matthieu Baerts wrote:

> Hi Jiri,
>
> On 12/07/2022 14:12, Jiri Olsa wrote:
>> On Tue, Jul 12, 2022 at 11:06:38AM +0200, Matthieu Baerts wrote:
>>> Hi Jiri, Mat,
>>>
>>> On 11/07/2022 23:21, Mat Martineau wrote:
>>>> On Mon, 11 Jul 2022, Jiri Olsa wrote:
>>>>
>>>>> The btf_sock_ids array needs struct mptcp_sock BTF ID for
>>>>> the bpf_skc_to_mptcp_sock helper.
>>>>>
>>>>> When CONFIG_MPTCP is disabled, the 'struct mptcp_sock' is not
>>>>> defined and resolve_btfids will complain with:
>>>>>
>>>>>  BTFIDS  vmlinux
>>>>> WARN: resolve_btfids: unresolved symbol mptcp_sock
>>>>>
>>>>> Adding empty difinition for struct mptcp_sock when CONFIG_MPTCP
>>>>> is disabled.
>>>>>
>>>>> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
>>>>> ---
>>>>> include/net/mptcp.h | 4 ++++
>>>>> 1 file changed, 4 insertions(+)
>>>>>
>>>>> diff --git a/include/net/mptcp.h b/include/net/mptcp.h
>>>>> index ac9cf7271d46..25741a52c666 100644
>>>>> --- a/include/net/mptcp.h
>>>>> +++ b/include/net/mptcp.h
>>>>> @@ -59,6 +59,10 @@ struct mptcp_addr_info {
>>>>>     };
>>>>> };
>>>>>
>>>>> +#if !IS_ENABLED(CONFIG_MPTCP)
>>>>> +struct mptcp_sock { };
>>>>> +#endif
>>>>
>>>> The only use of struct mptcp_sock I see with !CONFIG_MPTCP is from this
>>>> stub at the end of mptcp.h:
>>>>
>>>> static inline struct mptcp_sock *bpf_mptcp_sock_from_subflow(struct sock
>>>> *sk) { return NULL; }
>>>>
>>>> It's normally defined in net/mptcp/protocol.h for the MPTCP subsystem code.
>>>>
>>>> The conditional could be added on the line before the stub to make it
>>>> clear that the empty struct is associated with that inline stub.
>>>
>>> If this is required only for this specific BPF function, why not
>>> modifying this stub (or add a define) to return "void *" instead of
>>> "struct mptcp_sock *"?
>>
>> so btf_sock_ids array needs BTF ID for 'struct mptcp_sock' and if CONFIG_MPTCP
>> is not enabled, then resolve_btfids (which resolves and populate all BTF IDs)
>> won't find it and will complain
>>
>> btf_sock_ids keeps all socket IDs regardles the state of their CONFIG options,
>> and relies that sock structs are defined even if related CONFIG option is disabled
>
> Thank you for the explanation. I didn't know about that.
>
> Then it is fine for me to leave it in mptcp.h. If it is not directly
> linked to bpf_mptcp_sock_from_subflow(), I guess it can stay there but
> maybe better to wait for Mat's answer about that.
>
>> if that is false assumption then maybe we need to make btf_sock_ids values optional
>> somehow
>

I'd rather keep the full mptcp_sock definition in net/mptcp/protocol.h 
since moving it would require also moving a few other structs it depends 
on.

Defining the empty struct in mptcp.h is fine with me and it sounds like 
that meets the needs of btf_sock_ids - but I'd like a v2 of this patch 
that moves the new empty struct declaration next to the inline 
btf_mptcp_sock_from_subflow function in mptcp.h

--
Mat Martineau
Intel