bpf.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [QUESTION] TCP connected socket selection
@ 2021-01-11 18:33 Konstantinos Kaffes
  2021-01-11 19:19 ` Jakub Kicinski
  0 siblings, 1 reply; 3+ messages in thread
From: Konstantinos Kaffes @ 2021-01-11 18:33 UTC (permalink / raw)
  To: bpf

Hi everyone,

It is the first time I am posting to a kernel mailing list so please
let me know if this question needs to be directed elsewhere.

I have been using BPF to programmatically steer UDP datagrams to
sockets using the "sk_reuseport" hook.

Similarly, I would like to identify request boundaries within a TCP
stream/connection and programmably forward requests to different
sockets *after* a connection is established. Is there a way to do that
in the kernel using BPF?

Thank you in advance,
Kostis

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [QUESTION] TCP connected socket selection
  2021-01-11 18:33 [QUESTION] TCP connected socket selection Konstantinos Kaffes
@ 2021-01-11 19:19 ` Jakub Kicinski
  2021-01-11 19:48   ` Konstantinos Kaffes
  0 siblings, 1 reply; 3+ messages in thread
From: Jakub Kicinski @ 2021-01-11 19:19 UTC (permalink / raw)
  To: Konstantinos Kaffes; +Cc: bpf

On Mon, 11 Jan 2021 10:33:49 -0800 Konstantinos Kaffes wrote:
> Hi everyone,
> 
> It is the first time I am posting to a kernel mailing list so please
> let me know if this question needs to be directed elsewhere.
> 
> I have been using BPF to programmatically steer UDP datagrams to
> sockets using the "sk_reuseport" hook.
> 
> Similarly, I would like to identify request boundaries within a TCP
> stream/connection and programmably forward requests to different
> sockets *after* a connection is established. Is there a way to do that
> in the kernel using BPF?

Sounds like what KCM does.

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [QUESTION] TCP connected socket selection
  2021-01-11 19:19 ` Jakub Kicinski
@ 2021-01-11 19:48   ` Konstantinos Kaffes
  0 siblings, 0 replies; 3+ messages in thread
From: Konstantinos Kaffes @ 2021-01-11 19:48 UTC (permalink / raw)
  To: Jakub Kicinski; +Cc: bpf

On Mon, 11 Jan 2021 at 11:19, Jakub Kicinski <kuba@kernel.org> wrote:
>
> On Mon, 11 Jan 2021 10:33:49 -0800 Konstantinos Kaffes wrote:
> > Hi everyone,
> >
> > It is the first time I am posting to a kernel mailing list so please
> > let me know if this question needs to be directed elsewhere.
> >
> > I have been using BPF to programmatically steer UDP datagrams to
> > sockets using the "sk_reuseport" hook.
> >
> > Similarly, I would like to identify request boundaries within a TCP
> > stream/connection and programmably forward requests to different
> > sockets *after* a connection is established. Is there a way to do that
> > in the kernel using BPF?
>
> Sounds like what KCM does.

Thanks! KCM would work but it is unclear to me how I can specify the
policy used by the multiplexor to match messages to specific sockets.

The documentation has examples on how to do the delineation but
nothing on how to specify the matching policy.

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2021-01-11 19:49 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-01-11 18:33 [QUESTION] TCP connected socket selection Konstantinos Kaffes
2021-01-11 19:19 ` Jakub Kicinski
2021-01-11 19:48   ` Konstantinos Kaffes

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).