Re: [RTnet-developers] Test programs for SOCK_RAW and ETH_P_ALL

[Jan K. <jan...@we...>]

Jorge Almeida wrote:
> Jan,
>=20
> In attach follows the test cases for the SOCK_RAW and ETH_P_ALL feature=
s. Its two files written in C code.
> One for sending messages, and other to read they using the ETH_P_ALL fe=
ature.

Thanks, will have a look.

>=20
> Please test the patch i send to you, after the refactoring that you nee=
d to do, and comment.

Well, if you read my comment, this is first of all about fixing your
patch. Therefore I would appreciate some comments on my findings as well
(disagree, ack, answers to my questions...). Even better would be an
updated version that is at least correct wrt/ rtskb_clone and resolves
the layering issue between stack_mgr.c and af_packet.c.

>=20
> The problem about the sending messages still happen, i only can send 16=
 messages. Please check if you have the same problem in one of your machi=
nes.
>=20
> In the loopback device i can send several messages (my test send 1000) =
but none is received in the rcv program.=20
> The loopback device works in a different way?

Yeah, loopback device is likely no good idea as it by-passes the stack
manager, see the code. That driver actually needs to be included in the
ETH_P_ALL patch as well, but we could do this later by extracting the
common features of the stack manager loop and the loopback driver into
one common function.

Jan