[Click] Click and kernel patches

[Eddie Kohler]

Hi Lars,

Lars Bro wrote:
> Hi Eddie
> 
> I am working on a test bench for our product, and I use a server with 
> several network interfaces where I send packets on one interface and 
> then listen where and when these packets arrive. Basically, I have 
> written a "Correlator" element that keeps packets arriving on input 0 
> for a certain time and checks packets arriving on all other inputs 
> against these. The output is packets with correlation information: 
> packet # xxx with paint anno yyy were seen on these input ports at with 
> these delays.
> The correlation output is sent as UDP via ToHost to the recording 
> application.
> 
> So, I can do with "FromDevice", "ToDevice" and "ToHost".
> 
> I certainly think that the "unmodified" kernel is the way to go, and 
> would very much like to help testing it.
> 
> One question, though: So far, I have intercepted a network connection by 
> using two ports, and configuring a bridge "brctl addbr..." When loading 
> Click, the bridge stops working, but of course FromDevice ... ToDevice 
> is used within the Click configuration to make the same bridge with 
> whatever interception is needed. I guess it is a bad idea to have the 
> bridge module loaded at all, or will Click still be able to "take over" 
> somehow?

So I believe that Click will take over from the bridge module, although any 
bridge that you've set up will stop working -- Click ignores the "brctl". 
Nevertheless, it will be safer to run without the bridge module loaded.

Eddie


> 
> yours,
> Lars Bro
> 
> On Thu, Feb 11, 2010 at 2:08 AM, Eddie Kohler <kohler at cs.ucla.edu 
> <mailto:kohler at cs.ucla.edu>> wrote:
> 
>     Hi Lars,
> 
>     My plan for the future is to further develop the linuxpatchless
>     infrastructure, which lets us run Click on unmodified kernels.
> 
>     So I would love for people to test this, such as you.  Let us know
>     how it goes.  The only thing missing that I know of is FromHost.
> 
>     Eddie
> 
> 
>     Lars Bro wrote:
> 
>         Hi,
> 
>         I am using Click for testing of our wireless product, and I
>         experience an
>         increasing demand for support of newer kernels. The reason for
>         this is
>         hardware of course, but also wireless(802.11) drivers have been
>         changing a
>         lot lately.
> 
>         I am now working with Matteo Croce's patches for 2.6.31 (on
>         2.6.31.9), this
>         is not as easy as I had thought. I can also see that Adam
>         Greenhalgh has
>         done some work in separating syntax changes from core changes.
> 
>         What I would like to know is whether this kind of work is
>         already ongoing,
>         or if there is an official plan for the next kernel patch.
> 
>         As I can see, Matteo's patch also includes changes for Click
>         itself, and
>         this should be very carefully considered.
> 
>         So I suggest that we agree on a "next" kernel version to
>         support, so we can
>         work on it in common.
> 
>         Any votes?
> 
>         yours,
>         Lars Bro
>         _______________________________________________
>         click mailing list
>         click at amsterdam.lcs.mit.edu <mailto:click at amsterdam.lcs.mit.edu>
>         https://amsterdam.lcs.mit.edu/mailman/listinfo/click
> 
>