[LiSA-Devel] Stability of latest master branch

[Radu Rendec]

On Thu, 2013-09-26 at 14:06 +0300, Claudiu Ghioc wrote:

> So I took a look firstly on the code which builds the running
> configuration. It seems to crash in
> "build_config_global" (cfgbuild.c:281) when calling the switch API
> function "igmp_get". Now, if we look at this function, it receives as
> parameters the sw_ops structure, a buffer to be filled with
> information and a pointer to an integer variable where it will return
> if IGMP snooping is activated. At the above mentioned line the second
> parameter is NULL (the buffer). I believe that the purpose of this
> piece of code is to determine whether IGMP is on or off.
> The "linux" implementation does not check the buffer parameter and it
> crashes when it tries to copy a bitmap to the buffer. Should I change
> the "igmp_get" function to be able to receive NULL as a second
> parameter and to check only if IGMP is on or off?

I've just looked at the "lisa" backend implementation. It copies the 2nd
parameter directly into the kenrel argument structure, so basically it
hands over the problem to the kernel code. The kernel code does check if
the argument is NULL.

So just to keep the API consistent across all backend implementations,
the function should be modified to check for NULL as well.

What about the other issue (not being able to add new interfaces into
the switch)?

Thanks,

Radu