On Wed, Jan 7, 2026 at 4:45 PM Oliver Hartkopp <[email protected]> wrote:
>
> Hi Francesco,
>
> On 06.01.26 21:39, Francesco Valla wrote:
> > Hello Harald, Oliver,
> >
> > On Tue, Jan 06, 2026 at 08:43:42PM +0100, Oliver Hartkopp wrote:
> >>
> >>
> >> On 06.01.26 17:50, Harald Mommer wrote:
> >>>> With the plain 'cangen' you are not really flooding the interface, since
> >>>> you are only sending a random CAN frame every 200ms. The only way I can
> >>>> reproduce this behaviour in a consistent manner is running from the host:
> >>>>
> >>>>       while true; do cansend vcan0 134#00; done
> >>>>
> >>>> which seems to generate the maximum amount of traffic.
> >>>>
> >>>> This is not of course a realistic bus load, but is leading the system
> >>>> (at least on my setup) to a corner case somewhere.
> >>>
> >>> I have no idea how long the shell needs for a loop, always used cangen -g 
> >>> 0 to stress the setup which is most probably faster than the shell 
> >>> interpreter, and sometimes did this for both directions (RX and TX).
> >>>
> >>> Full load is a realistic setup. And even if it was not, if something 
> >>> stopped working or worse crashes torturing the setup this was a problem.
> >>>
> >>
> >> Yes. cangen -g 0 -i <interface> creates full load - even on real CAN
> >> interfaces. You can also generate fixed content if you want to omit the
> >> generation of randomized content. 'cangen -?' prints a help text.
> >>
> >
> > I agree with both of you - I was simply arguing that a plain 'cangen'
> > with no parameters is not really loading the interface.
> >
> > For some reason, I was only able to trigger the unwanted behavior with
> > cansend in a while loop and not with cangen -g 0, even with fixed ID and
> > payload. However, I suspect the issue is a matter of timing and
> > coincidences rather than load level.
>
> Yes, the difference is, that you open a new CAN socket each time with
> cansend, while cangen opens one socket and pushes lots of frames into it.
>
> Btw. this should not lead to a stuck CAN interface.
>
> Is the interface usable from another terminal (with cansend/candump) or
> how does this 'stuck interface' look like?
>

I reported the issue here:
https://github.com/rust-vmm/vhost-device/issues/923. It seems it is
the device implementation. I am working on a fix.

Matias


Reply via email to