Re: [Xenomai] Kernel oops during rtnet loopback usage on x86_64 (e1000e)

[Pintu Kumar <pintu.ping@gmail.com>]

> FWIW, all of my rtnet fixes that went in since 3.0.6 have been tested on
> x86 exclusively.

For us, even the xeno-test/net_udp fails on both x86/arm.
I also verified the udp test (loopback) that exists on rtnet repo:
https://git.code.sf.net/p/rtnet/code
Under: examples/generic/

Then, I also developed a common rtnet udp server/client using posix skin.
When I build and verify it on normal kernel it works fine.
But when I build it with Xenomai posix skin, I get the kernel oops on
both x86/arm.

Now it fails with xenomai-3/next (latest) and with Kernel 4.9, on x86.
Earlier (1.5 months ago), I remember, rtnet loopback worked on x86_64
with 3.0.6, with "nosmap".
 If I run the same on Virtual Box Ubuntu 32-bit with same Kernel 4.9, it works.

So, I wonder, normally how do you verify rtnet sample application.
Please share the test applications and the steps you follow.


Thanks,
Pintu

On Wed, Jun 27, 2018 at 11:40 PM Philippe Gerum <rpm@xenomai.org> wrote:
>
> On 06/27/2018 07:29 PM, Pintu Kumar wrote:
> > On Wed, Jun 27, 2018 at 10:47 PM Greg Gallagher <greg@embeddedgreg.com> wrote:
> >>
> >> I don't think Beaglebone is supported currently in RTNet.
> >>
> >
> > But I am just checking loopback interface with 127.0.0.1
> > I hope that should work on all boards and arch, even without
> > installing the net driver.
> >
>
> You must be the first and only user giving any feedback about rtnet over
> ARM if my memory serves me well. So either all others have been hiding
> behind for all these years, or you may be the only specimen of that species.
>
> FWIW, all of my rtnet fixes that went in since 3.0.6 have been tested on
> x86 exclusively.
>
> --
> Philippe.