The problem seems to be that CONFIG_RTNET_SELECT_SUPPORT is not defined when
compiling rtnet.
In the rtnet configure script, this is automatically defined when
$CONFIG_XENO_OPT_RTDM_SELECT is set to 'y', but I have no idea how this is
supposed to be set. It was certainly not set when I configured rtnet.
On Tue, Oct 12, 2010 at 4:27 PM, Gilles Chanteperdrix <
gilles.chanteperd...@xenomai.org> wrote:
> Johan Cockx wrote:
> >
> >
> > On Tue, Oct 12, 2010 at 3:38 PM, Gilles Chanteperdrix
> > <gilles.chanteperd...@xenomai.org
> > <mailto:gilles.chanteperd...@xenomai.org>> wrote:
> >
> > Gilles Chanteperdrix wrote:
> > > Johan Cockx wrote:
> > >>
> > >> On Tue, Oct 12, 2010 at 3:04 PM, Gilles Chanteperdrix
> > >> <gilles.chanteperd...@xenomai.org
> > <mailto:gilles.chanteperd...@xenomai.org>
> > >> > Ok: __real_socket is not called, the
> > XENOMAI_SKINCALL3(...) code
> > >> > returns zero in this case.
> > >>
> > >> Ok. zero should be translated in something like 1024 - 128.
> > And the
> > >> reverse operation happen in select code.
> > >>
> > >>
> > >> If I understand you correctly, the socket call is returning a
> valid
> > >> Xenomai file descriptor as it should. So, the next question is:
> > why is
> > >> this file descriptor (or the corresponding mask) not recognized
> > as such
> > >> in select()? Do you have any further suggestions on how to debug
> > this?
> > >
> > > As I just explained. verify that __wrap_socket returns 1024 - 128,
> or
> > > something like that. And that this value gets translated back to 0
> in
> > > select code.
> >
> > Normally, since there is only one file descriptor in the fdset, the
> test
> > which fails should be first_fp_valid_p.
> >
> >
> > I guess you are looking at the kernel code? I was still looking at the
> > user space code for __wrap_select. The arguments of that function are
> > as expected: nfds=897, and the bit for file descriptor 896 is set in
> > __readfds. I guess the XENOMAI_SKINCALL5 macro expands to a kernel
> > call, but sorry I have little experience with kernel programming. Does
> > this go straight to the __select function in skins/posix/syscall.c? I
> > can see the first_fp_valid_p call there.
>
> You can assume that it goes straight. It is a bit more complex than
> that, but the details do not really matter for this case.
>
> >
> > This is the function where I previously put a printk call. I just
> > checked dmesg again, and the printk'd message now appears (it seems
> > that `dmesg | tail -f` isn't working correctly for me, but `dmesg |
> > tail` is). Also, when I change syscall.c, a `make-kpkg --rootcmd
> > fakeroot --initrd kernel_image|`| at the top of the source tree somehow
> > does not rebuild anything. Until now, I have done a clean + complete
> > rebuild for the whole kernel for each change, but that takes quite some
> > time. Is there a better way to do this?
>
> Well, build your kernel without make_kpkg. But you can put printks all
> over the places in select code, this will save you compilations.
>
> --
> Gilles.
>
------------------------------------------------------------------------------
Beautiful is writing same markup. Internet Explorer 9 supports
standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 & L3.
Spend less time writing and rewriting code and more time creating great
experiences on the web. Be a part of the beta today.
http://p.sf.net/sfu/beautyoftheweb
_______________________________________________
RTnet-users mailing list
RTnet-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/rtnet-users
