[OLSR-users] Mips-Processor: Plugins will SEGV

[Andreas Tønnesen]

Hi Sven,

First of all - I highly recommend not using the CVS version for plugin
testing. There has been a lot of olsrd internal changes and most of the
plugins require quite some rewrites to work with this version(this will
ofcause be fixed before the next release).

This issue has been discussed before:
http://www.olsr.org/pipermail/olsr-users/2004-October/000233.html
But I don't think any conclusions were drawn there... But you could check
out the mips-i vs. mips-ii issue mentioned there.
I guess it could be something uclibc related. I will put it on my todo
list - but it will probably be some time before I can look into it.

Regarding the ampersand in the function pinter assignments I believe the &
is optional in this case. To me it looks better using it - but as with
most such things I guess it's a matter of taste :)

- Andreas

> Hi,
>
> recently, I've read the posting about the dyn_gw on this list. This sounds
> promising, so I tried to run that plugin on my WRT54g. But: loading the
> plugin will stop olsrd from running with a SEGV message. Seems to be some
> traditional bug in olsrd. Further investigation reveals:
>
> - Plugin runs OK on Intel (both Win and Linux)
> - Plugin fails miserably on Mips (both WRT54g and Meshcube)
> - Plugin fails both for 0.4.7 and for 0.4.8 (cvs version 30-oct)
>
> Further debugging will help here, so I've patched
> lib/dyn_gw/src/olsrd_plugin.c:
>
> *** Snippet from olsrd-0.4.8-cvs-2004-10-30 ***
> int
> register_olsr_data(struct olsr_plugin_data *data)
> {
>   /* IPversion */
> fprintf(stderr, "register_olsr_data(%p)\n", data);
>   ipversion = data->ipversion;
> fprintf(stderr, "ipversion(%d)\n", ipversion);
>   /* Main address */
>   main_addr = data->main_addr;
> fprintf(stderr, "main_addr(%d)\n", main_addr);
>
>   /* Multi-purpose function */
> fprintf(stderr, "olsr_plugin_io gleich!!!\n");
> fprintf(stderr, "1olsr_plugin_io(%d)\n", data->olsr_plugin_io);
>   olsr_plugin_io = data->olsr_plugin_io;
> fprintf(stderr, "2olsr_plugin_io(%d)\n", olsr_plugin_io);
>
> *** /Snippet ***
>
> This will output the following if started on MIPS:
>
> ---------- Plugin loader ----------
> Library: olsrd_dyn_gw.so.0.1
> OLSRD dynamic gateway plugin 0.1 by Andreas Tønnesen
> Checking plugin interface version.... 1 - OK
> Trying to fetch plugin IO function....OK
> Trying to fetch register function....OK
> Running registration function...
> register_olsr_data(0x7fff6c50)
> ipversion(2)
> main_addr(268449888)
> olsr_plugin_io gleich!!!
> 1olsr_plugin_io(4267792)
> Segmentation fault
>
> I'am fiddling a bit with variable storage classes, and have some success
> with adding "static" and the " = 0" initializer in the *.h where
> "olsrd_plugin_io" is defined. However, this is not the only function
> pointer
> used, so it's segfault again. Anybody knows, if this is a gcc cross
> compiler
> bug? Or just something with DS!=SS so one have to use compiler overwrites
> "CALLBACK global do not assume Data and Stack on same segement" here?
> Another comment: Is there a superfluous ampersand in the "olsrd_plugin_io
> =
> &whatever_function;"? My C++/ASM knowledge is bailing out here, especially
> the parentesis jungle used to define function pointers in C/C++ alway
> confuses me...
>
> Regards, Sven-Ola
>
> _______________________________________________
> olsr-users mailing list
> (spam-protected)
> https://www.olsr.org/mailman/listinfo/olsr-users
>


---------
Andreas Tønnesen
http://www.olsr.org