On Wed, 24 Aug 2011 11:02:53 +0200, Florian Fainelli <flor...@openwrt.org>
wrote:
> Hello Andrew,
> 
> On Wednesday 24 August 2011 07:34:19 Andrew wrote:
>> Hello all,
>> I'm currently working on the rtl8196 OpenWRT port. My initial patches
can
>> be found here:
>> http://invyl.ath.cx/cgit/cgit.cgi/openwrt/
>> 
>> I'm currently using external rsdk-linux- toolchain, that came along
with
>> the bsp from the trendnet router.
>> The actual hardware is a edup pocket router.
>> It's kind of old: (gcc 3.4).
> 
> This toolchain has two problems from an OpenWrt perspective: it is in
> binary 
> form and it is old. Fortunately, I have some patches around to support
> lexra 
> core based on gcc-4.1.2, which should be easy to forward port.

No, it's not definetely binary-only... I had to accomplish quite a quest 
to make it compile a month or so ago. uClibc looks quite recent (0.9.30). 
I haven't diffed it to vanilla, yet. 

>> 
>> The current state is:
>> * Initial stuff commited to OpwnWRT makefiles,
>> * Initial patches to make the kernel somewhat compile are ready.
>> I'm now working out the linker problems and writing the image packager
>> Makefile
>> 
>> My questions here are:
>> Is OpenWRT okay with this archeological gcc? Or should I expect more
>> weird
>> bugs?
> 
> I think you will hit some strange bugs.
> 
>> Since the CPU is LexraCore it misses some patented instructions so
usual
>> mips-linux- toolchain won't work
>> without hacing it. There used to be a patch at
>> gcc maillists, but it doesn't look like it made it into the main
branch.
> 
> No it's definitively not in mainline gcc, I do not think anyone ever
> submitted 

http://gcc.gnu.org/ml/gcc-patches/2005-05/msg01249.html
The only thing I found.

> patches to get it supported, though this should not be too hard. The
> patents 
> have now expired (they expired in dec. 2006 iirc), so there should not
be
> any 
> political issue anymore.
> 
>> 
>> What conserns the kernel, I tried to apply the minimal set of patches
to
>> make this stuff work.
>> However, some subsystems are quite a mess. This is the current state:
>> * board support is sh*tty in the bsp, symlinked to out of kernel dir
and
>> with little board-specific code.
>> needs a full rewrite.
> 
> Yes, it does.
> 
>> * mtd dev support needs some cleaup (shouldn't be that difficult)
>> * lan and wifi are messy, but kind of work, I just didn't have the
>> ocasion
>> to have a closer look at them
> 
> I have a device here based on the RTL856x core, which is using some
binary 
> blobs to do layer-4 offloading, this should hopefully not concern you,
> since 
> RTL8196 has a real Ethernet MAC.

The only blob I saw was fw that loads into wireless. May be I didn't look
good enough.
 
>> * gpio support sucks: they made it a char dev incompatible with mostly
>> anything. Board-specific code in the driver
>> itself. Needs a full rewrite.
>> * Some files in the kernel are symlinks to something like
>> /home/$codemonkeyname/somedir..../file.c. Hopefully I've
>> cleaned that all.
> 
> If you are interested, I can send you clean patches which add support
for 
> RTL856x, on which you could also based yours, because RTL8196 is not
that 
> different from RTL865x

Yes, that would be awesome. Thanks. 

Regards, 
Andrew
_______________________________________________
openwrt-devel mailing list
openwrt-devel@lists.openwrt.org
https://lists.openwrt.org/mailman/listinfo/openwrt-devel

Reply via email to