On Sat, 27 Aug 2011 13:02:14 +0200, Florian Fainelli <[email protected]> wrote: > Hello, > > On Saturday 27 August 2011 12:45:41 Andrew wrote: >> Hello all, >> I'm currently working on rtl8196 port, and those weird guys from realtek >> implemented there a different architecture rlx in kernel. >> Basically, much is borrowed from mips, but small differences here and >> there are present. I applied openwrt command hack to rlx, however, that >> didn't work out of the box. >> To get patch-cmdline actually patch rlx's vmlinux I had to widen the >> SEARCH_SPACE ( >> http://invyl.ath.cx/cgit/cgit.cgi/openwrt/commit/?id=38a13e566f179a12803a39 >> 1ca941cad1f806db08 ). The next thing was, that despite the image gets >> patched, and I see the correct text hexdump'ing the image, the kernel >> command line stays the same. Looks like some other place needs to get >> patched, but I just can't firure where to look. Any ideas? > > The bootloader most likely overrides the command-line as well, so make > sure > your prom code concatenates both command-line.
Got it, I'll try to investigate. >> Thanks in advance. >> >> I have now fixed the image packing for initial support for different >> boards, and, hopefully I'll get spi flash up&running next week. What >> should >> the state of the port be, before it's ok to send in the patches to the >> trunk? :) > > I had a quick look at the files in your repository: > > - please merge support for rtl8196c and rtl8196b, they do not differ that > much This is just the original realtek's code. If you're talking about stuff in mips/ it looks like dead code, since it doesn't even get compiled. > - please use the patches that I posted on the mailing-list as much as you > can, > the code is much cleaner and integrates better into the existing mips > architecture (special checksum, memcpy functions etc ...) > - remove all the dead code ifdefed'out with #if 0 I see. I'll try to diff rsdk gcc with the original and see if they have some useful optimisations there I can use as well. Cleanup is subject of the future work, since there's a lot of redundant and dead code. (I have a strange feeling the whole arch/mips part is dead in realtek's tree) > - port your patches to 2.6.39 at least, 3.0 would be even better One really big question here: should I try go on with their RLX architecture in kernel, or should I try to go off with plain mips? It does look diffing rlx and mips that they DO have some differences. Mostly commented mips code here and there. > - remove all the unnecessary hacks introduced by patches-2.6.30, they are > not > required > > - clarify the license of the tools located in image/imgtools, they do not > seem > to be open source (they have no license at all) This is the worst part of it. All the files I have are from TRENDnet TEW622-BRP GPL release (Should the GPL cover the tools, or not?), but I've got no idea how should I clarify those (other then rewriting from scratch which is pain) Regards, Andrew _______________________________________________ openwrt-devel mailing list [email protected] https://lists.openwrt.org/mailman/listinfo/openwrt-devel
