1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

New Linux Kernel?

Discussion in 'Tomato Firmware' started by arrmo, Dec 13, 2013.

  1. arrmo

    arrmo Addicted to LI Member

    Hi,

    Just a though, but ... is anyone else interested in trying to evolve towards a more current kernel? I assume the right way to do it would be to have patches, that apply to the kernel - then updates would be easier.

    Thoughts? Or a crazy idea?

    Just thinking out loud ... :).

    Thanks!
     
  2. RMerlin

    RMerlin Network Guru Member

    Can't be done unless someone has a few thousand dollars to buy an SDK from Broadcom, with the wireless driver source code.
     
  3. eahm

    eahm LI Guru Member

    How much exactly?
     
  4. RMerlin

    RMerlin Network Guru Member

    No idea. I don't know either if you can obtain it without being an actual company.

    The point was that you cannot change to a different kernel without having the source code for the wireless driver and related components.
     
  5. leandroong

    leandroong Addicted to LI Member

  6. Victek

    Victek Network Guru Member

    Ralink chipset (RT-N56U) evolved to actual kernels updating their drivers ...
     
  7. koitsu

    koitsu Network Guru Member

    Respectfully (but a bit tongue-in-cheek): sounds to me like you're volunteering to study it. Let us know the results. :)
     
  8. Marcel Tunks

    Marcel Tunks Networkin' Nut Member

    There are probably people at Asus, Belkin/Linksys, and Netgear who think that new kernel features might be useful. None of their routers seem to be on 3.x. I'm guessing that Broadcom's licenses have some unfavorable contractual obligations or excessive costs associated with them. Otherwise I would say that a small crowdfunding campaign would be fun.

    If people are desperate for new kernel features, then they should probably move to an openwrt derivative or a software solution (eg. Debian firewall.)
     
  9. arrmo

    arrmo Addicted to LI Member

    Hi,

    Good discussion - thanks folks! But two more dump questions (my speciality unfortunately ... :(),
    - are the interfaces really changing? Just not sure, but it impacts if we can use "old" drivers or not.
    - how do releases like OpenWRT achieve this?

    Thanks again!
     
  10. leandroong

    leandroong Addicted to LI Member

  11. RMerlin

    RMerlin Network Guru Member

    Moving to a new kernel for Asus would mean losing HW acceleration, since that module is pre-compiled by Broadcom for the specific kernel Broadcom supports in their SDK. So, it won't happen.

    What "new kernel features" do you need anyway on a home router?
     
  12. RMerlin

    RMerlin Network Guru Member

    Because as pointed out, Padavan is for completely different hardware than what Tomato supports.
     
  13. RMerlin

    RMerlin Network Guru Member

    OpenWRT uses very old driver code, which has its whole set of problems associated with it.

    As for DD-WRT, Newmedia has an SDK to develop with.
     
  14. leandroong

    leandroong Addicted to LI Member

    I think, padavan FW has that HW acceleration, under firell->Netfilter there is option for
    "Maximum Connections", options available as follow
    1. 8192
    2. 16384 (HW_NAT FoE Max)
    3. 32768
    4. 65536
    5. 131072 (slow)
    6. 262144 (very slow)
     
  15. Marcel Tunks

    Marcel Tunks Networkin' Nut Member

    None in SOHO settings, so I use Tomato! :)

    Would be fun to play with the bufferbloat fixes and to eventually play with NFTables, but I don't have access to a network that needs those features. Maybe some day when I quit my day job...

    @leandroong Padavan and the hardware on which it runs are different from the ground up. An analogy is that jets and helicopters both fly, but you can't swap parts between them. Your enthusiasm is appreciated, but it's just not going to happen. Enjoy Padavan, and enjoy Tomato!
     
  16. RMerlin

    RMerlin Network Guru Member

    I read a bit on NFTables a few months ago, and I'm really not a fan. It sounds like your typical design done by an engineer who wants to do something that looks cool, but is far less practical than existing solutions, all in the name of "it looks nifty on paper". Most basic operations that takes one line with 3 parameters in iptables will take 3x the amount of typing, which means 3x the chances of mistakes.
     
  17. Marcel Tunks

    Marcel Tunks Networkin' Nut Member

    Too bad. It did look nifty on paper. ;)
     
  18. koitsu

    koitsu Network Guru Member

    {OFF-TOPIC}
    The sooner someone ports OpenBSD pf or FreeBSD pf (they are very similar but not quite the same internally) to Linux, the happier all Linux people will be. Trust me, pf beats the living crap out of iptables as far as simplicity and understanding what does what. Sorry for advocating BSD kinda out of left field, but Linux's firewalling stacks have been an abysmal clusterf*** since the original introduction of ipfwadm. pf is absolutely amazing and "just makes sense" when you look at it. Of course, all the iptables modules on Linux that offer lots of great functionality, many of them do not exist for pf, so that would be a lot of work.

    If someone wants to see an actual live/real pf.conf on a production box, just let me know. I'm sure if I paste it, anyone even remotely familiar with firewalls or the concept of firewalls could understand it. Note that I don't have much experience with using NAT for pf, but supposedly the rules are just as simple (one-liners that just make sense).
    {/OFF-TOPIC}
     
  19. eahm

    eahm LI Guru Member

    I own two IT companies, that's why I was asking :)
     
  20. mstombs

    mstombs Network Guru Member

    Trouble is you will have to sign an NDA, so won't be allowed to tell us, I suspect no fixed pricelist - all dependent on expected sales/market sector. Broadcom seem to blow hot and cold with opensource there are occasional releases but never quite enough! They don't deal with end customers and I suspect they only really want to deal with the likes of Cisco, Belkin etc. We are still looking for technical information on the BCM4706/BCM5300 ejtag commands so we can get JTAG working for RT-N66U etc if you do get technical contact with them!
     
  21. Victek

    Victek Network Guru Member

    Contact me if you get any information... have one RT-N66U CFE bricked... ;)
     
  22. mstombs

    mstombs Network Guru Member

    I have a RT-N66U with JTAG pins, which is not bricked! It was dead when I got it but after some open-heart surgery to replace a blown regulator it seems fine, but not sure want to use for anything but development. I stll have 4 spare FR9888 s, but you need know someone with the tools/skills to be able to replace (an old solderer, not me!). Can identify the CPU with tjtag, zjtag or Urjtag , but that might be after CFE boot code puts it into MIPS mode.

    Back on topic Broadcom tantalizingly give away a header file giving clues

    http://repo.or.cz/w/tomato.git/blob...389:/release/src-rt-6.x/include/hndjtagdefs.h

    but we don't know how to use the MACROS and the multiple modes the cpu can work in. Presumably access to the Broadcom SDK would allow a router manufacturer to build/debug using JTAG as well as build wireless drivers.
     

Share This Page