OpenWrt Forum Archive

Topic: BCM 63xx chips - ADSL State of art

The content of this topic has been archived between 6 Feb 2018 and 7 May 2018. There are no obvious gaps in this topic, but there may still be some posts missing at the end.

It is interesting.
But it will be better if you will try to describe which register responsible for which operation.
States are good, but they may be not accurate because of different line parameters, connection type, etc.

I think you may use source code in this work, because your aim is only to get DSL hardware specification (thing which Broadcom should not hide from public).
If we will got this kind of specification, road for developing driver will be open.

But I will say again, is it possible, that Netgear has permission to public source code?

What im trying to say is, we take those registers at different stages, This may be a good read: http://www.mediafire.com/?aq0gew97tl2o548

[1] We can take a gpl tarball released by the manufacturer of 'ADSL Modem Router X'.
[2] We take that tarball and make changes needed to stop the drivers loading at startup that are responsible for, ATM http://en.wikipedia.org/wiki/Asynchronous_Transfer_Mode ADSL http://en.wikipedia.org/wiki/Asymmetric … riber_Line.
[3] We create a linux driver http://www.google.co.uk/search?q=linux+driver+writing. Which as i mentioned in my previous post: "creates a couple of char devices that we can use to dump the atm processor and adsl registers".
[4] We load the adsl drivers when we are ready.
[5] We use the usual web based or command line configuration interface to change settings,
[6] We use the char devices to dump the registers as a connection is attempted/made/in progress. Maybe we could make the char device update every time a change is made to the registers. Or on a timer?
[7] We compare changes, different vpi/vci, connection state, etc with the dumps of the registers.

I dont do C, im still learning. I do understand the hardware and physical connections on electronic devices, so i can understand asembly, memory and registers but C is just too much. I have quite a few broadcom based routers. I know the score with linux, bricking, jtag, etc. So im well up for frying a few routers.

Maybe I understand something wrong, but I think ATM is protocol (data link layer services that run over OSI Layer 1 physical links).
Also, as I know there are open source ATM drivers for linux.
And "ATM processor" and "ATM' are not the same things.

You are saying about ATM processor and ADSL processor registers.
But how you now that there are registers? Maybe they has some kind of io ports, DMA access, etc.
If you are saying about the list and names of registers, than you know the specification of hardware.
So if you know detailed specification of hardware, we need nothing else. Just write code.

We can't  dump registers, because we do not know how to access them, how many registers and so on.
If we will be able to read registers, then sure we can use described steps.

PDF doc in http://en.wikipedia.org/wiki/Asymmetric … riber_Line describes "ATM protocol" but not any device driver.

You can't get the full spec or datasheets without paying large sums of cash to broadcom. What i said before is pretty much standard routine for reverse engineering hardware.

I can tell you where the atm processor registers start, 0xfffe2000 for 6338 and 6358. 0xfffe4000 for 6348.
I can tell you where the adsl registers start, 0xFFFE1000 for 6338. 0xFFFE3000 for 6348 and 6358.
I think the adsl io mem starts at 0xFFF00000.

If i could find a way to see whats going on without programming anything in C i would have already started on the spec.

That's looks good to rewrite DSL from the scratch, but as for me, this work can take a long time to make job done. And can happen on the finish we will have DSL driver for discontinued chipsets. Why not adopt existing supported kernel for openwrt build tree? Take a brief look to the neufbox Openwrt fork https://forum.openwrt.org/viewtopic.php?pid=107660 IMHO, it can be a good start point to have working Openwrt on our ADSL roters.

OMFG... roll

You lot just keep completely missing the point.

This is about openness.

If you want proprietary drivers tainting your kernel then why not just keep the frigging factory firmware.

I have done a shit load of research but it seems nobody here has the same initiative as me.

My girlfriend gave birth over the weekend so im not going waste any more time on this.

Sure, birth is much more important thing.
Wish you and you girlfriend easy and happy birth.

I understand your point of view.

But seems that you do not understand the volume of work.
For my rough evaluation it may take a lot of time. Think near a half of year.
Write (1 day) - test (1 day) - inform you (1 day) - write ...
Heave some experience in remote work. So think this project may take a lot of time.
So as for me I do not want to hurry to start this work. So I was thinking about what to do next.
Also here were holidays.

That is why I was thinking that adopt the existing code ("clean room") is much faster and much quality approach.
There is not many desire persons to develop. Maybe it is explained by flowing. Person who wants to use ADSL in OpenWRT just takes this sources and rebuilds OpenWRT with them. For home use (in not USA or other developed country) there is no big reason if the router clean licensed or not.

If you do not mind, I may start some simple work. As I understand, for first step you need the driver to dump registers of device.
Think it is simple and will not take much time. I may use native SDK from D-Link and create char device or TCP port to dump values.
Have some time now, so will do this exercise.

Sorry guys i havn't had much sleep over weekend so i had a bitch on last night.

Fellas, i understand that it will not be easy but, many hands make light work. The only people on the planet who really do understand how much work is necessary are the broadcom employees/contractees who wrote the the original drivers.

See guys, i dont think an easy, quick and dirty hack is the answer here. We need to open this up once and for all.

I have got the diff of broad-com kernel and generic pristine one done. Taken some effort to make sure that the code is what portable to newer ones. I successfully compiled bits and pieces of them, today I am expected to look at the DSL part of it.
I have been  partying around from Christmas to New Year and  been in holiday mood smile
New year greetz to you all too....
I have got some free time for next few days... and going slowly.. but surely .. moving forward.

I don't read all the thread so apologize me if this information is yet know but a little (I belive) group of people is yet working on this topic at
http://bcm63xx.sipsolutions.net/

Rock on, fairly recent stuff. I stopped checking that place out when they finished the 43xx spec. They are the guys who can.

Oh well, might as well file this thread under /dev/null then.

(Last edited by routednbooted on 11 Jan 2011, 23:39)

Is there any progress with ADSL drivers in opnewrt? Can you tell me? posts are from january.

No.
Not started yet.
If it will be done on Netgear sources (will be not simple actually), it will be not GPL. Just howto to build for yourself.

I decide to do no write from scratch.
Just will select devices carefully when buying...

ddlk wrote:

No.
I decide to do no write from scratch.
Just will select devices carefully when buying...

The problem is: how many devices can you select with opensource adsl drivers? . Or, more interesting for me, opensource VoIp drivers..

I am not familiar with drivers architecture, so cant even guess.
The best source is manufacturer...

Broadcom Joins the Linux Foundation

It was exciting news back in September when Broadcom unveiled brcm80211, the fully open and Linux-compatible driver for several of its 802.11n wireless chipsets. Now, however, it looks like the company will soon make what promises to be an even bigger announcement.

Similar Articles:

Linux Kernel Update Brings a New Wave of Hardware Support.
Linux Wi-Fi Gets Easier with New Broadcom Driver.
Linux and Windows 8's Secure Boot: What We Know So Far.
Ubuntu 11.04 'Natty Narwhal': A First Look at the OS.
Linux Mint 11 and Fedora 15: In One Week, Two Gems Debut.
Linux Foundation: Secure Boot Need Not Be a Problem.
Specifically, in a move that will be officially announced on Monday, the company has taken its Linux support to the next level by joining the Linux Foundation, with plans to extend its open development and collaboration with the Linux community.

Broadcom will continue to work on the Linux Driver Project, and it will also participate in The Linux Foundation Collaboration Summit, enabling it to work directly with community developers as well as other industry players and suppliers.

"There is no question: Linux has become a major platform for communications devices and technologies," said Michael Hurlston, senior vice president and general manager for Broadcom's WLAN line of business. "Our decision to open source the drivers for Broadcom's 802.11 chipsets is in response to our growing base of customers using Linux and is the first of what we expect to be many open development success stories."

Problem Solved

Historically, wireless networking has been a sticking point for some Linux users, including those with netbooks and laptops incorporating Broadcom chipsets, which typically have used proprietary drivers that don't work with the free and open source operating system.

Competing vendors such as Atheros and Intel began supporting Linux natively sooner than Broadcom did, so the company's move last fall was met with resounding applause.

Since the release of that new open driver, it has been integrated into the recent Linux kernel release 2.6.37, where it can be actively improved upon by the entire Linux community. It was also included in Ubuntu 10.10, or Maverick Meerkat, last fall.

Moving forward, Linux users of hardware with Broadcom chips will no longer have to give wireless another thought.

A Rapidly Growing Roster

Now, by joining the Linux Foundation, Broadcom is clearly signaling its intent to continue such efforts in the long term--a particularly exciting prospect given its portfolio of semiconductors for wired and wireless communications.

The addition of Broadcom, in fact, is one of many important additions to the Linux Foundation in recent months. Perhaps most notably, China Mobile joined the group in November, followed by Huawei and Mentor Graphics in December, for example.

IP management company Protecode and Timesys both joined this week, and Cybercom and GoAhead will join Broadcom in announcing their own membership next week.

As I noted back in September, it's all a testament to the growing power of Linux as a market force. Businesses are turning to it in increasing numbers, and so are individuals. It's great to see companies like Broadcom recognize that this is a significant market that can no longer be ignored.

I think no one gonna enable it's BCM 63xx chips with OpenWRT fw soon.

Some one know about ADSL driver?

some one knows help bump

No news on ADSL, no support existing, no support planned and no support possible in the foreseeable future. You can stop bumping this.

The discussion might have continued from here.