Libre wifi for Ben Nanonote (was: Re: Libre GNU/Linux PDAs)

Wolfgang Spraul wolfgang at sharism.cc
Sat Jan 16 23:26:29 EST 2010


Daniel,

> I'm vaguely aware of this, but since the driver is fully GPL most
> likely people will eventually get it working quite well; this seems to
> be what happened with ath9k, which sucked for a long time but in
> 2.6.32 seems to be very nice.

I highly doubt that will happen with the 6001. I don't want to go into the
details of the firmware revisioning here etc. It just won't happen.
Why do you think the KDDI microSD card (which may never be produced) has
a 6001? The 6001 is an old chip and not recommended or supported for new
designs. The 6002 moved away from flash so you need to load a binary blob
at boot time (more about that below).

> If no one from sharism.cc wants to do this, I can try, perhaps in
> concert with a friend who does speak Japanese.

Not at the top of my list right now.

> The binary blob makes the Spectec SDW-823 unusable and unpromotable by
> organizations and people I care about (fsf, rms). Lack of free as in
> freedom wifi on the ben nanonote would make the device much less
> useful.

Hmm. Good thing the Ben NanoNote has no Wi-Fi.
I don't think we should second guess what the fsf or rms would exactly
say about this, I'm listening... (do you have any recent statements?)

To me the difference between binary blob loaded at boot time, and binary
blob stored in flash, is no difference at all from a free as in freedom
perspective - _if_ the redistribution license for the binary blob is OK.
The binary blob in the KS-7010 chip is about 80 KiB, making it a good
target for reverse engineering. It may not use a typical ARM instruction
set though, but some more proprietary Japanese thing. Who knows.
Anyway reverse engineering this kind of thing won't succeed ultimately,
so if we continue down the SDW-823 path we would probably try to work with
Renesas.
Not on GPL'ing the firmware (won't happen) but on offering a thin firmware
or other firmware features that allow the Linux kernel more control over
the chip.

If the fsf or rms don't want to promote the Spectec SDW-823, fine by me.
Hopefully they then promote the USRP instead, or some other attempts at
making copyleft chips.
Think about it this way - the binary blobs are a reminder that there is
more software running that is non-free. My reaction to this wouldn't
be "let's hide it" (in flash such as in the Atheros 6001), but it would be
"let's write a free replacement".

Good that you mention the FSF: Mirko may know more about this but we looked
at their endorsement criteria for the Ben NanoNote, and think they are not
thorough enough for the type of copyleft hardware we are aiming for.
It seems for the FSF a device is endorceable if non-free software is hidden
well enough. That's very unfortunate, the FSF should not hand out awards
to those that are best in hiding non-free software in a device.

The FSF spent a lot of time thinking about the relatively subtle differences
between static and dynamic linking. Maybe they feel they want to focus on
that rather than going deeper into the messy embedded world.
The good news is, even without leadership from the FSF in this area, many
good things are happening, take USRP as an example.
The SDW-823 (or other current Wi-Fi) cards are intermediate steps. If we
get them to work well with our devices right now, I think that's also a
good thing. Even if in the long run I totally agree with you we could
do a lot of fun things if the firmware would be easily rebuildable from
source.

> It would be good if you could get Renesas to GPL the SDW-823 firmware :-)

See above. First step is to make it more Linux-friendly, 'thin firmware' so
that the Linux kernel can have more control. This will be at odds with
power consumption (suspend more behavior), but it's a first step.

> Note that while this may seem like a little thing, the nonfree
> libertas firmware ended up being a big issue / time waster for a
> similar project, the OLPC XO.

Sure I know. Big time waste for everyone, including Marvell. But what's your
point? What is OLPC doing now?

> It was so bad that they ended up trying to reverse engineer the
> firmware, but I don't think they ended up being very successful with
> that.

Correct.

> I can find references to specific problems / bug former XO devs to
> talk about this if qi-hardware people are interested.

He, you can point them to our project, the Ben NanoNote - always welcome!
Thanks for your feedback, very much appreciated!
Wolfgang

On Sun, Jan 17, 2010 at 09:51:01AM -0500, Daniel Clark wrote:
> On Sat, Jan 16, 2010 at 6:57 AM, Wolfgang Spraul <wolfgang at sharism.cc> wrote:
> >> It looks like there is an Atheros AR6002 microSD wifi card [1,2,3,4];
> >> anyone have contacts with Japanese companies / KDDI / Mitsumi
> >> Electric?
> >
> > Openmoko used Atheros (6001) in the FreeRunner, and it was not a good
> > experience, for a number of reasons. Everybody tried hard, including
> > Atheros, but in the end we didn't make it. I think Atheros 5xxx and
> > 9xxx have fairly good Linux drivers, and the 2425 as well (former ZyDAS).
> > 6xxx is a troublemaker.
> 
> I'm vaguely aware of this, but since the driver is fully GPL most
> likely people will eventually get it working quite well; this seems to
> be what happened with ath9k, which sucked for a long time but in
> 2.6.32 seems to be very nice.
> 
> > For the KDDI card, can you order it? AFAIK it was a 'demo-only' product
> > so far and no real production is planned.
> [...]
> > But that's several steps down the road, not sure how far we make it.
> > If someone finds out news about the KDDI card, or gets in touch with them,
> > please keep the list posted we are definitely interested...
> 
> It is not orderable now by consumers at the moment. I think I read it
> was planned to be available sometime early this year, but who knows.
> My idea is that if someone affiliated with an OEM (such as sharism.cc)
> contacted Mitsumi Electric (perhaps someone who spoke Japanese as well
> would help), they would be able to do a bulk order or reseller
> arrangement, and/or get more information about actual planned
> availability.
> 
> If no one from sharism.cc wants to do this, I can try, perhaps in
> concert with a friend who does speak Japanese.
> 
> > The only generally available and working microSD Wi-Fi card I am aware of
> > is the Spectec SDW-823. It was out of stock for a few months at the end
> > of last year, but now available again in Taiwan.
> > Spectec's #1 distributor is Expansys, with fulfillment centers globally.
> > At Expansys the SDW-823 is listed as "due Jan 28" right now (on the UK
> > site), I guess if people put in orders at Expansys this should move
> > forward. The cards definitely exist in stock in Taipei.
> 
> Spectec SDW-823 also seems to be available from ebay in the US.
> 
> > We are evaluating how well those cards actually work in our Ben NanoNote.
> > Antenna sensitivity, power consumption, driver features.
> > The Spectec SDW-823 has a Renesas (former KeyStream) KS-7010 Wi-Fi chip
> > inside. GPL driver, binary firmware blob.
> 
> The binary blob makes the Spectec SDW-823 unusable and unpromotable by
> organizations and people I care about (fsf, rms). Lack of free as in
> freedom wifi on the ben nanonote would make the device much less
> useful.
> 
> > We do have fairly good relations with the KeyStream department inside
> > Renesas, if the evaluation of the SDW-823 on our NanoNotes comes out
> > positive, we will probably go back to Renesas and talk about new features
> > for the firmware (all the things Linux people like, such as thin firmware,
> > monitor/AP mode, mesh, etc).
> 
> It would be good if you could get Renesas to GPL the SDW-823 firmware :-)
> 
> Note that while this may seem like a little thing, the nonfree
> libertas firmware ended up being a big issue / time waster for a
> similar project, the OLPC XO.
> 
> It was so bad that they ended up trying to reverse engineer the
> firmware, but I don't think they ended up being very successful with
> that.
> 
> I can find references to specific problems / bug former XO devs to
> talk about this if qi-hardware people are interested.
> 
> Happy hacking,
> -- 
> Daniel JB Clark | http://pobox.com/~dclark
> 
> _______________________________________________
> Qi Developer Mailing List
> Mail to list (members only): developer at lists.qi-hardware.com
> Subscribe or Unsubscribe: http://lists.qi-hardware.com/cgi-bin/mailman/listinfo/developer




More information about the discussion mailing list


interactive