[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Debian hardware certification


I started a thread on debian-private, because I didn't know where to
post. So there's already a (very short) discussion that started. Here's
my original message:

On 05/21/2011 07:11 PM, Thomas Goirand wrote:
> Hi there,
> I didn't know where to post, so I'm posting here. Let me know where I
> should post, if I'm doing wrong.
> I was at the CCCC (China Cloud Computing Conference) in Beijing the
> last 3 days. There, I had a chat with Nick, from chinaskycloud which
> works tighly with SuperCloud (a Supermicro subsidiary in China).
> We talked about the possibility to have their hardware being
> certified as compatible with Debian, and have them advertize about
> it on their website product pages.
> The plan would be to test the hardware (probably with a live CD
> using a KVM over IP). If it doesn't work, see what driver isn't
> present, and if the backported kernel has the fix. If it does,
> in some cases, we could add a patch in a Debian point release, if
> it's not too intrusive.
> Having a hardware certified program would increase adoption of Debian
> among server users. It will also help Debian fans to buy the correct
> hardware they need.
> So:
> 1/ Do we have already such hardware certification program?
> 2/ If we don't, can we start one?
> 3/ If we make one, can someone volunteer to make a logo?
> 4/ Is there any volunteer to test hardware, as only me doing
> it wouldn't be enough?
> 5/ Could we create a front desk for such a hardware certification?
> Your thoughts are welcome. I'll wait for suggestions until I reply
> to Nick.
> Thomas Goirand (zigo)

And here's my follow-up with some of the answers:

On 05/21/2011 08:00 PM, Bastian Blank wrote:
> They can already declare Debian a supported system. They have to do the
> support for that and can work with Debian to get the work done.

That's what we were discussing. Who should they get in touch with? I am
not aware of any front desk for such hardware certification.

> They are
> even allowed to use the Debian Open Use Logo to show this.

My point wasn't to use a "normal" Debian logo, but a new one
specifically for the purpose of saying that Debian runs on the hardware.

> are you aware that all of the supermicro motherboards with built-in
> KVM-over-IP functionality provide that functionality with the support
> of an on-board processor which runs an embedded linux-based system?
> supermicro attempts to comply with the GPL by publishing the tarballs
> provided by supermicro's upstream vendors:
>     ftp://ftp.supermicro.com/GPL/
> but i would be hard pressed to tell you how to use those tarball dumps
> to produce the firmware images provided for actual usage on individual
> controllers here:
>     http://www.supermicro.com/support/bios/firmware0.aspx
> more importantly, the tarball dumps and firmware images contain
> non-free blobs (in particularly, compiled java byte code).

I agree that is a big issue, and we should never vouch for that.
However, that's not the goal here.

> while i prefer supermicro products to those produced by their
> competitors, i would object to the use the use of the term "correct
> hardware" here or the use of the debian project's name to officially
> endorse a product containing non-free software.

The issue is that we wont ever find a server with a fully open BIOS, at
least not tomorrow. Intel didn't make the same kind of announcement as
AMD did with coreboot support, so it's going to get hard to get things
totally open. And the IPMI thing is tightly related to the BIOS (the
BIOS can change the IP of the KVM over IP).

I however agree. I think we can set different level of correctness. One
that would say Debian has been tested and it runs on the hardware
(boots, network is working, hard driver controller works), and another
one that certifies the product as open source compliant and we endorse
it. The later could come after the former.

On 05/21/2011 09:53 PM, Marco d'Itri wrote:
> Hardware certification does not just mean that the OS boots, but that
> if there are any driver-related bugs the entity who certified the
> software will fix them.

This is precisely the goal. The discussion I had with Nick was to let DD
have access to such hardware, and see if things can get fixed if there's
issues. I already did such a work testing a patch for the network driver
of X8STi-F (thanks to the help of some people on the kernel lists), and
the result was that the Intel e1000e 82574L network driver patch from
2.6.27 got backported in the Debian 2.6.26 kernel, and included in Lenny
5.0.4 point release (as it was a small, non-intrusive, patch). This was
done thanks to the help of my Californian hardware supplier that gave me
access to the hardware. What I'm thinking about is doing it on a bigger

I do believe that the kernel team does good enough work to fix issues by
the way. They did in the past...

Thomas Goirand (zigo)

P.S: Please keep Nick as Cc, as I don't believe he reads

Reply to: