Re: carl9170 fw 1.9.3 on the binary archive page

Re: carl9170 fw 1.9.3 on the binary archive page

[Christian Lamparter]

Hello,

On Tuesday 24 May 2011 14:39:16 Marius Ladegaard Meyer wrote:
> I am currently maintaining packages in the AUR of Arch Linux for the
> carl9170 driver and firmware. For the firmware package I use the binary
> firmware files available at
> http://www.kernel.org/pub/linux/kernel/people/chr/carl9170/fw/ I see in the
> git repo of carl9170fw that version 1.9.3 was tagged two months ago, and I
> would like to provide the newest version to Arch users. Would it be possible
> for you to upload a binary to the aforementioned page please?

making new releases is not the issue. But "testing" everything in
every possible combination in different environments, with
different configurations and different devices is a huge task...

But I'm always baffled by distributions.
Why do they insist on having a binary firmware image, when
they can build one from scratch as well... After all, they do
so for every other application/packet, right?

Regards,
	Chr

Re: carl9170 fw 1.9.3 on the binary archive page

[Christian Lamparter]

On Tuesday 24 May 2011 20:49:17 Marius Ladegaard Meyer wrote:
> I'm sorry if I was not quite clear. I am not an "official" maintainer for
> Arch Linux in any way, I simply wrote a package script for other users to
> download from the Arch User Reposiroty. With the binaries on the mirror I
> linked to, this is almost a one-liner, whereas a personal build for every
> user is a little bit more complicated (since I'm not a dev I cannot push a
> binary package to the official repos, and the AUR forbids binaries).
It would be a easier if the firmware didn't need its own toolchain and
could use the one provided by the distro.

> But I figure now that I might want to ask the dev's at Arch to make a package
> of this instead. I tried once or twice to build the fw from git (and filed you
> a bug about a missing dir if you remember) but I did not end up with a
> useable carl9170-1.fw.
now you see, this is interesting. You tried once or twice to build the fw, but
the resulting firmware image was not usable in the end? Can you explain in
detail why it/they was/were "not usable"?

Did the driver reject the image?
Did it you get a "-110" (timeout) during init?
interface created? scan works? can connect?
...

or is it more like:
"actually stuff works pretty much like before, but because this is not
official and it doesn't work "better" it must be broken."

Regards,
	Chr

Re: carl9170 fw 1.9.3 on the binary archive page

[Christian Lamparter]

On Wednesday 25 May 2011 21:26:32 Marius Ladegaard Meyer wrote:
> I built the firmware again today. After building the toolchain, I ran the
> autogen.sh script and selected the default values on all the things I was
> asked about. The script completed succesfully and created the files
> carl9170.bin, carl9170.dsc, carl9170.elf and carl9170.fw in the carlfw
> folder. I have attached the output in a text file.
output looks sane.

> It is at this point I cannot figure out what to do. The carl9170-1.fw files
> on the mirror just needs to be put in the /lib/firmware directory, and the
> kernel will find them when I plug in the device. I tried to copy all the
> aforementioned carl9170.* files to /lib/firmware, but all I see from dmesg
> is
> 
> [ 5416.799516] usb 2-6: new high speed USB device using ehci_hcd and address
> 5
> [ 5417.092852] usb 2-6: reset high speed USB device using ehci_hcd and
> address 5
> [ 5417.257236] usb 2-6: firmware not found.

the firmware filename is important. The driver can only load the firmware if it
is stored in the firmware folder with the right filename: "carl9170-1.fw"

the "-1" was introduced as a result of a rather long discussion about
firmware versioning best practices:
http://wireless.kernel.org/en/developers/Documentation/firmware-versioning

anyway cd into carl9170fw-source/carlfw and
cp carl9170.fw /lib/firmware/carl9170-1.fw
[If /lib/firmware is the correct directory]

this should do the trick.
> and iwconfig gives me no interface. I'm 100% sure that the error is between
> the chair and the keyboard on this one, but I cannot find any documentation
> on this anywhere.
oh no, you are 100% right. This was not documented anywhere in plain English,
but now it should be [just pushed].

Thanks,
	Chr