New Xilinx Drivers, GNSS Reciver USB Driver & Habana Labs Updates Land In Linux 5.17

[phoronix]

Phoronix: New Xilinx Drivers, GNSS Reciver USB Driver & Habana Labs Updates Land In Linux 5.17

Sent in to the Linux kernel on Friday were the "char/misc" updates as the smorgasbord of kernel changes not fitting formally within other areas of the kernel. The char/misc changes range from AI accelerator driver updates to new Xilinx code to other random changes littered throughout...

https://www.phoronix.com/scan.php?page=news_item&px=Linux-5.17-char-misc

[milkylainen]

Don't understand how a generic USB driver for GNSS is implemented?
There is nothing generic about GNSS receivers?

If the driver was a NMEA-parsing USB-serial port or something then yes,
but I don't see why that would need to live in kernelspace.

A Sierra Wireless GNSS USB driver can be implemented, but that is hardly generic.

[tuxd3v]

A Sierra Wireless GNSS USB driver can be implemented, but that is hardly generic.

What I take from it, is that there exist some reference design of USB GNSS, and Sierra probably used that reference design to implement it.

So if it was implemented using a hardware reference design , it will be a generic driver.
If everybody that implements GNSS via USB, follow the reference design, they will be able to use the same driver..

Now the problem would come for those that decide to deviate from the reference design, those probably will have troubles using the generic driver..

[numacross]

Don't understand how a generic USB driver for GNSS is implemented?
There is nothing generic about GNSS receivers?

If the driver was a NMEA-parsing USB-serial port or something then yes,

It looks like that's exactly what it is.

[milkylainen]

What I take from it, is that there exist some reference design of USB GNSS, and Sierra probably used that reference design to implement it.

So if it was implemented using a hardware reference design , it will be a generic driver.
If everybody that implements GNSS via USB, follow the reference design, they will be able to use the same driver..

Now the problem would come for those that decide to deviate from the reference design, those probably will have troubles using the generic driver..

I took a look at the code. It doesn't really do anything.
A GNSS device and read and writes to the USB descriptor.
Doesn't actually do any NMEA parsing, or anything useful related to a GNSS intrinsics...
or anything else for that matter.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/gnss

I mean, I get that the kernel needs a GNSS-layer but this is like...
I don't get it? Johan is an experienced developer.

Do I get my own device layer pushing bits and bobs around without doing anything useful in the vanilla tree?
This could easily been done out of tree until it matures quite a bit more.

[tuxd3v]

I took a look at the code. It doesn't really do anything.
A GNSS device and read and writes to the USB descriptor.
Doesn't actually do any NMEA parsing, or anything useful related to a GNSS intrinsics...
or anything else for that matter.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/gnss

I mean, I get that the kernel needs a GNSS-layer but this is like...
I don't get it? Johan is an experienced developer.

Do I get my own device layer pushing bits and bobs around without doing anything useful in the vanilla tree?
This could easily been done out of tree until it matures quite a bit more.

You right, there are only the usb stuff..
It will probably come in the months ahead..