On Mon, Jun 25, 2018 at 02:55:51PM +0200, Lorenzo Bianconi wrote:

quoted

Hi all,

Hi Stanislaw,

quoted

On Mon, Apr 09, 2018 at 04:48:19PM +0200, Stanislaw Gruszka wrote:

quoted

On Mon, Apr 09, 2018 at 04:26:42PM +0200, Lorenzo Bianconi wrote:

quoted

quoted

I would like to integrate the driver to kernel via mt76 driver, i.e.
add USB hooks and mt76x0 mac/phy code to mt76. This will open
possibility to develop support for mt76x2 USB devices as well as mt76x0
PCIe devices in mt76.

I have already started supporting mt76x2 USB devices in mt76 since register map
is pretty similar to PCIe devices:
https://github.com/LorenzoBianconi/wireless-drivers-next/tree/mt76x2u
I added some usb utility routines so I think we can integrate mt76x0 in mt76 as
well

Great, I'll start to integrate mt76x0 on top of your tree.

So I started to do integration here:
https://github.com/sgruszka/wireless-drivers-next/commits/mt76x0-draft

However since driver is self containging, I think better would be just
submit the driver into mt76/mt76x0/ dir upstream and do code merging work as
follow-up patches posted on the mailing list. Patches then could be reviewed
on regular basic. This will provide support for new mt76x0 devices in kernel
quicker. Conflicts with mt76x2u and not yet upstreamed mt7603 could be resolved
on the fly.

I did a quick review of the code and it seems (please correct me if I
am wrong) there is
a lot of duplicated code with mt76/mt76x2u and mt7601u drivers (i.e:
mcu/eeprom/mac code

Yes, however there are some subtle differences too.

is quite the same of the ones used in mt76x2u). Moreover mt76/mt76x2u has been
refactored in order to expose usb and mt76x2_common modules where you
can use better
802.11 aggregation (using mac80211 per-sta queuing) and A-MSDU support (using
tx/rx usb scatter-gather). Moreover mt76x2u has been tested/used by
various users till now.
So since mt76x0 will be deeply modified I guess it would be better to
start integrating the driver with
mt76/mt76x2u before been merged upstream otherwise will end-up with a
lot of integration commits.

Not sure why many integration commits in upstream is a problem. I think
having patches posted on mailing list is better than doing them in my
"private" tree without any review.

What do you think?

I was thinking about posting mt76x0 driver in a subdir (there is sill 
some cleanup work need to be done there), wait for upstream mt76x2u
integration, then post patches that remove duplication between mt76x2
and mt76x0 and add support for mt76x0e on the way.

Thanks
Stanislaw