On Sun, Aug 02, 2020 at 11:58:26AM +0200, Stefan Sperling wrote:
> On Sat, Aug 01, 2020 at 09:16:55PM -0700, Bobby Reynolds wrote:
> > Hi Misc!
> > 
> > Just purchased an Intel AX200 so I could try out the new support in 6.7.
> > First and foremost, many thanks to those working on the wifi stack :)
> > 
> > I ran into some firmware issues with the 6.7 release, but I saw some
> > relevant entries in the -current changelog, and the snapshot from
> > Monday worked quite well! Unfortunately, after running sysupgrade this
> > afternoon my iwx0 device is no longer working. It shows up in dmesg but
> > not in ifconfig output, and I don't have network connectivity.
> > 
> > Any tips on how I might further debug the issue? I'm rather new to
> > OpenBSD but quite comfortable with debugging and digging into source
> > code.
> 
> > "Intel Wi-Fi 6 AX200" rev 0x1a at pci1 dev 0 function 0 not configured
> 
> I needed to tweak the iwx device matching code for AX201 because Intel
> is playing silly games with PCI product IDs.
> 
> Please send me the output of pcidump -vvv so I can make sure that
> your device is properly detected again.

Nevermind, after taking another close look at the Linux code I could
already commit a fix. So please compile a kernel from -current or wait
for a future snapshot. Your device should then work again.

In case of AX201 we need to be careful about which devices are matched by
the driver because there are technical differences such as which firmware
image is required. In addition to matching the PCI product ID, the driver 
must also match the PCI subsystem product ID against a table of known values.

For consistency I wrote similar matching code for AX200 devices, and it
correctly matches my AX200 device.

But in the AX200 case, the Linux driver uses subsystem product IDs only for
marketing purposes: It makes the operating system advertise some AX200 devices
with distinct trademarked names. There are no technical differences so we can
just go back to matching any AX200 device as we did before.

Sorry about the breakage.
Going forward, I'll try to keep in mind that some things drivers written
by vendors will do are not done for technical reasons...

Reply via email to