Re: Regression with the latest iwlwifi-9260-*-46.ucode

[Luca Coelho <luca@coelho.fi>]

On Fri, 2019-08-09 at 22:00 +0300, Thomas Backlund wrote:
> Den 06-08-2019 kl. 16:04, skrev Takashi Iwai:
> > On Mon, 05 Aug 2019 14:03:55 +0200,
> > Now we got a feedback from the latest linux-firmware (20190726) and
> > surprising the result was negative.  The dmesg after the cold boot is
> > found at:
> >    https://bugzilla.suse.com/show_bug.cgi?id=1142128#c26
> > 
> > The kernel is 5.2.3, so it should be new enough.
> > 
> > If anything else needed (or something missing), let us know.
> > 
> 
> I read on some forum that some commented that the "Add support for SAR 
> South Korea limitation" fix is needed, but it seemed weird...
> 
> Anyway, with theese on top of 5.2.7
> 
> 39bd984c203e86f3  iwlwifi: mvm: don't send GEO_TX_POWER_LIMIT on version 
> < 41
> f5a47fae6aa3eb06  iwlwifi: mvm: fix version check for GEO_TX_POWER_LIMIT 
> support
> 0c3d7282233c7b02  iwlwifi: Add support for SAR South Korea limitation
> 
> 
> We have confirmation from an affected user that its now stable with both 
> older and newer firmwares...
> 
> And we earlier tried with only the:
> 39bd984c203e86f3  iwlwifi: mvm: don't send GEO_TX_POWER_LIMIT on version 
> < 41
> 
> But that did not help (not that I really expected it since its loading 
> version 46 firmwares anyway)
> 
> So my guess is that the newer firmware actually subtly expects to get 
> the behaviour of the:
> 0c3d7282233c7b02  iwlwifi: Add support for SAR South Korea limitation
> 
> Of course that's still guessing and I assume only Intel fw guys can 
> verify that...

Yes, you need the 3 patches.  The first two should solve the
"BAD_COMMAND" issue and the last one fixes the "NMI_INTERRUPT_WDG"
issue.

The first two are already in v5.3-rc4 and in v5.2.9 stable.

I'm going to send the third one to stable now.

--
Cheers,
Luca.