Re: [PATCHv2] backports: Update hwsim defconfig

Mathy Vanhoef <[email protected]> Wed, 18 May 2022 15:52:38 +0200
Newsgroups org.kernel.vger.backports
Message-ID <20220518155238.42d1d464@zbook-mathy>
On Mon, 11 Apr 2022 09:58:31 +0200
Johannes Berg <[email protected]> wrote:

> On Mon, 2022-04-11 at 15:39 +0800, Mathy Vanhoef wrote:
> > In the current hwsim config no rate control algorithm is enabled.
> > This means that, when using defconfig-hwsim, the mac80211_hwsim
> > fails to load with the following error:
> > 
> > 	[  132.155183] ieee80211 phy0: Failed to select rate
> > control algorithm [  132.155196] ieee80211 phy0: Failed to
> > initialize rate control algorithm
> > 
> > Fix this by enabling rate control algorithms in the provided hwsim
> > config.  
> 
> Makes sense, thanks.
> 
> > Experiments were also less reliable when WEXT was enabled. So this
> > patch also disables WEXT when using defconfig-hwsim to make the
> > default more reliable for users.
> >   
> 
> Huh? How so? Maybe some other tool was present on the system that was
> using wext, and disabling wext made that tool no longer detect the
> wireless or so? I cannot see how else it would be "less reliable" with
> wext enabled, if you don't really use it.

I checked this in more detail: when using kernel 5.13.0-30-generic on
Ubuntu 20.04 with backports-5.15.5-1 the kernel crashes when adding
interfaces in case WEXT has been enabled in backports (in my case when
inserting mac80211_hwsim in a virtual machine). The problem is that
wdev->netdev is NULL inside the function cfg80211_init_wdev, causing
the kernel to hang. The same happens on kernel 5.15.40-1-lts on Arch
Linux (this is currently the most up-to-date LTS kernel on Arch Linux).

Interestingly, on kernel 5.13.0-41-generic on Ubuntu 20.04 this isn't a
problem because the field wdev->netdev is then initialized.
--
To unsubscribe from this list: send the line "unsubscribe backports" in