Hi Jonas,
On 10/02/2025 18:23, Jonas Smedegaard wrote:
> Quoting Matthieu Baerts (2025-02-09 23:49:37)
>> 9 Feb 2025 22:15:50 Jonas Smedegaard <d...@jones.dk>:
>>> Quoting Matthieu Baerts (NGI0) (2025-02-09 19:56:27)
>>>> iwd doesn't provide all the same features as the ones from
>>>> wpa_supplicant, especially everything not related to the Wireless world,
>>>> e.g. Ethernet authentication (bug#956457), or some more specific
>>>> features like MACsec.
>>>>
>>>> wpa_supplicant and iwd can then be used in parallel, for different
>>>> purposes.
>>>
>>> When you install only the iwd package, then you can also install the
>>> wpasupplicant package, and carefully configure them to not step on each
>>> others' toes.
>>>
>>> What the package network-manager-iwd offers is relieving the user of
>>> manual configuration: It is ensured that iwd works together with
>>> network-manager, but since network-manager recommends wpasupplicant,
>>> it is not adequate to provide a network-manager config snippet, because
>>> wpasupplicant will still be installed and will in its default
>>> configuration interfere with the default configuration of iwd.
>>
>> I see, but in fact, thanks to the config file provided by this package,
>> at the next reboot, NM will use iwd and leave wpasupplicant alone.
>> So no conflicts.
>>
>> iwd is just one WiFi backend that is loaded after having read the config.
>> NM will not try to use both at the same time.
>
> I might be wrong, but my understanding (and as I recall my personal
> experience as well) is not a problem of which helper tool
> network-manager interacts with, but instead that wpasupplicant by
> default loads itself as a daemon which somehow (listening to same
> socket, I guess) causes iwd to fail to load or operate reliably.
>
> See https://wiki.debian.org/NetworkManager/iwd for related info.
I guess you can have the conflict if you set IWD as the new backend,
then restart NM without stopping wpasupplicant that was already launched
from a previous instance. (I don't know if it can still be an issue today.)
If there is a reboot, the issue will not occur (that's what I had on my
side) because wpasupplicant will not be used as backend by NM.
In other words, there is a risk of issues only during the session where
IWD has just been installed, when the NM service is restarted but the
wpasupplicant one is left running. If someone wants to use IWD right
away, it will have to stop wpasupplicant as mentioned in the wiki. Do we
need to cover this case?
I don't think it is needed, but then a post-install script could also
stop wpasupplicant and restart NM.
Now that I'm thinking about that, with the conflict, wpasupplicant will
be removed, but NM will not be restarted and will try to continue using
it. I don't think it is a good situation. Keeping wpasupplicant seems
safer. (Or a post-install script is needed.)
>>> Thanks for the proposal, but I disagree with this one.
>>
>> (I think there is a small mixed-up in the changelog file, because it is
>> saying the opposite, but that's a detail :) )
>
> What mixup, more specifically? I like perfect changelogs. :-)
The changelog for the last version mentioned this:
* relax binary package network-manager-iwd
to not conflict with wpasupplicant;
update long description;
closes: bug#1094603, thanks to Matthieu Baerts
Did you not remove the entry for the bug#1095606 instead of this one?
Cheers,
Matt
