Baptiste BEAUPLAT wrote:
> Systemd, in buster, will ship the file /lib/modprobe.d/systemd.conf,
> which contains the following:
>
> options bonding max_bonds=0
>
> # Do the same for dummy0.
>
> options dummy numdummies=0
>
> This breaks any configuration that an administrator could have added to
> /etc/modprobe.d regarding the dummy and bonding modules.
We need more information about why an administrator might have done
this, since otherwise for a start it's impossible to guess what would
go wrong as a result. VMs with sabotaged networking, or what? Is
there some other bugreport where we could read about these symptoms?
> For instance, a file in /etc/modprobe.d/dummy.conf containing:
>
> options dummy numdummies=1
>
> Will result in the following being executed by modprobe:
>
> insmod /lib/modules/4.19.0-5-amd64/kernel/drivers/net/dummy.ko
> numdummies=1 numdummies=0
>
> And the original configuration will be overridden.
>
> The only way to force modprobe to use local configuration is to rename
> the /etc/modprobe.d/dummy.conf file to /etc/modprobe.d/systemd.conf.
This overrides the /lib/modprobe.d/systemd.conf entirely, doesn't it?
In which case you'd lose any other things in that file that systemd
might be depending on. Checking on a Buster box I see it also defines
"options bonding max_bonds=0" - if I want to avoid overriding that,
would it be better to use a name like /etc/modprobe.d/zz-local.conf?
> I thinks this should be documented in the release notes as admins would
> need to be aware of that.
And/or quite possibly about max_bonds=0? I'm afraid I know even less
about what symptoms that might have.
--
JBR with qualifications in linguistics, experience as a Debian
sysadmin, and probably no clue about this particular package
