Any risk installing from Debian backports if it requires new kernel?

Hi, guys

I was trying to install wireguard (to use as a client) on my PC running PureOS (kernel 4.19.0-13-amd64 #1 SMP Debian 4.19.160-2 (2020-11-28)).

I enabled buster-backports (those are the correct ones, aren’t they?) (only main) and when I went to install wireguard running apt -t buster-backports install wireguard I was prompted with

The following additional packages will be installed:
  linux-image-5.9.0-0.bpo.2-amd64 linux-image-amd64 wireguard-tools wireless-regdb
Suggested packages:
  linux-doc-5.9 debian-kernel-handbook openresolv | resolvconf
The following NEW packages will be installed:
  linux-image-5.9.0-0.bpo.2-amd64 wireguard wireguard-tools
The following packages will be upgraded:
  linux-image-amd64 wireless-regdb
2 upgraded, 3 newly installed, 0 to remove and 180 not upgraded.
Need to get 50,8 MB of archives.
After this operation, 284 MB of additional disk space will be used.

Do you see any risk in taking that change? Is that compatible with the free kernel in PureOS?

I don’t really have those concepts clear. I appreciate your help! (I wouldn’t have a problem with 5.9 kernel on itself, I run that a few months on my very old core2duo macbook from 2007 on Arch, now running 5.10, without any problems. This PC should be even more stable. But I don’t really understand if it would “work” fine with all the “free” parts in PureOS. Thanks in advance!)


Hm, I’m betting PureOS still uses ext4… My usual advice would be “take a BTRFS snapshot and see”.

It’s probably fine. The kernel team is very good at not breaking userspace. If it fails, you can likely roll back without significant trouble. It might include some not-FSF-free bits of firmware, if debian uses a single package for the kernel and firmware, but it will be DFSG Free, which is in some ways more rigorous than the FSF guide (and in some ways less).

Edit: As should be obvious, make sure you have any important data backed up (offsite!) first, and don’t do it when you have a deadline relying on the machine-in-question pending.