User reports a kernel taint message following updating to kernel 5.5.5. Has anything changed in kernel 5.5 that explains this? This is the message they're seeing: https://ask.fedoraproject.org/t/kernel-tainted-after-running-updates/5487...

My understanding of these messages is: 1. the proprietary nvidia driver is being loaded and used 2. since the module is not signed, the kernel is tainted

But I'm not sure what else I can infer about kernel taint. Are all the other lockdowns still in place? Ideally the user would register a key using mokutil, and sign the module. But if they don't do that, they are still better off than if they disable UEFI Secure Boot to avoid getting the kernel taint message, correct? Unless the user has gone to the trouble of self signing a proprietary

On Thu, Feb 27, 2020 at 5:54 AM Justin Forbes <jforbes(a)redhat.com&gt; wrote: > > No, nothing has changed here, loading a proprietary module has marked the kernel as tainted for a very long time. If you went back to 2.6 kernels, you would see a similar message about the kernel being tainted. The message has expanded a bit over the years as we check for things like module signatures, etc, but the end result is the same the taint flag is P for proprietary module. Gotcha. > Unless the user has gone to the trouble of self signing a proprietary module, and adding that key to the keyring, UEFI secure boot had to be disabled to even load the module. Module signatures are used and checked outside of secure boot as well. Still, even if they do sign the module and add that key to enable the module to work with secure boot, the kernel will be tainted P. Is it technically possible for the Fedora signing key to be used to sign a 3rd party key, thereby allowing the loading of 3rd party modules signed with that 3rd party key? Policy wise, is it likely that could be done? e.g. trusting the RPM Fusion Nvidia and Broadcom kernel modules? This is something that will *not* be done or considered. If those third

ostensibly we want users to leave it enabled. But because self-signing modules is tedious, possibly quite a lot of users are just disabling UEFI Secure Boot. I'm not sure if it's possible to make this work out of the box for users, but it would be nice to not just make it a documentation problem.