qyliss changed the topic of #spectrum to: A compartmentalized operating system | https://spectrum-os.org/ | Logs: https://logs.spectrum-os.org/spectrum/
cole-h has quit [Ping timeout: 246 seconds]
Advacar has quit [Ping timeout: 240 seconds]
multi has quit [Ping timeout: 276 seconds]
multi has joined #spectrum
multi has quit [Ping timeout: 250 seconds]
multi has joined #spectrum
cole-h has joined #spectrum
cole-h has quit [Ping timeout: 252 seconds]
Advacar has joined #spectrum
TimFroehlich has joined #spectrum
Advacar has quit [Ping timeout: 240 seconds]
cole-h has joined #spectrum
<samueldr> oh, 2014
<samueldr> what are the chances that this is still the case?
<qyliss> not sure
<qyliss> samueldr: do you think it would be sensible to suggest re-enabling it if other distros have done so?
<samueldr> you have the answer in the back half of your sentence
<qyliss> alrighty :)
<samueldr> kernel configuration is a good kind of cargo cult driven development
<samueldr> you still have to question the decisions, somewhat
<qyliss> there should be a website where I can browse kernel config options and see a table of distros that set or don't set each one
<samueldr> but if everyone does it, generally it means it's safe/good
<samueldr> qyliss: agreed
<samueldr> so
<samueldr> where does fedora keep their kernel config? ;)
<samueldr> I learned recently that many distros (fedora was observed) don't expose /proc/config[.gz]!
<qyliss> yeah, they put it in /boot instead for some reason
<samueldr> on an installed system
<samueldr> not on their installation iso!
<qyliss> ah wonderful
<samueldr> that's the kind of collaboration I'd like to see happen cross-distros
<samueldr> well, not decision per se
<samueldr> ah, duh, I corrected a sentence I never typed out
<samueldr> it would be interesting to have a table laid out, with reasoning where it matters
<qyliss> # CONFIG_INTEL_IOMMU_DEFAULT_ON is not set
<qyliss> aww
<samueldr> that's where having the reasoning explained would be helpful :)
<samueldr> oh
<samueldr> >> Selecting this option will enable a DMAR device at boot time if one is found. If this option is not selected, DMAR support can be enabled by passing intel_iommu=on to the kernel.
<qyliss> wow, according to repology we're the only ones to package rc kernels
<qyliss> the upstream default is Y IIRC
<samueldr> is there an issue with passing the intel_iommu=on flag?
<samueldr> I should have looked at what the option did first :)
<samueldr> I thought it served to enable the IOMMU feature
<samueldr> (or _a_ feature of)
<qyliss> it's slightly annoying, and also means users don't get the security properties of IOMMU by default
<samueldr> hmmm, but since you say upstream defaults to =y it just blurs things even further!
<qyliss> (ngl though my main motivation here is the "slightly annoying" part0
<qyliss> def_bool y
<samueldr> you fell for it!
<samueldr> *that* is the upstream defconfig
<samueldr> sometimes it doesn't agree with defaults in their Kconfig
<qyliss> ugh
<qyliss> multi, nicoo: either of you know where online the Debian kernel config lives?
<samueldr> qyliss: so I guess kernel.org is +1 -1 :)
<qyliss> looks like changing the default isn't the right thing to do, then
<qyliss> but I'll keep thinking about the comparison site idea -- feels like it would be very useful and not too difficult
<samueldr> yeah, while it would be better to have more insight from other distros, at a glance from the defconfig, archlinux and fedora, sounds like it's not desirable
<samueldr> the most difficult part is finding the source of the config files :)
<qyliss> i did find a debian bug report where they enabled it
<samueldr> though maybe extracting from a kernel binary is the easiest universal way
<samueldr> since NixOS *also* does not ship a .config file
<qyliss> was thinking about that, but we also don't have a stable URL where one can download the latest kernel binary
<samueldr> would be interesting, too, to include two columns from kernel.org: "no defconfig" and "defconfig"
<qyliss> yeah
<qyliss> Debian enabled it in 2019, no idea if it's still enabled
<samueldr> yeah, there won't be a stable url for any distros, but you could go from the hydra path in one hope to get the nix store path
<samueldr> one hop*
<samueldr> with the "last build from a finished eval" link
<qyliss> shame repology doesn't do binary package URLs
<qyliss> (to my knowledge)
<qyliss> nicoo: thank you!
<nicoo> You are welcome
<qyliss> hmm, it doesn't even mention that option
<samueldr> from a bystander point of view, the only thing lacking from salsa is a curated list of repos that are the source of packages
<samueldr> qyliss: probably means it ends up being the default
<qyliss> or it lives in one of the arch-specific subdirectories
<samueldr> either that or there's... yeah that
<samueldr> *and* could also be catting at the end of the kernel.org defconfig
<samueldr> which could muddy up the waters for what "defaults" means
<samueldr> they don't have the option because they don't have the option
<qyliss> oh gosh yeah that'd be another thing I'd have to deal with
<samueldr> qyliss: I have this terrible "normalization" tool for Mobile NixOS https://github.com/NixOS/mobile-nixos/blob/master/bin/kernel-normalize-config
<samueldr> where it takes the kernel config from the repo, passes it through the menuconfig, and spits back out the config
<samueldr> but even that is not complete
<samueldr> you'd need to build the kernel config like the distro does it
<samueldr> since some could realistically be catting segments together
<samueldr> e.g. chromeos does that
<qyliss> yeah
<samueldr> extracting from the kernel binary is probably the only way to do it safely
<qyliss> not if they don't include the config in the binary
<samueldr> yep
<qyliss> i.e. fedora and debian
<samueldr> ah, lacking /proc/config[.gz] does not mean IKCONFIG is disabled entirely!
<qyliss> oh I see
<samueldr> IIRC with fedora I extracted the config using binwalk
<samueldr> (there's also a tool that serves specifically to extract the config from binary kernels, better than binwalking it)
<qyliss> is that scripts/extract-ikconfig?
<samueldr> most probably!
<qyliss> i've used that before when I didn't know about /proc/config.gz
<samueldr> heh
MichaelRaskin has joined #spectrum
raboof has quit [Ping timeout: 260 seconds]
acertain has quit [Read error: Connection reset by peer]
acertain has joined #spectrum
raboof has joined #spectrum
multi is now known as singlet
singlet is now known as multi