Notices by 🐧sima🐧 (sima@chaos.social)

Monday after spring break right back getting dragged into (gpu) memory management fun. I guess time to re-drop my tongue-in-cheek take on this:

memory management is trivial, right until you add reclaim

then it becomes impossible

#kernel

extremely simplified, but if you find yourself in a really broken system, you have two options

ragequit and (try to) nuke it from orbit, to build something better

derail, sabotage and maybe improve it from within

you can't do both, because the bridges you need for the second strategy are the same bridges you converted into plasma with the first

and lucky me, there's a lot of things this applies to rn, giving me way too much plausible deniability

how to change the kernel[1]

1. assemble a sufficient coalition of willing fools
2. do it
3. if it works, ask for forgiveness, if it fails, quietly bury it and try the next thing

the more public success you pile up, the easier this gets. but if you fail at step 1, because your ego gets in the way, or you lack the political skills, or you think talking about anything non-technical is verboten, it will be endless amounts of pain and frustration

1: anything you want to change really

#kernel

this is just wild to me

"big subsystem has no new maintainer"

last few times someone stepped down in drm we didn't just look for anyone, but someone who complements the team well. so specifically a new drm maintainer with armsoc or distro background or things like that, to make sure the overall subsystem is well represented

https://lore.kernel.org/all/87wmefguqt.fsf@kernel.org/

#kernel

@karen @conservancy random silly fact, January 1st isn't a random day, it's when the sun starts rising earlier again in the northern hemisphere. which is not the same as the shortest day, because of the ellipsoid path the earth takes around the sun

so quite a thing to celebrate for people living up north who get the big sad when the sun isn't there

hm I thought there was a blog or rant or kernel documentation page about persistent device naming, similar to stable kernel abi nonsense from @gregkh but I can't find it, was I dreaming?

https://www.kernel.org/doc/Documentation/process/stable-api-nonsense.rst

#kernel

@vaurora I'm learning the hard way how much a difference testosterone makes. could have been me pre-transition - well except the flip-flops, I really only wear those in summer, need to look somewhat decent and all that

now I just bought an over knee length down parka because constantly freezing is just not very funny 🥶

reading a cv ...

"Error free coding in C"

👀

that's really not a thing, I'm sorry

@mekkaokereke I threw a big donation at the local trans* support group this morning

they're ramping up a program to support trans/queer immigrants and asylum seekers, so I'm hoping this is the place my money will be most effective

@mekkaokereke I really hope I don't have to read your threads again in 4 years

but realistically I think I need to hope that I read your threads again in 4 years. because the more likely alternative is that there's going to be even less of an election than now or 4 years ago, with even more tricks to make sure only the right (well, white) people can vote

kangrejos in copenhagen was great, but the absolute 🤯 moment was when I learned that GFP_ stands for get_free_pages

I had no idea. and @gregkh next to me had the same https://xkcd.com/1053/ experience

#kernel

@lina @gregkh I just thought it's general protection fault, misspelled, and made no sense. not the first acronym I'll never understand, so just shrugged for the past few decades of looking at kernel code

sometimes I wonder whether people genuinely believe you can be a complete asshole in one place and totally friendly in another, with some of the same people in both places, and then expect that your bad reputation doesn't transfer ...

sometimes I have regrets about drm having so much driver specific uapi

it's worse when the it's driver specific values in a generic api like the drm format specificiers consisting of a fourcc format code and an opaque 64bit layout modifier. yup someone designed a modifier that's not opaque to generic userspace like compositors

#kernel

result of the croissant baking course

#food

@linear I'm on the board of that org, and we shouldn't have this. but it's a bit complicated

1. drm subsystem never enforced the real name policy, and @airlied and me as subsystem maintainers put that on record https://lore.kernel.org/dri-devel/CAKMK7uG+TeATXctJaXBgSRxpinDdtOhGa+o2CMPaPtO1QyHtJA@mail.gmail.com/

it's now also officially fixed for the kernel overall https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=d4563201f33a022fc0353033d9dfeb1606a88330

2. some projects require the https://www.developercertificate.com/ and there's the common misconception that this requires a "real name". it doesn't, and never has

1/2

@linear @airlied

3. if we do have projects that have a real name policy in their docs or require it, we need to fix it because that just flat out violates our CoC and values as an org. I'd much appreciate if you can reach out (in private if you prefer) so we can sort this out

will the real X please stand up, please stand up, please stand up

lol my dear employer's IT departement telling me to order a new laptop because windows 10 is eol and apparently they can't upgrade that stuff ...

meanwhile the linux machines here very much only get retired when the hardware is dead or just too old

once more for the peanut gallery

nvidia _choose_ to not support the wayland stack. or well more generally, the new linux kms+egl gpu stack, since there's a lot other cool things you can built on top of it that aren't wayland

nvidia _choose_ to implement crypto locked down fw in a way that blocks open drivers (even apple got this right!). and no one else can fix it, because it's actual real crypto

nvidia _choose_ to not fix any of this

don't buy nvidia if you don't like this

#kernel