[SOLAR_FIELDS]

The fact that you had to take a whole paragraph to explain the contortionist arrival at something that isn't even really super clear after you explained it (you kinda pointed the finger both at end users and at distro maintainers simultaneously) and essentially boils down to "well, you as the end user need to be following kernel CVE's and can't trust distro maintainers to do it" does in fact indicate that there is a deeper issue at play here. You might say "well, there's no implicit chain of trust here". You might be right, but is that really the most effective way of doing things? Of course Linux is Use it at your Own Risk, but is there not a concept of "we as a collective community should get together and try not to drop the ball on some serious shit?"

In terms of something actionable, and maybe someone more well versed in how the distros work can tell me why this is a bad idea, but shouldn't there be a documented process and channel for critical CVE's to be bubbled out to distro maintainers who then have some sort of SLA for patching them and sending them downstream to end users? Perhaps incentives are not aligned to produce this outcome.

[akerl_]

To be more blunt: if you’re paying for a product, the vendor owes you whatever things they committed to. If you’re a Redhat customer and your agreed SLA with Redhat for this kind of security fix was passed by, go be mad at Redhat. (I don’t think Redhat is bad here, they’re just the vendor most known for a commercial offering from the lists here. I would say the same thing about Ubuntu Pro)

Otherwise, it’s on the end user. Distro volunteers don’t owe you anything. Kernel devs don’t owe you anything.

I don’t care about what would be the most effective way of doing things. I care about what folks involved actually owe to each other, and distro volunteers don’t owe users any kind of active chasing of remediation due to the user’s threat model.

The idea of making some kind of streamlined process that solves what you didn’t like about this vulnerability’s remediation is that it ignores basically all the complexity. Like “what about distros that don’t abide by embargoes” or “what distros count as ones that matter” or “what about all the vulns that aren’t in Linux, they’re in software that’s packaged across many operating systems”.

[SOLAR_FIELDS]

Right, you’re saying “system is working as designed”, and I’m agreeing, but I’m saying “the system as designed kind of sucks, how can we make it better”?

[akerl_]

I disagree that it sucks. It leverages a ton of people putting in their time and resources, and relies on system operators being active participants.

This vulnerability is, for some threat models, a really big deal. A security group found the vulnerability. They disclosed it. It was patched.

Folks here have gotten all kinds of bent out of shape that the groups involved didnt do things in the way each internet commenter would have liked. But this is the system working.

[i_think_so]

> This vulnerability is, for some threat models, a really big deal.

This vulnerability is, for other threat models, a death sentence.

> A security group found the vulnerability. They disclosed it. It was patched.

It was patched only after some people who should have been notified well in advance happened to notice something was up. That is NOT HOW IT'S SUPPOSED TO WORK.

For as long as the unpatched window remains open, skids will mess around and break things. Organized crime teams will use it for some really nasty hacking/ransomware/exfil/extortion/whatever. I guarantee you, this vuln is powerful and widespread enough that intel orgs will use it to kill targets, if they haven't already been using it for years. And if they have, we can just bank on them pulling out all the stops to take advantage of the remaining time for wreaking havoc. Make a project out of it and see if you can guess some of the future headlines.

Certain folks might not care much because they are citizens of one or more of those orgs' nations, so those targets are welcome to die in their opinion. That's fine. You do you, I'll do me, we'll all just go on doing our thing. But it's all fun and games until the wrong target gets hit and now there's a pact between the Germans and the Austrians being invoked and a few dozen million Europeans die. Or a geopolitical hotspot flares up and overnight 20% of the global petroleum supply chain grinds to a halt. Use your imagination. This vuln is a digital magic wand that is trivially usable to cast Avada Kedavra and somebody neglected to tell 99.99% of the Good Guys about it.

How is this different from any other day? Because now we've got a world-changing vuln out in the wild with no distro mitigation on day 1, and who the hell knows how many unscrupulous actors poised to take advantage of it before the fun and games stops. There will be no adults in the room when the miscreants decide to deploy while they still can.

Is this vuln going to start the next world war? Probably not. I don't expect it to and I hope and pray it doesn't. But leaving a vuln like this undisclosed to the very people whose job it is to protect us all is playing with fire. Not matches; more like a 10-grams-less-than-critical mass of plutonium.

sam is right to be pissed and he's doing a very good job of hiding it, because he knows that his users are at the mercy of TPTB in the Linux kernel world. Somebody's head needs to roll for this, and I don't mean some dude the CIA wants to hax0r because he's next on the list.

[walletdrainer]

> This vuln is a digital magic wand that is trivially usable to cast Avada Kedavra and somebody neglected to tell 99.99% of the Good Guys about it.

A Linux LPE is a nothingburger unless you’re relying on the Linux kernel to enforce internal security boundaries, which would simply be foolish.

[i_think_so]

The PoC exploit code in python (3.10+) fits comfortably in 1k bytes. An unminified version that works for even older versions of python is just a hair under a 1500 byte packet payload, modulo headers for your preferred method of delivery. I can only guess how much it could be shrunk down to only the shellcode.

Now, y'all tell me, since I'm not a web guy. How hard is it going to be to tweak this lovely little pathogen into some kind of browser exploit? It just needs to be combined with a sandbox escape to work on current versions, right? Difficult but quite worth investing the time and effort to develop if that's your line of business. If that happens, every at-risk Tails user is going to have to stay offline for a while, unless they want to play the drone lottery.

Or how about chaining it with any of the as-yet unpatched bugs in gawd-only-knows how many web services out there that have poor input sanitization code? That bug now graduates from a DoS crash causer to a root grab. Good luck stopping it with your fancy AI Behavioral Analysis security tools. They better be fast. The sploit is going to do its work in two packets, maybe three. Fun times.

Lucky for us systems monkeys, it's not like anybody is spending billions of dollars to develop vuln finding AI tools right at this very second. So there shouldn't be many unpatched web services holes.

Oh, wait.

Of course, as the grey hats can already tell you, the really delicious part of this thing is how it's going to become the LPE tool of first resort for any APT that's already inside ur base killin ur doodz.

Nothingburger? This nothingburger is going to root a million OS instances before we know what hit us.

[tptacek]

Start a distro with your preferred upstream tracking policy.

[SOLAR_FIELDS]

Is that the only option here? It’s certainly being framed as such.

[nchmy]

Fwiw, I'm completely with you on this. The folks you're communicating with seem utterly miserable, and don't seem to be communicating in good faith.

Not sure what the solution could/should be, but surely there could be a better, easier mechanism for kernel to advise all distro maintainers who care, and for those distro maintainers to subscribe in some way. Whether any distro maintainers do so (let alone do something about the vuln notifications) would be entirely up to them. There could also be some easier way for end users to see what the distros' policies on this are, such that they can take that into account when selecting a distro.

[akerl_]

It seems odd to call me utterly miserable and then suggest I’m not communicating in good faith.

We don’t have to agree, but the site rules are pretty clear that swipes like that aren’t ok.

That kind of distro maintainers and kernel devs communication path already exists: the linux-distros@ mailing list. But since anybody can read it, posting “hey everybody, this is a security patch” has basically the same effect as the security researcher posting, in terms of disclosing the vuln to bad actors.

Given that anybody can make a Linux distro, and Linux distros aren’t generally either capable or interested in background checking their teams or policing their individual security practice, it doesn’t seem possible to have a communication channel that distros can sign up for that lacks this problem.

[komali2]

Just as a purely intellectual exercise, what changes about this if we leave aside ideas of "owe," "deserve ," and "earn?"

There's not really an enforcement mechanism in FOSS like there is in capitalism world, it just comes down to what we want our part of the world to look like. So I think we'd think more clearly if we leave aside the ideas like "who owes who what." I think it's fun to imagine what sort of motivations and incentives there are if we put away the money ones.

[Brian_K_White]

"leave aside ideas of "owe," "deserve ," and "earn?""

Nonsensical string of words with no meaning.

If you want something that someone else isn't giving you, you have the option to try to do it yourself, or try to compel someone else to give you what you want somehow. Feel free to idk pay someone to track the kernel list and 4000 others and send you heads-ups? Try to pass a law to make people do what you want since you don't care about words like "owe"?

[komali2]

> If you want something that someone else isn't giving you, you have the option to try to do it yourself, or try to compel someone else to give you what you want somehow.

Yes, exactly, the opposite of paying, since when you pay someone something they owe you whatever you paid for.

If we leave aside owe, deserve, and earn, we can start discussing things like what we want our kernel ecosystem to look like, how we can make it safer, etc, without being burdened by these concepts.

It's a simple intellectual exercise, that's all. If you're having a strong reaction to it, imo that'd make it even more fun for you to participate.

[Brian_K_White]

But there was no intellectual excercise. Only a complaint with no proposal.

You want someone to do something for you for some other reason than that they owe you.

They already are doing something for you that they don't owe you. They are writing software that you benefit from. You just want them (or somebody) to do something else that they don't owe you.

They aren't, because they don't owe you and it's not something they want to do for fun, and so since the problem is they don't owe you, you wish to set aside words like "owe".

Well sure. Looks like you found the problem and the solution alright. Why didn't anyone else think of that?

[komali2]

I don't feel like I'm complaining, I feel like I'm asking how else someone would frame it without leaning on the concepts mentioned. What changes about the dynamic then?

[QuiEgo]

Agree on this so hard. Why does everyone expect instant patches and SLA-like infrastructure from unpaid volunteers?

If you want that, buy a commercial distro of linux, or use Windows. That's a huge part of Microsoft's value proposition to enterprise - they pay people to stay on top of security patches for you. Same with RedHat and others.

Expecting anything of unpaid volunteers is unreasonable.

> THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM “AS IS” WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, REPAIR OR CORRECTION.

[hnfong]

> In terms of something actionable, and maybe someone more well versed in how the distros work can tell me why this is a bad idea, but shouldn't there be a documented process and channel for critical CVE's to be bubbled out to distro maintainers who then have some sort of SLA for patching them and sending them downstream to end users? Perhaps incentives are not aligned to produce this outcome.

Who decides who is a trustworthy distro maintainer? In the open source world everyone is equal, no favorites are chosen. If your point is that the distros backed by companies making at least $x million revenue a year should get priority disclosure... pretty sure somebody will take issue with this.

And it's not like a hypothetical issue either. Given the high stakes, bad actors are highly incentivized to masquerade as some small scale niche distro until they get their effectively free zero day CVE.

[SOLAR_FIELDS]

I wanted to share with you that I took some time to reflect on this response and I now agree that this is correct. To ask Linux to add some central authority, no matter how good-intended it is, is antithetical to how Linux works. We must accept that this is a side effect of the purism that requires Linux to function. Can we do better? Probably. But we shouldn’t give up the point of why we are doing this in the first place - and to declare such centralized authority would be doing so, IMO

[PearlRiver]

The real advantage of Microsoft is that there is someone you can sue!

Linux like every open source project is just a bunch of people who are YOLOing it. Not something you use for your fortune 500 critical mission infrastructure.

[anygivnthursday]

https://github.com/microsoft/azurelinux

[sirl1on]

I thought this is why red has exists?

[kjs3]

Only if you are paying them. If you don't have a service contract for RHEL, you have no grounds to sue.