This is before exploits routinely started getting catchy names and even before MySpace was a thing. Let alone Twitter and Facebook was around to share news of the impending CPU apocalypse
Too bad so few people listened
@cypnk Wait, does that coincide with the consumerization of IT?!?
And the willful "just do it!" that started coming from the C-Level offices around the same time?
@cypnk how do you know it wasn't a different bug? The Intel pdf detailing the errata they're talking about is a 404...
@cypnk did OpenBSD already clear the page tables when context switching before the recent disclosures?
@hisham_hm That I'm not sure, but I think their implementation of pmap is such that exploiting is not possible. I could be wrong though
"writing to the %cr3 will flush the entire TLB"
It's also important to keep in mind that OpenBSD implements this and many other features that make the first few stages of an exploit extremely difficult to impossible. Of course, this also affects performance to some degree, which is a reason why other OS vendors haven't implemented it
@clacke Oh gosh! This is has been a problem ever since Theo left NetBSD
But let's not forget that OS vendors do routinely put performance over security. Most other vendors also have to worry about breaking software, which OpenBSD doesn't really concern itself with (but they do help port authors to migrate their code to be more secure)
Their approach is "break before compromise". I don't know of any other OS vendor who does this
@cypnk @clacke I am a NetBSD guy, but I have *always* strongly admired OpenBSD's unwavering "no fucks given" attitude toward doing (what they consider) the right thing. Not only are they happy to break/miss out on things, they also take on a *tonne* of extra work (e.g. developing their own implementations of CVS, NTP and now even SSL) in order to avoid compromising their principles. No other OS project comes close to that dedication.
@solderpunk @clacke And it's a huge drain on resources to do that too. But obviously they have very different goals. I remember Theo describing OpenBSD as a "research operating system", which I think is the more accurate description
I love NetBSD too in that you can pretty much run it on anything with a CPU
@cypnk That is a true and valuable property of NetBSD (which will, e.g. run on the RaspberryPi 1 unlike OpenBSD or Minix 3), but I do sometimes regret that there is such a strong and singular association between the project and portability. It makes people think that there is no good reason to run NetBSD on x86/amd64, which is not so. I love it just as much for its minimalism and cleanliness (although that is much worse than it used to be when I first became an enthusiast), for pkgsrc and more
@solderpunk That is a shame. And its userland is very good too, which is why I think Minix is now using it as theirs
All projects undergo increases in complexity. Such is life :)
@solderpunk @cypnk @clacke
I've also seen them make pains to keep their code pretty portable. I watched an exchange with an Android developer (as in developer of the Android OS) where has asked about a change so they could keep continue using some code in their libc. OpenBSD actually broke it out into a separate file to make that easier for them.
OpenBSD has a "no fucks given" reputation, but they want things to be better for everyone.
@clacke Oh, and I remember Linus yelling at some developer (not for the first time), "WE DO NOT BREAK USERSPACE!" 😂
OTOH, OpenBSD breaks software all the time in the name of security. For other OS vendors, this approach really isn't possible, or at least practical. So I don't blame them entirely for not being able to make these changes
@cypnk Sorry to nitpick, it's definitely interesting. But MySpace has very much already been a (very popular) thing in 2007. Facebook and Twitter might not have been that well known yet but where both operational already.
@syp I didn't include it in the original post because it's a bit more technical, but here's the sourcecode for their implementation of pmap
(Server seems to be having issues right now)
These fixes were taking place since the early 2000s
@cypnk Thanks, will take a look when the server is up again. But I did not even mean to doubt that OpenBSDs/deraadts early skepticism about those features was right.;)
Would you be kind enough to point where exactly parallel branch vulnerabilities are discussed in this documents?
I'm not a cpu specialist and they are hard to read...
@LienRag This was more a highlight of the fact hat OpenBSD devs were aware of the potential problems these early architectural trends would have in (then) current and future software
I *think* this is part of those mitigations:
(It's quite long, so search for "TLB"). I'm also not an expert so I could be wrong
@cypnk If that's not an indictment of their collective inability to communicate with the outside world, I don't know what is. 😜
@paco This is a *very* old problem with the team
But to be fair, a lot of these mitigations do introduce a pretty serious performance hit. Not to mention it breaks pretty much all userland software so even if other OS vendors agreed on the problem, they probably wouldn't have fixed it the same way
@cypnk I'm pretty sure this has been exploited for 15+ years...
I don't for a second think that this big of a hole hasn't been used before.
@thegibson I'd be shocked if this wasn't in the catalog of a at least one 3 letter agency already
Yeah, the cat-and-mouse game isn't practical or feasible going forward. The only safe procedure from here on is proactive auditing, but I don't know how many OS vendors will expend the energy or budget (or have either) to do so
Yeah read that too. eleven years is a damn big time span in computing.
I was planning to buy a new CPU within the next two years, but now it seems you can only chose between the big security hole (Intel) or the smaller one (AMD).
But from what I've been reading a good chunk of the performance lead thad Intel exploited the hole market on, came from these security holes.
@thegibson I'm actually kinda glad this Meltdown/Spectre debacle happened
On the surface, it's terrible, but it's exactly the kick in the pants we all needed. The Snowden revelations didn't really do much. Ransomware annoyed us, but it only got a few flunkies fired. Data leaks are *yawn*
Now we have a situation where whole businesses are going to be running on CPUs as slow as molasses and the productivity hit will hopefully be the final wakeup call about dangers of "first to market"
@cypnk They will not change... those at the decision making level won't see the problem..."cost of doing business.
The CIO/CISO always has the junior seat at the conference table.
a bit like in the old back issues of Popular Mechanics where they're talking about how homes might be underwater in the far off year of 2000
though isn't "being extremely paranoid about everything", like, openbsd's policy on everything?
after all, even a stopped clock shows the correct time twice a day.
@devurandom Let's just say extremely paranoid about most things ;)
I think the security aspect gets a lot of the exposure, but really, they're just trying to write code in a clear and coherent way. "Quick hacks" get discarded in favor of maintainable code, The result is you get secure code as a nice bonus
That's not unique to OpenBSD, or even OS design. Lots of other software projects adopt the same principles
Server run by the main developers of the project It is not focused on any particular niche interest - everyone is welcome as long as you follow our code of conduct!