Microsoft now blaming the EU for the crowdstrike issue, because the EU made them open their APIs to third party developers in 2009 is hilarious.
If your APIs had any kind of gracious error handling, this wouldn’t be an issue. Fix your bad code, Microsoft, stop pointing fingers.
@thelinuxEXP This is not how any of this works...
> If your APIs had any kind of gracious error handling, this wouldn’t be an issue. Fix your bad code, Microsoft, stop pointing fingers.
Memory access doesn't go through APIs. As such, there's no "error handling" that could fix it on Microsoft side. If memory is sufficiently corrupted, the kernel itself cannot do shit. The only thing they can do is generic handlers for invalid memory access CPU exceptions -- which is exactly what they did.
While yes, the CrowdStrike Sensor Linux version would not have this issue, it's *not* due to better error handling in Linux. It's due to Sensor using eBPF on Linux as of recent times: Linux already supplies user-space APIs to do monitoring that Sensor can just consume instead of implementing the functionality itself.
On Windows Sensor has to implement it by itself (which, yes, *that* you can blame on Microsoft!). OTOH, Windows does by now also support eBPF (so in theory CrowdStrike could port that to Windows), *but* CrowdStrike Sensor on Windows does more than the Linux version AFAIK. Meaning, an equivalent implementation on Linux could have just as bad of an impact.
... error handling of memory corruptions through kernel modules, and I thought I've seen all insane takes on this thing.
@ljrk Thanks for the condescending tone, it’s always appreciated.
Obviously, that’s not what I meant. What I meant is “why the gell do you leave your OS in a state where it can’t get out of a BSOD without manual intervention”.
Of course, to know what I meant, it would have required reading a bit of the conversation underneath this message, instead of assuming things ;)
@thelinuxEXP Well, you're the one who was shitting at the Microsoft devs. And I'm always for shitting on Microsoft, but that post was just absurd.
And I'm tired of people tooting bad takes that pretty clearly spell out what they meant and then claiming "oh, I meant something completely different from what I wrote that in no way relates to the original text".
But whatever, even the new take is absurd, especially since Linux has the same problem: You can trivially get Linux into a state where it panics and requires manual intervention.
And yes, I'm unfriendly at people who think they can shit on devs who work on a friggin' hard topic like kernel development w/o having a damn clue about the topic at hand. Please, take an os dev kernel class and bang your head against walls before writing such a bullshit post.