Specifically, the weak spot is in “maple tree,” a new data structure system for VMAs introduced in Linux kernel 6.1 that replaced the “red-black trees” and relied on the read-copy-update (RCU) mechanism.
Maple Tree also recently caused intermittent failures in some of my CPU-intensive tasks, in such an obscure way that I only found out by dumb luck that it was a kernel bug. I expect it will be great eventually, but it’s feeling pretty rough at the moment. I’m thinking this code should have had more testing and maturing before going mainline.
Damn. If the Maple Tree code is bugging out under CPU-intensive tasks, that would explain a lot about how my system’s been behaving since I moved to 6.1. Thanks for the heads-up, and I guess I should compile another new kernel.
Yep, got failures building GrapheneOS and the devs of that ROM made a big fuss on their Twitter when they encountered the failure themselves. The kernel devs really messed up with the way they deployed this thing
Maple Tree also recently caused intermittent failures in some of my CPU-intensive tasks, in such an obscure way that I only found out by dumb luck that it was a kernel bug. I expect it will be great eventually, but it’s feeling pretty rough at the moment. I’m thinking this code should have had more testing and maturing before going mainline.
Damn. If the Maple Tree code is bugging out under CPU-intensive tasks, that would explain a lot about how my system’s been behaving since I moved to 6.1. Thanks for the heads-up, and I guess I should compile another new kernel.
Rcu is mostly broken, it’s been a nightmare for a decade, building on top of that seems suicidal.
I know rcu failures are just symptoms of other issues, but building on top of it doesn’t help matters.
Yep, got failures building GrapheneOS and the devs of that ROM made a big fuss on their Twitter when they encountered the failure themselves. The kernel devs really messed up with the way they deployed this thing