The Linux 5.2+ "Register Corruption" Bug / Golang Issue Was A One-Line Kernel Caching Issue

[hotaru]

This bug is not just triggered by doing just any AVX work.

It requires (from here):

1. A thread doing AVX operations using YMM registers.
2. A signal where the kernel must fault in stack pages to write the signal context.
3. Context switches. Having a single task isn't sufficient.

this is true, but anyone doing the first of those now needs to verify that the other two didn't happen. in practice, this means they have to run the computation again with a fixed kernel.

[pal666]

subj ultimately shows usefulness of go

[CommunityMember]

subj ultimately shows usefulness of go

Actually, what it shows is that (again) Google has the resources to go deep when things go squirrely, at least in code they use and care about. With enough crash reports, all bugs are shallow.

[milkylainen]

we don't actually know that the fallout has been tiny:

Well. I was under the impression that the context was corrupted. Not the actual data itself.
So more likely to cause invalid code and state than data corruption.
A wild guess...

[hotaru]

Well. I was under the impression that the context was corrupted. Not the actual data itself.
So more likely to cause invalid code and state than data corruption.
A wild guess...

nope. if you look at the original Go bug report, it was about corrupted data.