I've now completed the kernel git bisect. git bisect says the commit
which fixed this issue is 29722cd4ef666705b2eda1c3ba44435488e509eb
("x86/asm/entry/64: Save R11 into pt_regs->flags on SYSCALL64
fastpath").This fits in with the discovery on the rr side that the rr commit which started causing the test suite to deadlock the kernel is https://github.com/mozilla/rr/commit/36ac389649e3e84caa46d64487813c018857f831 which is an R11-related change. rr upstream have now committed https://github.com/mozilla/rr/commit/11e67a66a2218effb9381b9ac86a2085b0f5d5d1 which is essentially working around this kernel bug. However, since "unprivileged user can deadlock the kernel" is a local-user DoS exploit I think the fix is still worth backporting. I haven't attempted to analyze the kernel change to determine why the kernel was crashing before or if you could do something more interesting than just crash it by setting r11 to something other than 0 in the userspace code. ** Changed in: linux (Ubuntu) Status: Incomplete => Confirmed -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1535307 Title: running 'rr' make check causes kernel "hard LOCKUP" To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1535307/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
