Don't override kernel panic/oom handlers in userspace

[mkroening]

Fixes hermit-os/kernel#43.

std's panic handler in userspace seems to do weird stuff that leads to page faults instead of a proper panic. This fixes the issue by renaming the kernel handler symbols making both coexist.

[stlankes]

Where do you know that this was the reason for hermit-os/kernel#43?

[mkroening]

I experimented with using panic! in CpuFrequency::detect, as described in hermit-os/kernel#43.

Though I am not sure, what x86_64-unknown-hermit's std panic_handler does exactly to provoke the page fault/invalid opcode, making sure the kernel uses its own x86_64-unknown-hermit-kernel no_std panic_handler, does work for CpuFrequency::detect as well as RxQueues::get_next from hermit-os/kernel#209.

It is possible that is has to do with std's panic_handler calling core::intrinsics::abort, which emits ud2 instructions, but that's speculation. The instruction pointer in the invalid opcode exception handler does not make sense, so I can't confirm this.

[stlankes]

bors r+

[bors]

Build succeeded:

• Format check (ubuntu-latest, nightly)
• KVM Tests
• Test (macOS-latest, nightly)
• Test (ubuntu-latest, nightly)
• Test (windows-latest, nightly)