With the CoW changes, there are two things that are not working any more:
hyperlight_main tracing - because tracing logic that relied on the fact that before each hlt instruction made by the halt function, the traces were sent to the host.
This does not hold true any more because the hlt instruction is written in assembly after the call to generic_init.dispatch_function - in the future PRs for CoW, most likely, dispatch_function will have the same approach.
We need to consider some changes to accommodate the new approach.
Furthermore, the Info trace for the dispatch_function is not shown because the tracing state is reset right after the call to the function comes from the guest. This produces warnings because the trace event closing the span is sent to the host, but no open span event is sent.
With the CoW changes, there are two things that are not working any more:
hyperlight_maintracing - because tracing logic that relied on the fact that before eachhltinstruction made by thehaltfunction, the traces were sent to the host.This does not hold true any more because the
hltinstruction is written in assembly after the call togeneric_init.dispatch_function- in the future PRs for CoW, most likely,dispatch_functionwill have the same approach.We need to consider some changes to accommodate the new approach.
Furthermore, the
Infotrace for thedispatch_functionis not shown because the tracing state is reset right after the call to the function comes from the guest. This produces warnings because the trace event closing the span is sent to the host, but no open span event is sent.