openvmm
openvmm copied to clipboard
mesh: remove global timeout thread
For convenience, mesh cancel contexts are timed out via a global timeout thread. Find a better way to manage this, e.g.:
- Pass a
pal_asyncdriver tountil_cancelled. This isn't great because we want to avoid apal_asyncdependency in the public API. - Require the platform to be initialized with an explicit driver for handling timeouts. This works OK but means every program using mesh has to remember to do this in
main. - Grab the driver from the current task. This seems easy to screw up since we're not always running on a task.
None of these solutions is obviously the right one. But it's awkward to have this extra thread.