Comment 19 for bug 1735546

@mhcerri
Thanks for the explanation! I built the kernels directly from the git repo and I didn't use the binary packages (I thought they should be the same). What's your repro-rate? Yesterday as Chris tested linux-azure-edge 4.13.0-1004.4 + my pull request, the repro rate was only ~1%, meaning we have to do "unload/reload netvsc" several hundred times to repro the hang or the calltrace. Previously without the pull request, Chris mentioned it was much easier to repro the issue. So it looks to me the bug is timing-sensitive, and somehow my pull request made it difficult to repro it.