Peter Memishian wrote: > > It's possible. Unfortunately, I'm not sure how to verify that since the > > second fault blows away any state related to why the panic occurred in > > the first place. > > I've found debugging stack overflows easier on SPARC. Does it reproduce > on SPARC? >
I've found the same, but in this case, the double-page-fault only exhibited itself on x86 (note the past tense). :-) Phil fixed this is his putback this morning. It turns out that I accidentally merged-out a change he'd made to os/startup.c to explicitly modload() the "dev" module in startup_modules(). Phil can probably give you details regarding why this omission leads to a double-page-fault (and why the "dev" module needs to be loaded in startup_modules()), but he noticed that this was missing from the Clearview gate after we did an IP Observability code walkthrough yesterday. -Seb
