Ah, so it turns out if you just hack the VMMR0_DO_CALL_HYPERVISOR case of the switch statement in vmmR0EntryExWorker to always do nothing and just return VINF_SUCCESS, then you can run purely with REM. I managed to get up to a GUI login in Ubuntu running in this environment.
cheers, Tom On Tue, Oct 27, 2009 at 8:24 PM, Tom Birch <[email protected]> wrote: > BTW my host OS is Mac OS X 10.6, and the host machine is a > Nehalem-based MacPro (i.e. has VT) > > On Tue, Oct 27, 2009 at 8:15 PM, Tom Birch <[email protected]> wrote: >> I'm trying to get VBox running with as little dependency on the kernel >> as possible, so no hardware acceleration, no raw mode, etc. everything >> just running in QEMU. I'm running VBoxSDL because it supports explicit >> arguments, and it seems to work. Running >> >> VirtualBox.app/Contents/MacOS/VBoxSDL --startvm Ubuntu --nohwvirtex >> --norawr3 --norawr0 >> >> will load my Ubuntu VM fairly slowly (definitely slower than if I >> don't specify --nohwvirtex), it gets up to Gnome, and the only calls >> into R0 while it is running are VMMR0_DO_GVMM_SCHED_HALT and >> VMMR0_DO_GVMM_SCHED_WAKE_UP (as shown by instrumenting >> vmmR0EntryExWorker) >> >> The problem is that VMMR0_DO_CALL_HYPERVISOR is called exactly once, >> right after VMMR0_DO_VMMR0_INIT. This goes through all the switcher >> code code that I'm trying to avoid. I don't understand why it's doing >> this, and it seems to only happen once and have everything else >> actually done up in userland by QEMU, any ideas? >> >> cheers, >> Tom >> > _______________________________________________ vbox-dev mailing list [email protected] http://vbox.innotek.de/mailman/listinfo/vbox-dev
