KVM General

[1] Marc: In PL1, an UNDEFINED instruction can trap even if its condition code is false. See the SWAP emulation code in the kernel, for example. This is because the UNDEF is generated at the decoder level.

[2] Marc: Moving to "Boot in HYP mode", and remove the old booting convention that uses the secure mode.

[3] dmart: The final state of my work is in the arm/hyp-entry* branches here: http://git.linaro.org/gitweb?p=people/dmart/linux-3-arm.git;a=summary. The zImage support wasn't finished, and the patches are not really ready to upstream. I think that the core support in arm/hyp-entry ought to be simplified to a simpler stub that does nothing except return to the caller with the mode changed to Hyp.

[4] Wont fix

[5] Won't fix as covered on the list archives and technically obvious

Work items:
[marc-zyngier] Support entering the kernel in HYP mode[3]: DONE
Injecting undefined instruction into the guest: DONE
[rusty-rustcorp] Handle false conditional traps[1]: DONE
[rusty-rustcorp] Handle undefined instructions properly in the main kernel[4]: DONE
[marc-zyngier] Remove obsolete SMC support[2]: DONE
Remove obsolete SMC support from the boot wrapper: DONE
Document reasons for not running the host in hyp mode[5]: DONE
[cdall] Host Thumb-2 MMIO support @ 2012-07-02: DONE
Guest LPAE support: DONE