On Fri, Nov 22, 2013 at 06:17:51PM +0000, Peter Maydell wrote: > On 22 November 2013 18:11, Christoffer Dall <christoffer.dall@xxxxxxxxxx> wrote: > > On Fri, Nov 22, 2013 at 05:17:14PM +0000, Peter Maydell wrote: > >> + * Emulate a virtual board which works by passing Linux all the information > >> + * it needs about what devices are present via the device tree. > >> + * There are some restrictions about what we can do here: > >> + * + we can only present devices whose Linux drivers will work based > >> + * purely on the device tree with no platform data at all > >> + * + we want to present a very stripped-down minimalist platform, > >> + * both because this reduces the security attack surface from the guest > >> + * and also because it reduces our exposure to being broken when > >> + * the kernel updates its device tree bindings and requires further > >> + * information in a device binding that we aren't providing. > >> + * This is essentially the same approach kvmtool uses. > > > > nit: questionable value of this last line in the comment. > > Well, it says we're not doing something completely bonkers > of our own devising ;-) > ;) > > + vbi->bootinfo.board_id = -1; > > > > board_id = -1 ? > > This is what the kernel's boot protocol says you should > pass as the board ID register for boards which are new > enough that they only support booting via device tree > and not the old-style board ID number system. > ok, thanks. -Christoffer _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
