On 27 February 2017 at 17:49, Thomas Huth <th...@redhat.com> wrote: > On 27.02.2017 14:16, Marcel Apfelbaum wrote: >> Why do we need the "-kernel" parameter for the null-machine? I mean - >> what are the use cases? > > Since it is now possible to instantiate a CPU and memory with the > null-machine, you can use it as a simple instruction set simulator board > now. Start QEMU with something like "-M none -cpu xxx -m 1G -S -s" and > then connect a remote GDB to interact with the emulated CPU. This is > e.g. useful for CPU models that do not have a matching emulated machine > yet. Adding the possibility to use the "-kernel" parameter for this, > too, would make things a little bit more comfortable on certain targets, > since you then don't have to fiddle with the generic-loader device.
But it still won't do the expected magic -kernel does for you on x86 or ARM or..., so it's just confusing to use the same option. We have a command line option for "just load the binary please", and that's the generic-loader. If you think generic-loader's syntax is a bit tedious you could argue for a convenience option that was syntactic sugar for it (in the same way that -hda is sugar for -drive and -device options), but that shouldn't be limited to the null-machine. thanks -- PMM
