On 15 November 2012 18:42, Eduardo Habkost <ehabk...@redhat.com> wrote: > DeviceState CPUs have to be reset too, and DeviceState uses the > reset-handler system to make sure DeviceState objects are reset, so it > won't be softmmu-specific. > > An alternative is to make empty stubs for qemu_[un]register_reset(), and > not move the reset-handler system to *-user. But somehow I feel better > having a working qemu_register_reset() function (and then adding a > qemu_devices_reset() call to *-user) than having a fake > qemu_register_reset() function and having to use a different API to > reset the CPUs on on *-user.
cf the pretty nasty stuff in linux-user/main.c which currently calls cpu_reset() but only for some targets. -- PMM
