On 2011-02-21 13:15, Paolo Bonzini wrote: > On 02/21/2011 01:00 PM, Jan Kiszka wrote: >> qemu_thread_* services do not truly belong here. >> >> What about defining SIG_IPI as QEMU_SIG_KICK which would be provided by >> qemu-thread-posix.h? Then the qemu-thread-posix.c could implement >> qemu_thread_kick() again. > > The function is really specific to VCPU threads. It is simply a wrapper > for the common bit of qemu_kick_vcpu and qemu_mutex_lock_iothread. I'll > rename it to qemu_cpu_kick_thread and make it accept a CPUState.
...or that way. We should just try to respect the abstraction layers. Jan -- Siemens AG, Corporate Technology, CT T DE IT 1 Corporate Competence Center Embedded Linux
