On Mon, Sep 19, 2011 at 12:54 PM, Frediano Ziglio <fredd...@gmail.com> wrote: > Now that iothread is always compiled sending a signal seems only an > additional step. This patch also avoid writing to two pipe (one from signal > and one in qemu_service_io). > > Tested and works correctly with KVM enabled. Performances are only sligthly > better (as I expected). strace output is more readable. > > Doubts: > - any sense having two patches and not only last one?
Personally I would just send one here. > - is ok if KVM is disabled? more testing required If I understand correctly it used to interrupt guest code execution whereas now it simply invoked a callback in the I/O thread. Normally this callback will raise a guest interrupt and get the vCPU to notice that I/O has completed. So I think this should work. Stefan
