On Fri, Apr 13, 2012 at 5:59 PM, Paolo Bonzini <pbonz...@redhat.com> wrote: > Il 13/04/2012 11:52, Zhi Yong Wu ha scritto: >>> > FWIW, live snapshot during guest boot (block_set_io_throttle ide0-hd0 0 >>> > 500000 500000 0 500000 500000) died immediately without the patches. >>> > With the patch I could do 20 snapshots before I stopped. >> Why do you not use qemu_aio_flush before loop start? it can fix the >> issue said by me. > > I won't modify a patch without understanding what's going on. From my > understanding (which may well be wrong), replacing qemu_aio_wait() with > qemu_aio_flush() should have no effect in this patch. > > In other words, you didn't provide enough justification to make me > understand why there is a problem and why the change fixes it. Did you > see a problem with all the patches, or with this one in isolation (which only applied this patch. > actually would not compile alone, and this should be a red flag alone)? > How did you reproduce it? During guest with IDE device boot up, at runtime set this device's bps=100. > > Paolo
-- Regards, Zhi Yong Wu
