On 18 October 2011 15:32, Avi Kivity <a...@redhat.com> wrote: > On 10/18/2011 04:06 PM, Avi Kivity wrote: >> > I didn't spot this the first time round, but this is wrong. >> > We shouldn't be g_malloc0()ing the MemoryRegion -- it should >> > be an element in some suitable device struct. >> > >> > I think the right thing to do here is probably first to do the >> > (fairly trivial) conversion of the icp_control code to be a >> > sysbus device, then do the memory region conversion on top of that. >> >> It's not any less wrong than the original code, which also leaked >> memory. I'll merge it and let any further conversion take place on top. > > I meant, it's not any more wrong than the original code, though my > original statement is also correct.
Either way, if you're happy to take it into your tree that's fine with me :-) -- PMM