On 11/21/2012 02:59 AM, Stefan Hajnoczi wrote: > On Tue, Nov 20, 2012 at 01:24:08PM -0500, Don Koch wrote: >> On 11/20/2012 07:31 AM, Stefan Hajnoczi wrote: >>> +void hostmem_init(Hostmem *hostmem) >>> +{ >>> + memset(hostmem, 0, sizeof(*hostmem)); >>> + >>> + hostmem->listener = (MemoryListener){ >>> + .begin = hostmem_listener_nop, >>> + .commit = hostmem_listener_commit, >>> + .region_add = hostmem_listener_region_add, >>> + .region_del = hostmem_listener_section_nop, >>> + .region_nop = hostmem_listener_region_add, >> Really? > Yes. I will rename the function in the next version of the patch to > make it clear this is correct. > > Since we rebuild the entire regions list on every update transaction, we > care about the "nop" (unchanged) regions too. They need to be added to > the new list we're building. > > The deleted regions are ignored because we're building a new regions > list from scratch anyway. > > Stefan Ah, OK. I caught this mail later than my last response (for whatever reason, the last patch you sent went to my inbox, this response got dumped in my qemu folder - go figure).
-d