在 2013-05-27一的 11:00 +0200,Gerd Hoffmann写道: > On 05/24/13 04:02, li guang wrote: > > 在 2013-05-24五的 10:24 +0900,Isaku Yamahata写道: > >> On Wed, May 22, 2013 at 01:37:41PM +0800, li guang wrote: > >>> 在 2013-05-22三的 14:28 +0900,Isaku Yamahata写道: > >>>> Why? > >>>> And it breaks pointer operation like > >>> > >>> the fact is I can't guess why gpe->sts is defined uint8_t > >>> but the real hardware is 32-bit width. > >> > >> Which section of ACPI spec? > > > > seems ACPI SPEC does not define the length, > > but, real hardware does, e.g. ICH9. > > Still fail to see what exactly you are trying to fix. > > qemu can continue work with uint8_t internally just fine. memory api > will split dword access by guests into 4 byte accesses. ich9_gpe_ops > looks correct. > > >>> I expand it to 32 because the future usage for me will > >>> access bit beyond 8. > > gpe->sts is an array, so you already have more than 8 bits. > > >> I don't have strong opinion wheter uint8_t or uint32_t. > >> But the current patch is broken as the pointer offset > >> operation needs to be fixed. > > Oh, and it breaks live migration too. >
Yes, I will re-consider this. Thanks!