> From: Paolo Bonzini [mailto:paolo.bonz...@gmail.com] On Behalf Of Paolo > Bonzini > Sent: Monday, 7 December 2015 01:53 > On 07/12/2015 06:20, Andrew Baumann wrote: > > Yeah, I considered doing that, but figured that those cases > > (incorrectly-sized register writes in 16-bit mode) are indicative of > > a pretty badly screwed-up guest, and was going for a minimal patch. > > It probably makes sense to change them for consistency, though. > > I think those should be fixed by modifying lan9118_*_mem_ops and adding > .valid.{min,max}_access_size. Not for 2.5, however. (Probably these > patches should also be 2.6 + qemu-stable rather than 2.5).
Just to clarify: would you guys like me to prepare such a patch? I'm not familiar with the memory op APIs, and don't have a good setup for testing this device emulation any more (and certainly not in 16-bit mode!), so would prefer to defer to someone else. BTW, I also see no great urgency for these patches. They're minor fixes, and it would be good to have them in, but it's certainly not a regression as the code has been that way for ages. Andrew