On Thu, Jun 14, 2018 at 09:54:41AM +0200, BALATON Zoltan wrote: > On Thu, 14 Jun 2018, David Gibson wrote: > > On Wed, Jun 13, 2018 at 04:13:57PM +0200, BALATON Zoltan wrote: > > > I don't see the problem. The addr register selects the register to read or > > > write. It is set by the first write when the device is accessed the first > > > time (this is denoted by addr == -1 (or really any negative value). The > > > device has 20 registers and trying to read any register outside addr > > > between > > > 0-19 will result in returning 0 and logging a warning about invalid > > > register > > > in m41t80_recv. What could fail here when guest sends garbage? It will set > > > addr to invalid value and try to read non-exitent register and get an > > > error > > > just like for any other nonexistent value of addr (or start to read from > > > register 0 if it manages to set a negative value). All writes of registers > > > are ignored currently (except setting addr by the first write). What > > > should > > > be enforced here? > > > > Ah, I see your point. I mean strictly we should match the hardware > > behaviour if you write garbage addresses here, but really I don't > > think it matters much. > > Problem is like usual I have no idea what the real hardware does. I've only > seen the datasheet, never seen real device and have no way to test. All the > clients I've tested seem to be OK with the current emulation so unless > someone has more info on how this should work I think we can live with this > version and then fix it later if found to be needed.
Ok, fair enough.
--
David Gibson | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_
| _way_ _around_!
http://www.ozlabs.org/~dgibson
signature.asc
Description: PGP signature
