Re: [hatari-devel] Disk driver memory access checks

[Eero Tamminen <oak@xxxxxxxxxxxxxx>]

Hi,

On 10/13/18 9:09 PM, Nicolas Pomarède wrote:
> Le 13/10/2018 à 18:53, Uwe Seimet a écrit :
> > > if transfer is made using DMA, does DMA generate bus error when trying
> > > to write to TOS ROM or is this not handled at all and "ignored"
> > > (contrary to what would happen if the cpu try to write to ROM)
> > >
> > > One could test this on real HW by just using xbios to read a floppy
> > > sector into ROM.
> >
> > Assuming that the DMA chip generates a bus error in such a case, what's
> > the most similar approach in the case of nf_scsidrv? Just returning an
> > error code as a result of the ST memory check probably not.
> > Of course, I'm aware that this is a rather esoteric problem.

The main thing is to protect the emulation process i.e. prevent
Hatari itself doing invalid memory access.

What else driver does in such a corner case is less of a concern
from my point of view.  For now, you could log an error and
later once it's checked on real HW, add proper error response.


	- Eero

> I would need to check in the code how to raise a bus error from outside 
> the cpu emulation ; in the meantime, the safest is just to ignore writes 
> to ROM (ie : considerer all the data were transfered but ignore the writes)
> A you say, it's a corner case and we wouyld need to check on real HW 
> first the behaviour for DMA FDC and for other DMA in TT/Falcon