Re: [hatari-devel] memory setup segfault

[Laurent Sallafranque <laurent.sallafranque@xxxxxxx>]

Hi Nicolas,


That's a great job, my Falcon now reboots into STf mode without core dump.
Thanks a lot, it's really appreciable.
I'll do more tests for TT, STe, ... but I'm confident in your patch ;)

Last remark, but probably not very important, as I mentioned it in a 
previous mail, when I boot in falcon mode, I don't have the memory check 
of the TOS.
Normally, the TOS displays the amount of memory, and then, it tests the RAM.
I don't have this since hatari v2.0 or close version.

Regards
Laurent



Le 09/11/2023 à 22:36, Nicolas Pomarède a écrit :
> Le 07/11/2023 à 23:14, Nicolas Pomarède a écrit :
> > Le 06/11/2023 à 20:30, Thomas Huth a écrit :
> >
> > > I can reproduce the crash, but only with --enable-debug.
> > >
> > > I can also reproduce it by starting without hatari.cfg with:
> > >
> > >   hatari --machine ste --tos tos206fr.rom --memsize 14
> > >
> > > Then go to the GUI, select 1 MB and return from the GUI.
> >
> > Thanks, I can get the crash too with this scenario.
> >
> > This is similar to previous crash when going from falcon to st with 
> > >= 4MB , MMU conf at ff8001 will store 0xf (invalid size for both 
> > memory banks) and MMU_Bank0_Size will be set to 0 at start, but will 
> > also keeps its 0 value after reset when calling memory_init with <= 4 
> > MB, which calls memory_map_Standard_RAM and trigger the crash
> >
> > I will look for a correct place to ensure MMU_Bank0_Size has a 
> > correct value in these cases.
>
> Patch pushed.
>
> STMemory_Reset() will now be called earlier during Reset_ST(), which 
> ensures the MMU conf used in STF/STE/TT at FF8001 will have a 0 value 
> at start and MMU_BankX_Size will be != 0 too (banks' size could be 0 
> when using more than 4MB)
>
> I tried a few combinations between Falcon, ST, 8MB, 1MB and I don't 
> get a crach anymore.
>
> Nicolas