Re: [hatari-devel] [REGRESSION] Access to non-existant IDE slave crashes Hatari

[Christian Zietz <czietz@xxxxxxx>]

Thomas Huth schrieb:

> I gave this a very quick try, but I was not able to reproduce the
> crash. Which machine type did you use? Did you use the GUI or the CLI
> to configure the IDE drive? Fast-boot enabled or disabled? Please
> provide exact steps how you ran Hatari.

Please find attached my configuration file and everything else that is
needed to reproduce the crash. As mentioned in my previous post, I was
using Hatari as of commit #1797e220:
http://antarctica.no/~hatari/hatari-1797e220/windows/

I still believe that any reports of critical regressions (and crashing
Hatari from a TOS program imho is such) should be investigated by more
than "a very quick try".

Further debugging Hatari's code shows the precise crash location to be
in ide_data_readw(), caused by the fact that s->data_ptr is NULL for the
(non-existant) slave device.
https://git.tuxfamily.org/hatari/hatari.git/tree/src/ide.c#n2457

> Anyway, thanks for providing the crash_hatari.S source code. Looking at
> that file, I wonder whether there is another bug in Hatari: You don't
> switch to supervisor mode before accessing the registers ... Is it
> possible to access the IDE registers in user mode on a real Falcon? I
> don't believe so, but could somebody who has a Falcon easily
> available please check this, just to be sure?

At least a lot of third-party IDE interfaces allow register access in
user mode, as well. Furthermore, the STBook was the other Atari with
built-in IDE. It becomes apparent from the STBook schematic that IDE
register access is possible there in either supervisor or user mode.
Hence, I wouldn't be surprised if Atari kept it that way for the Falcon.

Regards
Christian
--
Christian Zietz  -  CHZ-Soft  -  czietz@xxxxxxx
WWW: https://www.chzsoft.de/
PGP/GnuPG-Key-ID: 0x52CB97F66DA025CA / 0x6DA025CA

Attachment: hatari-crash.7z
Description: Binary data