[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: memory protection



Hi!

On Wed, Nov 19, 1997 at 12:44:46PM +0100, you wrote:
> |> Except Thing IS supposed to be MiNT-friendly. N.AES too. 
> |> Thing cannot even be started by N.AES correctly.
> Then why does it access memory it doesn't own?  That's still a bug.  If
> it's not bug-free (which will of course never happen for any non-trivial
> program) it should catch SIGSEGV and properly clean up in the handler.
Well, a common problem with memory protection under GEM is the AV 
protocol. For example, ST-Guide ships with standard program flags (i.e. 
full memory protection), but never uses memory protection flags for the 
blocks it sends via AV protocol. Thus, Thing is killed as soon as it 
receives ST-Guide's AV_PROTOKOLL-message, as the pointer to the 
applications's AES-name points to private memory.

And (maybe in connection with NVDI) after a GEM application has been 
killed due to memory violation, this sometimes crashes the whole system 
as the AESSYS tries to access memory formerly owned by the now-dead 
application (thus "type free"). Andd when AESSYS is killed, there's not 
much chance for a recover ...

And btw, what should a GEM application clean up in a signal handler? 
AFAIK, AES- and VDI-calls aren't allowed in signal handlers, anyway, so 
where's the point?


Ciao

Thomas


-- 
Thomas Binder (Gryf @ IRC)     binder@rbg.informatik.tu-darmstadt.de
PGP-key available on request!  gryf@hrzpub.tu-darmstadt.de

Attachment: pgpSZhwQCfLZW.pgp
Description: PGP signature