Re: [MiNT] Memory protection

["Anthony Jacques" <anthonyj@jacquesa.freeserve.co.uk>]

Hi

>> Actually, #3 should already be that way.  The problem is that a
number
>> of the new protocols (e.g. OLGA) don't work well under memory
protection
>> (so I hear, anyway).  Hence was born #2, which is less powerful than
#3
>> but more flexible.
>
>This is a problem of the application. If they use the right Mxalloc()
>mode there is no problem (also if private virtual address rooms are
>implemented).

OLGA includes things like passing pointers to functions, which will
cause problems as the code will be in private memory that doesn't belong
to the calling app...

Obvious solutions:

1) Load entire program into global memory using header flags... watch
for the crashes.
2) Each application includes code to copy chunks of code into
Mxalloc()'d chunks, and perform all relocation that is required.
3) Change protocol to avoid this... extremely unlikely, IMO.

You will NOT get many applications doing the second option, and most
will forget to set the flags for 1 and users may not bother (this is the
current situation with apps that break under memory protection... thats
why most people consider MP to be broken).

I think its right to consider options which might make MP more
widespread, even if they are "weaker".

Anthony
----------------------------------+----------------------------------
Anthony Jacques     IRC: AnthonyJ | The boy stood on the burning deck
                    ICQ: 11287923 | Whence all but he had fled...
anthonyj@jacquesa.freeserve.co.uk | Twit.
http://ultra.quake2.co.uk/        |                    Spike Milligan
----------------------------------+----------------------------------