Re: Testing PROT_NONE and other protections, and a surprise

[Jamie Lokier <jamie@shareable.org>]

Kyle Moffett wrote:
> >Can you confirm in a simple way that mapping a file, or some anonymous
> >memory, without PROT_READ, really isn't writable under MacOS X?  Can
> >you confirm it with a word write, if that would be relevant?
> 
> I hope I didn't make some stupid mistake in my program, but here it
> is, and here are my results.

Thanks for testing, Kyle.

It looks fine, although this is wrong:

>         mem = mmap(0,4096,PROT_WRITE,MAP_ANON|MAP_SHARED,-1,0);
> ...
>         if (mem == 0) return 1;

The error code is -1, aka. MAP_FAILED.

> Starting...
> Mapped memory!
> Address is 4000

That's a surprisingly low address.

> Bus error

Phew, I'm glad I decided to catch SIGBUS in the test program at the
last moment...

That's a historical BSD-ism.  They can't change it now, because
programs do trap and check for SIGBUS on that platform for protection
violations.

> I'll probably go file a bug with Apple now :-D

It might be a generic *BSD bug (for whatever value of * is used by MacOS X).

That would be interesting to know -- anyone here running *BSD on PPC
or any other architecture to test?

Of course it's an Apple bug as well :)

-- Jamie