linuxolator: amd64 Linux Test Project failures

[Divacky Roman]

On Wed, Jan 03, 2007 at 01:27:31AM +0100, Marcin Cieslak wrote:
> Divacky Roman wrote:
> >On Tue, Jan 02, 2007 at 09:55:47AM +0100, Alexander Leidinger wrote:
> >>Quoting Scot Hetzel <swhetzel at gmail.com> (from Sun, 31 Dec 2006  
> >>13:21:27 -0600):
> >>
> >>>That patch to mmap fixes the problems with mmap on amd64, and brings
> >>>the failed LTP testcases closer to the  i386 failed testcases.
> >>Thanks for testing, I try to get time to commit this.
> >
> >which makes me wonder what is MD on the linux_mmap* code. wouldnt it be 
> >better
> >to move that to some MI file instead?
> >
> >and there are tons of similar code.. for example linux_pause. how is this 
> >MD?
> >I'd vote for moving that code..
> >
> >opinions?
> 
> For example amd64 implements execute protection natively (PROT_EXEC), 
> while on i386 there is no way to separate this from PROT_READ. Other 
> platforms (sparc) may have different mmap implementations and we may 
> require different linux_mmap() behaviour (for example, on i386 PROT_EXEC 
> silently implies PROT_READ, on amd64 it does not have to be the case).
> 
> Having said that, we *may* be lucky and end up with identical mmap 
> emulation for all platforms. But I would prefer to test PROT_EXEC 
> compatibility first before we do that. Probably we should extend mmap 
> fingerprinter to test for expected PROT_EXEC behaviour.

hmm.. this is true.. I didnt think about this. Looks like we should scrutinize
every such case and decide where it belongs. The fact is that we will need
native amd64 linuxulator and it will (or at least should) use the compat/linux/
code..

thnx for the point marcin