Hi Tyler, I've gotten the ptlcalls working with arbitrary addresses (it'll map whatever physical address comes out of the CPUID instruction). Here's the updated ptcalls.h file:
https://gist.github.com/dramninjasUMD/7471138 I haven't taken the time to neatly #define the new sections to merge it into the old ptlcalls code yet; I just deleted the unused bits. I am going to ask one of the Kitten developers to sanity check my code first and do the cleanup later. Cheers, Paul On Wed, Nov 13, 2013 at 1:19 PM, <[email protected]> wrote: > Does 0xFFF00000U work for you? IIRC, this is the address that I'm using > for the ARM port and it worked with the Linux x86-64 images that I tried. > It would be if we could find a mutual address at some point. > > Tyler > > > After taking several failed avenues, it turned out that it is actually > > quite straightforward to get kitten running in marss. > > > > I am currently trying to go back and change the ptlcalls so that both > > linux > > and kitten can use the same physical address for the ptlcall MMIO. If I > > can > > get it to do this, then one could use an unmodified copy of marss to run > > kitten and the only thing that would need to be changed is the ptlcall > > code. Let me see if I can get this working today and then I'll post the > > ptlcall changes. > > > > -Paul > > > > > > On Tue, Nov 12, 2013 at 8:49 PM, <[email protected]> wrote: > > > >> Paul, > >> > >> I'd love to take a look at commits if it's not too much trouble. I'm > >> willing to venture that there are some commits that expose or fix bugs > >> that we haven't seen yet... > >> > >> Tyler > >> > >> > Hello all, > >> > > >> > I've recently been working on getting MARSSx86 working with the Kitten > >> > Lightweight Kernel that is intended for super computer nodes (see: > >> > https://software.sandia.gov/trac/kitten ). Essentially it's a lot > like > >> > Linux, but with a lot of the protection and platform ubiquity stripped > >> > out. > >> > It's an interesting target for kernel development as it should (in > >> theory) > >> > be way easier to add features. As the kernel targets x86_64 and > >> already > >> > runs in QEMU, it's a perfect fit for marss. > >> > > >> > Anyways, I'm happy to say, I was able to get got it working by > >> replacing > >> > the ptlcall mmio to /dev/map with some of kitten's own mapping > >> functions. > >> > I > >> > also had the change the ptlcall physical address to a lower address in > >> > memory (though, there might be a way to keep it the same). > >> > > >> > If there is any interest, I am willing to share my patches. > >> > > >> > -Paul > >> > _______________________________________________ > >> > http://www.marss86.org > >> > Marss86-Devel mailing list > >> > [email protected] > >> > https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel > >> > > >> > >> > >> > > > >
_______________________________________________ http://www.marss86.org Marss86-Devel mailing list [email protected] https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
