Re: [Xen-devel] XEN[ARM] Master not working on Allwinner A20

[Ian Campbell <Ian.Campbell@xxxxxxxxxx>]

On Tue, 2014-01-07 at 14:31 +0000, karim.allah.ahmed@xxxxxxxxx wrote:
> On Tue, Jan 7, 2014 at 1:58 PM, Ian Campbell <Ian.Campbell@xxxxxxxxxx> wrote:
> > On Sun, 2014-01-05 at 16:48 +0000, karim.allah.ahmed@xxxxxxxxx wrote:
> >
> >> If you still can't boot with any memory bigger than 128M, as a fast
> >> workaround you can apply this patch.
> >
> > I wonder if it might be possible to work around this by more carefully
> > selecting the load addresses for Xen+Linux+DTB+initrd, such that they
> > are packed into the top end of RAM, leaving a larger contiguous chunk
> > available at the beginning. e.g. if sizeof(Xen)=X and sizeof(Linux)=L
> > and sizeof(DTB)=D (all rounded up to 2M boundary) then load things at:
> >         MEMMAX-X:       Leave free for high relocation of hypervisor
> >         MEMMAX-X-L:     Load Linux here
> >         MEMMAX-X-L-D:   Load DTB here
> >         MEMMAX-X-L-D-X: Load initial Xen image here
> >
> > Ultimately this is because allocations need to be aligned to their size,
> > so on a 1GB system there are only two possible 512MB allocations, if
> > even one page is allocated in each half then it isn't possible to
> > satisfy things. I don't think the core allocator gives us the option to
> > do non-aligned allocations.
> 
> What if we allocated the dom0 from the boot allocator instead (before
> ditching it) ?

Have you checked if the boot allocator has the same constraints?

I'm also not sure if we have enough info during the early phase to know
what we are supposed to be doing (i.e. have we parsed the command line
yet?).

If you can come up with a patch we can consider it, but to be considered
for being 4.4 material it'd have to be pretty straightforward and
obvious.

Ian.



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel