Re: [Xen-devel] [[PATCH for-4.13]] xen/arm: mm: Allow generic xen page-tables helpers to be called early

[Julien Grall <julien.grall@xxxxxxx>]

Hi Stefano,

On 10/11/19 6:11 PM, Stefano Stabellini wrote:
> > This is not pretty, but I don't view this as critical to fix for Xen 4.13. So
> > I am thinking to defer this post 4.13.
> If the issue doesn't trigger on 4.13, then I agree we shouldn't try to
> fix it (badly) at this stage.
>
> But we do have a "minus phys_offset" operation in dump_hyp_walk, I don't
> follow why we wouldn't see a problem if Xen is loaded at 1MB on arm32.
I haven't said the issue cannot be triggered. I pointed out I don't view 
this as critical.
One of the reason is that I bet nobody ever run Xen below 1MB on Arm32. 
Otherwise they would have seen an error...
In other words, I am not going to plan to fix it for Xen 4.13. However, 
if someone wants to spend time on it (and have platform to test it), 
then patch are welcomed.
> Xen pa: 0x100000
> Xen va: 0x200000
> phys_offset: 0xfff00000
>
> test_va: 0x202000
> test_va - phys_offset = 0xffffffff00300200. But it should be 0x102000. >
> Given that the problem is in a BUG_ON maybe we could remove it? Or just
> rework the BUG_ON?
For arm32, dump_hyp_walk() is only called when the AT instruction fails 
to translate a physical address. You are already doing something wrong 
if you hit, so you will panic in either case. The only difference is you 
don't get the page-table dumped.
Cheers,

--
Julien Grall

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel