<div dir="ltr"><br><div class="gmail_extra"><br><div class="gmail_quote">On Thu, Mar 1, 2018 at 11:30 PM, Patrick Rudolph <span dir="ltr"><<a href="mailto:patrick.rudolph@9elements.com" target="_blank">patrick.rudolph@9elements.com</a><wbr>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span>Am Donnerstag, den 01.03.2018, 14:11 -0800 schrieb Julius Werner:<br>
> Hi Patrick et. al.,<br>
><br>
> Continuing from what you said on IRC, let's please discuss this<br>
> before you spend time to work on any major changes to the ARMv8 MMU<br>
> code. I don't think that should be necessary (especially changing the<br>
> GRANULE_SIZE which is a complication I'd really like to avoid<br>
> having), and I'd like to understand what you're trying to do first.<br>
> Can you explain what memory ranges you need to map on your SoC and<br>
> why they need to be so large?<br>
><br>
</span>Hi Julius,<br>
the Cavium SoC expose the MMIO space at 1 << 47 with a size of 1 << 47.<br>
As coreboot does use PA=VA mapping, I have to extend the VA space to 48<br>
bit and that seems to work fine with 64KB granule.<br>
<br>
The change in coreboot is minimal, I'm not sure if all armv8 CPUs<br>
support's that granule, but a Kconfig should do for now.<br></blockquote><div><br></div><div>Yes, the architecture reference manual for ARMv8 mentions that 4KB, 16KB, and 64KB are supported for VMSAv8-64.</div></div></div></div>