On 12/15/2017 01:19 AM, Mario Six wrote:
On Fri, Dec 15, 2017 at 9:05 AM, Mario Six <mario....@gdsys.cc> wrote:
Hi Simon,
On Thu, Dec 14, 2017 at 9:36 PM, Simon Glass <s...@chromium.org> wrote:
+Stephen, Tom
Hi Mario,
I've had to drop this since it breaks tegra. Stephen feels that this
is likely a bug in the patch rather than anything wrong with Tegra. Do
you have any thoughts? I can potentially try things out on the Tegra
boards I have.
Regards,
Simon
I think I've found the problem: It's an endianness issue. The
of_translate_address function seems to return a raw big-endian address, while
the of_read_number function translates it to CPU endianness before returning
it. On the MPC8308 I tested on this made no difference, but little-endian
systems will have problems. When I put a bogus dev_read_addr in the sandbox
serial driver, I see the effect as well.
Here's a hopefully fixed version (just the diff; I'll post a proper patch when
your tests succeeded); it corrects the issue on sandbox, so I hope it will work
on tegra as well:
...
Small correction:
Yes, that seems to work. I tested:
a) u-boot-dm works fine on Jetson TK1.
b) Applying your original patch reproduces the failure.
c) Applying this latest patch make everything work again.
Thanks.
_______________________________________________
U-Boot mailing list
U-Boot@lists.denx.de
https://lists.denx.de/listinfo/u-boot
