On Sun, Apr 6, 2014 at 6:42 PM, Olof Johansson olof@lixom.net wrote:
On Sun, Apr 6, 2014 at 11:02 AM, Olof Johansson olof@lixom.net wrote:
On Sun, Apr 6, 2014 at 1:34 AM, Maxime Ripard maxime.ripard@free-electrons.com wrote:
Hi Olof,
On Sat, Apr 05, 2014 at 07:24:11AM -0700, Olof's autobooter wrote:
Failed boards:
cubie multi_v7_defconfig : FAILED 1:06.34I think you have the same issue Kevin had. You're loading the DT too close to the kernel, which is erasing the dtb whenever it uncompresses itself.
Ah, will fix.
Didn't seem to help. Moved it up quite a bit and same failure.
FWIW, here's what's working for me w/pyboot:
loadaddr: 0x41008000 dtb_addr: 0x42000000 dtb: %(dtb_base)s/sun4i-a10-cubieboard.dtb initrd_addr: 0x43000000
cubie2 sunxi_defconfig : FAILED 1:21.10 cubie2 multi_v7_defconfig : FAILED 1:08.11 cubie2 multi_lpae_defconfig : FAILED 1:04.54What revision of u-boot are you using?
I think you have the same issue Kevin had too, where the bootloader doesn't set the arch timer's CNTFRQ, resulting in a crash in the driver.
My u-boot is fairly old, but it's very annoying to have to update it. CNTFRQ is easy to fix by adding a propertly in the device-tree.
Still, I probably have to do that to get PSCI in this case.
This is a huge pita. What's the canonical u-boot to build and boot right now, that I will update to this once and then never touch again?
PITA, Totally agree. Doubly so because sunxi u-boot is not mainline.
Here's what I had to do: first, I'm using u-boot-sunxi[1], but then had to add a remote for maz's tree[2] and I built from the 'wip/psci' branch of maz's tree. When I built, wip/psci was at commit commit b489e198e1d05ca7d5eb28e001c99284159705bd, so that's what I"m currently running on the cubie2, but it appears wip/psci has moved a bit since then.
Kevin
[1] https://github.com/linux-sunxi/u-boot-sunxi.git [2] git://git.kernel.org/pub/scm/linux/kernel/git/maz/u-boot.git