Kevin,
Here's the list of patches:
On Wed, May 07, 2014 at 12:51:56PM -0700, Kevin Hilman wrote:
Thomas Petazzoni thomas.petazzoni@free-electrons.com writes:
Dear Jason Cooper,
On Wed, 7 May 2014 08:11:30 -0400, Jason Cooper wrote:
iirc, these boot into an initrd. If so, are the errors below an artifact of the boot farm?
Gregory, Thomas, Ezequiel,
If these are legitimate errors, can you please take a look and tell me if anything needs backported?
Weird. I don't think anything has changed in terms of initrd handling recently, at least nothing specific to mvebu platforms. I remember discussing some U-Boot configuration tips at ELC with Kevin. Maybe he has changed his boot scripts/configurations a bit and it breaks on these platforms now? I'd like to hear from Kevin about this.
These are long standing boot failures that I had blacklisted from boot testing on stable, but in shuffling some things around, my blacklist got removed.
The problem here on the mvebu platforms is that the multi_v7_defconfig in v3.10.y is missing some kconfig options for using a bootloader-loaded initramfs. Specifically:
CONFIG_BLK_DEV_INITRD=y
c12d82b84353 arm: multi_v7_defconfig: Enable initrd/initramfs support
CONFIG_ARM_ATAG_DTB_COMPAT=y
a0396b9bd5a4 ARM: multi_v7_defconfig: enable ARM_ATAG_DTB_COMPAT
Also for the buildroot initramfs I use, I need:
CONFIG_DEVTMPFS=y CONFIG_DEVTMPFS_MOUNT=y
9cc98f01e270 ARM: Make multi_v7_defconfig usable on hardware
This one is rather large.
These have been in multi_v7_defconfig s since v3.12 I think, but missing from v3.10, which was the early days of multi_v7_defconfig anyways.
Someone needs to submit a patch to stable for this to get this working, or maybe I should just stop booting multi_v7_defconfig on v3.10.
Well, the question I have now: how far along was the multi_v7 effort at v3.10? Was a multi_v7 v3.10 kernel actually usable, assuming the config options were set properly?
If it wasn't then yeah, lets stop booting v3.10 multi_v7. If it was, then I'll put together the email to stable for the patches to be backported.
thx,
Jason.