Skip to content

bsp (virt64): expose EL2 on every AVZ machine of a chain without ATF - #72

Open
daniel-rossier wants to merge 1 commit into
mainfrom
fix/virt64-avz-kernel-machine
Open

daniel-rossier wants to merge 1 commit into
mainfrom
fix/virt64-avz-kernel-machine

Conversation

@daniel-rossier

Copy link
Copy Markdown
Contributor

On the chains without ATF, __do_platform_boot_chain builds the QEMU machine in two places. Only the boot-ROM one (U-Boot read off the card) added virtualization=on under IB_HYPERVISOR="avz". The -kernel u-boot/u-boot one, kept for a position-independent U-Boot, always wrote a plain virt,gic-version=2, so AVZ would fault on its first EL2 system-register access there.

Both now go through __virt64_machine(hyp), taken from so3, where bsp_virt64.inc is now byte-identical. The __do_platform_deploy comment also says that __deploy_arm_common handles both boot shapes, which it does.

The -kernel path is latent here: virt64 always builds virt64_defconfig, linked at a RAM address, so the card path is the one taken. Tested in so3 (virt64, uboot chain): the same file deploys and boots SO3 with U-Boot from the card.

Only the machine reading U-Boot off the card got virtualization=on under
AVZ. The one handing U-Boot to QEMU with -kernel, kept for a
position-independent U-Boot, was always a plain "virt,gic-version=2", so
AVZ would fault on its first EL2 system-register access there. Both now
take the machine from __virt64_machine(), as in so3.

The deploy comment also says that __deploy_arm_common handles both boot
shapes, which it does.

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant