Re: [Qemu-devel] Using virtio-mmio

[Peter Maydell]

On 1 August 2013 11:41, Richard W.M. Jones <address@hidden> wrote:
> On Thu, Aug 01, 2013 at 10:32:14AM +0100, Peter Maydell wrote:
>> You might want to consider -M vexpress-a15, if you
>> want a setup that will let you use KVM (will probably
>> need to reconfig your kernel appropriately; may
>> need to discard "earlyprintk" if you compiled the
>> kernel with the dodgy "guess earlyprintk serial port
>> address based on exact revision-and-patchlevel of CPU"
>> option.)
>
> Unfortunately vexpress-a15 causes the kernel to fail to boot.
> I guess the Fedora kernel is not configured to work with A15 yet.

> $ grep VEXPRESS config-3.9.9-302.fc19.armv7hl
> CONFIG_ARCH_VEXPRESS=y
> CONFIG_ARCH_VEXPRESS_CORTEX_A5_A9_ERRATA=y
> CONFIG_ARCH_VEXPRESS_CA9X4=y
> CONFIG_SENSORS_VEXPRESS=m
> CONFIG_VEXPRESS_CONFIG=y
> CONFIG_REGULATOR_VEXPRESS=m

Actually the vexpress-A15 doesn't need any extra config
switches the way the vexpress-A9 does, because it's all
device-tree controlled.

Unfortunately at this point you run into the classic
issue of trying to get an ARM kernel running, which
is that a huge class of config errors all have the
failure mode "just sits there with no serial output".
This is remarkably user-unfriendly but I don't really
know how we could fix it since the underlying cause
is that serial ports on ARM aren't in a single standard
location, so if the kernel's not configured right it
won't find the serial port. I usually end up resorting
to running with gdb connected to qemu's debug stub
and fishing the kernel output out of its log_buf :-(

-- PMM