[Xen-devel] OMAP5 port.

[Andrii Anisov <andrii.anisov@xxxxxxxxxxxxxxx>]

Hello Ian,

Thank you for your answer about the GIC.

We are doing an experimental porting XEN on a TI OMAP5 based board.

Yep, TI OMAP5 is stillborn but it's a Cortex-A15 based SOC and we have some boards on a table.

So far we got XEN up and loading Dom0 Linux.

We have a linux 3.4 port for OMAP5. Backported xen/arm patches from git://xenbits.xen.org/people/sstabellini/linux-pvhvm.git .

Currently our Dom0 boot dies at xenbus_init.

Getting parameters HVM_PARAM_STORE_EVTCHN and HVM_PARAM_STORE_PFN from XEN returns 0, next access to xen_store_evtchn or xen_store_interface fields is blocked by xen. 

I did find how this parameters are acquired from XEN, but they are not initialized. Could you please clarify how arch.hvm_domain.params should be initialized. Unfortunately I got no clue about this from vexpress or linaro xen sources.

One more question: is http://lists.xen.org/archives/html/xen-devel/2012-09/msg01091.html the latest version of the patchset?  Is it available as a tarball?

Sincerely,

Andrii Anisov.

On Wed, Mar 6, 2013 at 2:21 AM, Ian Campbell <ian.campbell@xxxxxxxxxx> wrote:

On Tue, 2013-03-05 at 19:57 +0000, Andrii Anisov wrote:
>
> In a function arch_domain_create() I do see mapping of GIC CPU
> interface registers space of VM to point to the GIC virtual CPU
> interface. But can't find how XEN handles VM's accesses to GIC
> Distributor registers. Could you please explain me how is it done, or
> point to the correspondent code.

It's handled from the second stage fault handler, see the call to
handle_mmio from do_trap_data_abort_guest. This eventually ends up
calling vgic_distr_mmio_handler->{read,write} AKA vgic_distr_mmio_read
(and ..._write).

Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel