Hi Julien,

Thanks, it looks like the issue was with the way I verified this - cat on
/sys/firmware/devicetree/base/psci/compatible from dom0's console returns
only the first compatible string set by Xen no matter what.
I have verified by debugging that linux finds the right compatible string
in psci_dt_init().


On Thu, Feb 1, 2018 at 8:12 PM, Julien Grall <julien.gr...@arm.com> wrote:

> On 01/02/18 19:00, Mirela Simonovic wrote:
>> Hi,
> Hi Mirela,
>> I'm working on switching to PSCI v1.0 in Xen (needed to support suspend
>> > to RAM).
> I am about to send a series to add support for PSCI 1.1 (I need it for the
> spectre mitigation). I can send you a temporary branch if you want to see
> whether I missed anything.
>> In addition to updates in PSCI implementation, we assumed that compatible
>> property in psci device tree node should be extended with "arm,psci-1.0"
>> (not a must since this doesn't affect Dom0's functionality today, but I
>> would add it if possible).
>> Could you please let me know how to add "arm,psci-1.0" to compatible
>> property?
>> I assumed that in domain_build.c in make_psci_node() function I need to
>> add "arm,psci-1.0""\0" at the beginning of 'compat' string (although I
>> don't understand the "\0"). That doesn't seem to work. Only "arm,psci-1.0"
>> ends up in compatible property and we need strings for older versions as
>> well.
>> I also noticed that even "arm,psci" doesn't appear to be added into the
>> compatible property (without any code modifications from my side). I
>> checked what is generated by debugging and by 'cat' on
>> /sys/firmware/devicetree/base/psci/compatible from Dom0's console.
> From the DT specification, each compatible string should be separated by
> '\0'. Doing a hexdump -C in the hardware domain I can see that both
> arm,psci and arm,psci-0.2 are present on latest Xen. So I am not sure why
> it would not be present on your side.
> BTW, if you use printk, only the first compatible will be printed (printk
> stop at the '\0').
> Cheers,
> --
> Julien Grall
Xen-devel mailing list

Reply via email to