On Tue, Jan 6, 2026 at 3:45 PM Shivendra Pratap
[off-list ref] wrote:

On 1/6/2026 6:04 PM, Bartosz Golaszewski wrote:

quoted

On Mon, Jan 5, 2026 at 7:06 PM Shivendra Pratap
[off-list ref] wrote:

quoted

quoted

quoted

+static int __init psci_reboot_mode_init(void)
+{
+       struct faux_device *fdev;
+
+       fdev = faux_device_create("psci-reboot-mode", NULL, &psci_reboot_mode_ops);
+       if (!fdev)
+               return -ENODEV;

This will always create this device for everyone who includes this
module. Move the of_find_compatible_node(NULL, NULL, "arm,psci-1.0")
call from probe() here instead and don't create the device if it
fails.

Ack.
Will move both calls to init before creating the faux device.

psci_np = of_find_compatible_node(NULL, NULL, "arm,psci-1.0");
and
np = of_find_node_by_name(psci_np, "reboot-mode");
--

On a second glance - and I may be totally wrong - would it be possible
to switch to using the auxiliary bus and create this device from
drivers/firmware/psci/psci.c? That would be even cleaner.

Till v17, device was being created in psci.c. Lorenzo wanted to move it
outside psci similar to design of cpuidle-psci.

https://lore.kernel.org/all/aRIfc9iuC2b9DqI+@lpieralisi/ (local)

Thanks for the link. Right, there's no actual psci driver binding to a
struct device, rather we only have a set of functions called very
early into the boot process.

Nevermind this comment

Bart