On Tue, Aug 22, 2017 at 12:12 AM, Michael Ellerman <mpe@xxxxxxxxxxxxxx> wrote:
> Rob Herring <robh@xxxxxxxxxx> writes:
>
>> In preparation to remove the full path from device_node.full_name, use
>> of_find_node_by_path instead of open coding with strcmp.
>>
>> Signed-off-by: Rob Herring <robh@xxxxxxxxxx>
>> Cc: Benjamin Herrenschmidt <benh@xxxxxxxxxxxxxxxxxxx>
>> Cc: Paul Mackerras <paulus@xxxxxxxxx>
>> Cc: Michael Ellerman <mpe@xxxxxxxxxxxxxx>
>> Cc: linuxppc-dev@xxxxxxxxxxxxxxxx
>> ---
>> v2:
>> - rebased to linux-next and removed spurious change fro patch 1.
>>
>> arch/powerpc/platforms/pseries/vio.c | 4 ++--
>> 1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/arch/powerpc/platforms/pseries/vio.c b/arch/powerpc/platforms/pseries/vio.c
>> index aa5ca74316fa..5754572deb23 100644
>> --- a/arch/powerpc/platforms/pseries/vio.c
>> +++ b/arch/powerpc/platforms/pseries/vio.c
>> @@ -1357,9 +1357,9 @@ struct vio_dev *vio_register_device_node(struct device_node *of_node)
>> */
>> parent_node = of_get_parent(of_node);
>> if (parent_node) {
>> - if (!strcmp(parent_node->full_name, "/ibm,platform-facilities"))
>> + if (parent_node == of_find_node_by_path("/ibm,platform-facilities"))
>> family = PFO;
>> - else if (!strcmp(parent_node->full_name, "/vdevice"))
>> + else if (parent_node == of_find_node_by_path("/vdevice"))
>> family = VDEVICE;
>
> This leaks references to the looked up nodes.
>
> Both these nodes are defined in PAPR (our hypervisor spec), and both of
> them must have a device_type, either "ibm,platform-facilities" or
> "vdevice".
>
> Looking at the commit that added the code I don't see any particular
> reason it used the comparison against full_name, rather than using the
> device_type.
>
> So I'm inclined to do this instead:
>
> diff --git a/arch/powerpc/platforms/pseries/vio.c b/arch/powerpc/platforms/pseries/vio.c
> index 8a47f168476b..f26f906e6021 100644
> --- a/arch/powerpc/platforms/pseries/vio.c
> +++ b/arch/powerpc/platforms/pseries/vio.c
> @@ -1357,9 +1357,9 @@ struct vio_dev *vio_register_device_node(struct device_node *of_node)
> */
> parent_node = of_get_parent(of_node);
> if (parent_node) {
> - if (!strcmp(parent_node->full_name, "/ibm,platform-facilities"))
> + if (!strcmp(parent_node->type, "ibm,platform-facilities"))
> family = PFO;
> - else if (!strcmp(parent_node->full_name, "/vdevice"))
> + else if (!strcmp(parent_node->type, "vdevice"))
> family = VDEVICE;
> else {
> pr_warn("%s: parent(%s) of %s not recognized.\n",
>
>
> I've checked both Qemu and kvmtool add the device_type, and I'm fairly
> confident that PowerVM does too. Anyway I'll test it on all the machines
> I can find.
Okay, do you want me to respin the patch or will you update it with
the above change?
Rob
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
