The current ARM DT topology description provides the operating system
with a topological view of the system that is based on leaf nodes
representing either cores or threads (in an SMT system) and a
hierarchical set of cluster nodes that creates a hierarchical topology
view of how those cores and threads are grouped.
However this hierarchical representation of clusters does not allow to
describe what topology level actually represents the physical package or
the socket boundary, which is a key piece of information to be used by
an operating system to optimize resource allocation and scheduling.
Lets add a new "socket" node type in the cpu-map node to describe the
same.
Signed-off-by: Sudeep Holla <[email protected]>
---
.../devicetree/bindings/arm/topology.txt | 52 ++++++++++++++-----
1 file changed, 39 insertions(+), 13 deletions(-)
(Note patch generated with -b option to avoid 60+ of whitespace changes)
Hi Rob,
You had expressed your interest to generalise the CPU topology bindings
accross multiple architectures. Do you want to move to the generic
bindings before adding this $subject socket support or is it OK to
finalise on this and then move the majority(based on the agreement)
to generic binding.
Regards,
Sudeep
diff --git a/Documentation/devicetree/bindings/arm/topology.txt b/Documentation/devicetree/bindings/arm/topology.txt
index de9eb0486630..668483554cca 100644
--- a/Documentation/devicetree/bindings/arm/topology.txt
+++ b/Documentation/devicetree/bindings/arm/topology.txt
@@ -9,6 +9,7 @@ ARM topology binding description
In an ARM system, the hierarchy of CPUs is defined through three entities that
are used to describe the layout of physical CPUs in the system:
+- socket
- cluster
- core
- thread
@@ -63,21 +64,23 @@ nodes are listed.
The cpu-map node's child nodes can be:
- - one or more cluster nodes
+ - one or more cluster nodes or
+ - one or more socket nodes in a multi-socket system
Any other configuration is considered invalid.
-The cpu-map node can only contain three types of child nodes:
+The cpu-map node can only contain 4 types of child nodes:
+- socket node
- cluster node
- core node
- thread node
whose bindings are described in paragraph 3.
-The nodes describing the CPU topology (cluster/core/thread) can only
-be defined within the cpu-map node and every core/thread in the system
-must be defined within the topology. Any other configuration is
+The nodes describing the CPU topology (socket/cluster/core/thread) can
+only be defined within the cpu-map node and every core/thread in the
+system must be defined within the topology. Any other configuration is
invalid and therefore must be ignored.
===========================================
@@ -85,26 +88,44 @@ invalid and therefore must be ignored.
===========================================
cpu-map child nodes must follow a naming convention where the node name
-must be "clusterN", "coreN", "threadN" depending on the node type (ie
-cluster/core/thread) (where N = {0, 1, ...} is the node number; nodes which
-are siblings within a single common parent node must be given a unique and
+must be "socketN", "clusterN", "coreN", "threadN" depending on the node type
+(ie socket/cluster/core/thread) (where N = {0, 1, ...} is the node number; nodes
+which are siblings within a single common parent node must be given a unique and
sequential N value, starting from 0).
cpu-map child nodes which do not share a common parent node can have the same
name (ie same number N as other cpu-map child nodes at different device tree
levels) since name uniqueness will be guaranteed by the device tree hierarchy.
===========================================
-3 - cluster/core/thread node bindings
+3 - socket/cluster/core/thread node bindings
===========================================
-Bindings for cluster/cpu/thread nodes are defined as follows:
+Bindings for socket/cluster/cpu/thread nodes are defined as follows:
+
+- socket node
+
+ Description: must be declared within a cpu-map node, one node
+ per physical socket in the system. A system can
+ contain single or multiple physical socket.
+ The association of sockets and NUMA nodes is beyond
+ the scope of this bindings, please refer [2] for
+ NUMA bindings.
+
+ This node is optional for a single socket system.
+
+ The socket node name must be "socketN" as described in 2.1 above.
+ A socket node can not be a leaf node.
+
+ A socket node's child nodes must be one or more cluster nodes.
+
+ Any other configuration is considered invalid.
- cluster node
Description: must be declared within a cpu-map node, one node
per cluster. A system can contain several layers of
- clustering and cluster nodes can be contained in parent
- cluster nodes.
+ clustering within a single physical socket and cluster
+ nodes can be contained in parent cluster nodes.
The cluster node name must be "clusterN" as described in 2.1 above.
A cluster node can not be a leaf node.
@@ -164,13 +185,15 @@ levels) since name uniqueness will be guaranteed by the device tree hierarchy.
4 - Example dts
===========================================
-Example 1 (ARM 64-bit, 16-cpu system, two clusters of clusters):
+Example 1 (ARM 64-bit, 16-cpu system, two clusters of clusters in a single
+physical socket):
cpus {
#size-cells = <0>;
#address-cells = <2>;
cpu-map {
+ socket0 {
cluster0 {
cluster0 {
core0 {
@@ -253,6 +276,7 @@ cpus {
};
};
};
+ };
CPU0: cpu@0 {
device_type = "cpu";
@@ -473,3 +497,5 @@ cpus {
===============================================================================
[1] ARM Linux kernel documentation
Documentation/devicetree/bindings/arm/cpus.txt
+[2] Devicetree NUMA binding description
+ Documentation/devicetree/bindings/numa.txt
--
2.17.1
On 11/12/18 3:37 PM, Rob Herring wrote:
> On Wed, Nov 07, 2018 at 05:13:44PM +0000, Sudeep Holla wrote:
>> The current ARM DT topology description provides the operating system
>> with a topological view of the system that is based on leaf nodes
>> representing either cores or threads (in an SMT system) and a
>> hierarchical set of cluster nodes that creates a hierarchical topology
>> view of how those cores and threads are grouped.
>>
>> However this hierarchical representation of clusters does not allow to
>> describe what topology level actually represents the physical package or
>> the socket boundary, which is a key piece of information to be used by
>> an operating system to optimize resource allocation and scheduling.
>>
>> Lets add a new "socket" node type in the cpu-map node to describe the
>> same.
>>
>> Signed-off-by: Sudeep Holla <[email protected]>
>> ---
>> .../devicetree/bindings/arm/topology.txt | 52 ++++++++++++++-----
>> 1 file changed, 39 insertions(+), 13 deletions(-)
>>
>> (Note patch generated with -b option to avoid 60+ of whitespace changes)
>>
>> Hi Rob,
>>
>> You had expressed your interest to generalise the CPU topology bindings
>> accross multiple architectures. Do you want to move to the generic
>> bindings before adding this $subject socket support or is it OK to
>> finalise on this and then move the majority(based on the agreement)
>> to generic binding.
>
> Doesn't really matter to me as long as Risc-V folks are in agreement.
>
> Otherwise, this looks fine to me.
>
> Rob
>
>
I can apply this patch in my unify topology series and resend everything
together as one series.
Regards,
Atish
On Mon, Nov 19, 2018 at 05:18:42PM -0800, Atish Patra wrote:
> On 11/12/18 3:37 PM, Rob Herring wrote:
> > On Wed, Nov 07, 2018 at 05:13:44PM +0000, Sudeep Holla wrote:
> > > The current ARM DT topology description provides the operating system
> > > with a topological view of the system that is based on leaf nodes
> > > representing either cores or threads (in an SMT system) and a
> > > hierarchical set of cluster nodes that creates a hierarchical topology
> > > view of how those cores and threads are grouped.
> > >
> > > However this hierarchical representation of clusters does not allow to
> > > describe what topology level actually represents the physical package or
> > > the socket boundary, which is a key piece of information to be used by
> > > an operating system to optimize resource allocation and scheduling.
> > >
> > > Lets add a new "socket" node type in the cpu-map node to describe the
> > > same.
> > >
> > > Signed-off-by: Sudeep Holla <[email protected]>
> > > ---
> > > .../devicetree/bindings/arm/topology.txt | 52 ++++++++++++++-----
> > > 1 file changed, 39 insertions(+), 13 deletions(-)
> > >
> > > (Note patch generated with -b option to avoid 60+ of whitespace changes)
> > >
> > > Hi Rob,
> > >
> > > You had expressed your interest to generalise the CPU topology bindings
> > > accross multiple architectures. Do you want to move to the generic
> > > bindings before adding this $subject socket support or is it OK to
> > > finalise on this and then move the majority(based on the agreement)
> > > to generic binding.
> >
> > Doesn't really matter to me as long as Risc-V folks are in agreement.
> >
> > Otherwise, this looks fine to me.
> >
> > Rob
> >
> >
> I can apply this patch in my unify topology series and resend everything
> together as one series.
Thanks for that. You can drop RFC when reposting. Remember to use -b
for ignoring space changes on this patch along with -M for renames in
your original patch series.
--
Regards,
Sudeep