Skip to contentSkip to conversation

ARM CCI support

RFC PATCH New Lorenzo Pieralisi 5 participants 10 messages May 9, 2013 – May 21, 2013 Device tree bindings

lore
Share & tools
lore
#0 lore raw

RFC PATCH v3 1/1 ARM CCI support

This patch is an updated of a previous posting:

http://lists.infradead.org/pipermail/linux-arm-kernel/2013-May/166173.html

main changes in v3:
- Changed init to early_initcall and added code to deal with init calls ordering issues
- Removed platform driver infrastructure, replaced with DT layer
- Added missing pointer flushing
- Removed useless includes
- Implemented MPIDR stashing
- Added basic functions to initialize and carry out CPU ports look-ups
- fixed !CONFIG_ARM_CCI compile issue

main changes in v2:
- Changed way masters are connected to CCI slave ports in DT

   # now every master has to declare a phandle to a CCI control IF
 - Fixed warning split on different lines
 - Added comments to disable function
 - Moved driver to non-hotpluggable API (registration and probing bundled
   together)
 - Upgraded cache flushing API
 - Reworked the driver to use devm primitives
 - Added disable/enable by index API
 - Fixed port enable and added macro for snoops/DVM control

ARM multi-cluster systems rely on the cache coherent interconnect (CCI) to support inter-cluster cache coherency and distributed virtual memory (DVM) messaging.

CCI kernel driver requires tweaks to the kernel and device tree bindings in order to link specific resources to groups of logical CPUs so that resources can be associated to specific cpu sets (slave ports can be associated with a set of CPUs which are part of a cluster).

CCI slave ports must be associated with specific bus masters in the system so that whenever a CCI operation is requested for a specific master (ie disable its CCI slave port, read its PMU counters) the association can be carried out dynamically in the OS through dynamically built data structures. Bus masters in the system connected to CCI ports must define device tree properties, as described in the documentation contained in the patch, in order to explicitly connect ports and allow the OS to build up the required data structures.

The current CCI DT bindings define the CCI address space as the same one as the root device tree node, which means that the CCI bus can address the entire address space visible to CPUs.

Lorenzo Pieralisi (1):
drivers: bus: add ARM CCI support

 Documentation/devicetree/bindings/arm/cci.txt | 161 ++++++++++
 drivers/bus/Kconfig                           |   7 +
 drivers/bus/Makefile                          |   2 +
 drivers/bus/arm-cci.c                         | 420 ++++++++++++++++++++++++++
 include/linux/arm-cci.h                       |  61 ++++
 5 files changed, 651 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/arm/cci.txt
 create mode 100644 drivers/bus/arm-cci.c
 create mode 100644 include/linux/arm-cci.h
Signature · 1 line
1.8.2.2
#1 lore raw

RFC PATCH v3 1/1 drivers: bus: add ARM CCI support

On ARM multi-cluster systems coherency between cores running on different clusters is managed by the cache-coherent interconnect (CCI). It allows broadcasting of TLB invalidates and memory barriers and it guarantees cache coherency at system level through snooping of slave interfaces connected to it.

This patch enables the basic infrastructure required in Linux to handle and programme the CCI component.

Non-local variables used by the CCI management functions called by power down function calls after disabling the cache must be flushed out to main memory in advance, otherwise incoherency of those values may occur if they are sitting in the cache of some other CPU when power down functions execute. Driver code ensures that relevant data structures are flushed from inner and outer caches after the driver probe is completed.

CCI slave port resources are linked to set of CPUs through bus masters phandle properties that link the interface resources to masters node in the device tree.

Documentation describing the CCI DT bindings is provided with the patch.

---
 Documentation/devicetree/bindings/arm/cci.txt | 161 ++++++++++
 drivers/bus/Kconfig                           |   7 +
 drivers/bus/Makefile                          |   2 +
 drivers/bus/arm-cci.c                         | 420 ++++++++++++++++++++++++++
 include/linux/arm-cci.h                       |  61 ++++
 5 files changed, 651 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/arm/cci.txt
 create mode 100644 drivers/bus/arm-cci.c
 create mode 100644 include/linux/arm-cci.h
Documentation/devicetree/bindings/arm/cci.txt new file +161 -0
new file mode 100644
index 0000000..384b460
@@ -0,0 +1,161 @@
+=======================================================
+ARM CCI cache coherent interconnect binding description
+=======================================================
+
+ARM multi-cluster systems maintain intra-cluster coherency through a
+cache coherent interconnect (CCI) that is capable of monitoring bus
+transactions and manage coherency, TLB invalidations and memory barriers.
+
+It allows snooping and distributed virtual memory message broadcast across
+clusters, through memory mapped interface, with a global control register
+space and multiple sets of interface control registers, one per slave
+interface.
+
+Bindings for the CCI node follow the ePAPR standard, available from:
+
+www.power.org/documentation/epapr-version-1-1/
+
+with the addition of the bindings described in this document which are
+specific to ARM.
+
+* CCI interconnect node
+
+	Description: Describes a CCI cache coherent Interconnect component
+
+	Node name must be "cci".
+	Node's parent must be the root node /, and the address space visible
+	through the CCI interconnect is the same as the one seen from the
+	root node (ie from CPUs perspective as per DT standard).
+	Every CCI node has to define the following properties:
+
+	- compatible
+		Usage: required
+		Value type: <string>
+		Definition: must be set to
+			"arm,cci-400"
+
+	- reg
+		Usage: required
+		Value type: <prop-encoded-array>
+		Definition: A standard property. Specifies base physical
+			    address of CCI control registers common to all
+			    interfaces.
+
+	- ranges:
+		Usage: required
+		Value type: <prop-encoded-array>
+		Definition: A standard property. Follow rules in the ePAPR for
+			    hierarchical bus addressing. CCI interfaces
+			    addresses refer to the parent node addressing
+			    scheme to declare their register bases.
+
+	CCI interconnect node can define the following child nodes:
+
+	- CCI control interface nodes
+
+		Node name must be "slave-if".
+		Parent node must be CCI interconnect node.
+
+		A CCI interface node must contain the following properties:
+		- interface-type:
+			Usage: required
+			Value type: <string>
+			Definition: must be set to one of {"ace", "ace-lite"}
+				    depending on the interface type the node
+				    represents.
+
+		- reg:
+			Usage: required
+			Value type: <prop-encoded-array>
+			Definition: the base address and size of the
+				    corresponding interface programming
+				    registers.
+
+* CCI interconnect bus masters
+
+	Description: masters in the device tree connected to a CCI port
+		     (inclusive of CPUs and their cpu nodes).
+
+	A CCI interconnect bus master node must contain the following
+	properties:
+
+	- cci-control-port:
+		Usage: required
+		Value type: <phandle>
+		Definition: a phandle containing the CCI control interface node
+			    the master is connected to.
+
+Example:
+
+	cpus {
+		#size-cells = <0>;
+		#address-cells = <1>;
+
+		CPU0: cpu at 0 {
+			device_type = "cpu";
+			compatible = "arm,cortex-a15";
+			cci-control-port = <&cci_control1>;
+			reg = <0x0>;
+		};
+
+		CPU1: cpu at 1 {
+			device_type = "cpu";
+			compatible = "arm,cortex-a15";
+			cci-control-port = <&cci_control1>;
+			reg = <0x1>;
+		};
+
+		CPU2: cpu at 100 {
+			device_type = "cpu";
+			compatible = "arm,cortex-a7";
+			cci-control-port = <&cci_control2>;
+			reg = <0x100>;
+		};
+
+		CPU3: cpu at 101 {
+			device_type = "cpu";
+			compatible = "arm,cortex-a7";
+			cci-control-port = <&cci_control2>;
+			reg = <0x101>;
+		};
+
+	};
+
+	dma0: dma at 3000000 {
+		compatible = "arm,pl330", "arm,primecell";
+		cci-control-port = <&cci_control0>;
+		reg = <0x0 0x3000000 0x0 0x1000>;
+		interrupts = <10>;
+		#dma-cells = <1>;
+		#dma-channels = <8>;
+		#dma-requests = <32>;
+	};
+
+	cci at 2c090000 {
+		compatible = "arm,cci";
+		#address-cells = <1>;
+		#size-cells = <1>;
+		reg = <0x0 0x2c090000 0 0x1000>;
+		ranges = <0x0 0x0 0x2c090000 0x6000>;
+
+		cci_control0 : slave-if at 1000 {
+			interface-type = "ace-lite";
+			reg = <0x1000 0x1000>;
+		};
+
+		cci_control1 : slave-if at 4000 {
+			interface-type = "ace";
+			reg = <0x4000 0x1000>;
+		};
+
+		cci_control2 : slave-if at 5000 {
+			interface-type = "ace";
+			reg = <0x5000 0x1000>;
+		};
+	};
+
+This CCI node corresponds to a CCI component whose control registers sits
+at address 0x000000002c090000.
+CCI slave interface @0x000000002c091000 is connected to dma controller dma0.
+CCI slave interface @0x000000002c094000 is connected to CPUs {CPU0, CPU1};
+CCI slave interface @0x000000002c095000 is connected to CPUs {CPU2, CPU3};
drivers/bus/Kconfig +7 -0
index 0f51ed6..38ea85c 100644
@@ -19,4 +19,11 @@ config OMAP_INTERCONNECT
 
 	help
 	  Driver to enable OMAP interconnect error handling driver.
+
+config ARM_CCI
+	bool "ARM CCI driver support"
+	depends on ARM
+	help
+	  Driver supporting the CCI cache coherent interconnect for ARM
+	  platforms.
 endmenu
drivers/bus/Makefile +2 -0
index 45d997c..2bd9947 100644
@@ -6,3 +6,5 @@ obj-$(CONFIG_OMAP_OCP2SCP)	+= omap-ocp2scp.o
 
 # Interconnect bus driver for OMAP SoCs.
 obj-$(CONFIG_OMAP_INTERCONNECT)	+= omap_l3_smx.o omap_l3_noc.o
+# CCI cache coherent interconnect for ARM platforms
+obj-$(CONFIG_ARM_CCI)		+= arm-cci.o
drivers/bus/arm-cci.c new file +420 -0

420 lines not on this page · raw message

include/linux/arm-cci.h new file +61 -0

61 lines not on this page · raw message

Signature · 1 line
1.8.2.2
#2 lore raw

RFC PATCH v3 1/1 drivers: bus: add ARM CCI support

Quoted 34 lines · Lorenzo Pieralisi

On ARM multi-cluster systems coherency between cores running on different clusters is managed by the cache-coherent interconnect (CCI). It allows broadcasting of TLB invalidates and memory barriers and it guarantees cache coherency at system level through snooping of slave interfaces connected to it.

This patch enables the basic infrastructure required in Linux to handle and programme the CCI component.

Non-local variables used by the CCI management functions called by power down function calls after disabling the cache must be flushed out to main memory in advance, otherwise incoherency of those values may occur if they are sitting in the cache of some other CPU when power down functions execute. Driver code ensures that relevant data structures are flushed from inner and outer caches after the driver probe is completed.

CCI slave port resources are linked to set of CPUs through bus masters phandle properties that link the interface resources to masters node in the device tree.

Documentation describing the CCI DT bindings is provided with the patch.

---
 Documentation/devicetree/bindings/arm/cci.txt | 161 ++++++++++
 drivers/bus/Kconfig                           |   7 +
 drivers/bus/Makefile                          |   2 +
 drivers/bus/arm-cci.c                         | 420 ++++++++++++++++++++++++++
 include/linux/arm-cci.h                       |  61 ++++
 5 files changed, 651 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/arm/cci.txt
 create mode 100644 drivers/bus/arm-cci.c
 create mode 100644 include/linux/arm-cci.h

[...]

Quoted 68 lines
drivers/bus/arm-cci.c new file +62 -0
new file mode 100644
index 0000000..000102f
@@ -0,0 +1,420 @@
+/*
+ * CCI cache coherent interconnect driver
+ *
+ * Copyright (C) 2013 ARM Ltd.
+ * Author: Lorenzo Pieralisi <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed "as is" WITHOUT ANY WARRANTY of any
+ * kind, whether express or implied; without even the implied warranty
+ * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#include <linux/arm-cci.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/of_address.h>
+#include <linux/slab.h>
+
+#include <asm/cacheflush.h>
+#include <asm/smp_plat.h>
+
+#define DRIVER_NAME            "CCI"
+
+#define CCI_PORT_CTRL          0x0
+#define CCI_CTRL_STATUS                0xc
+
+#define CCI_ENABLE_SNOOP_REQ   0x1
+#define CCI_ENABLE_DVM_REQ     0x2
+#define CCI_ENABLE_REQ         (CCI_ENABLE_SNOOP_REQ | CCI_ENABLE_DVM_REQ)
+
+struct cci_nb_ports {
+       unsigned int nb_ace;
+       unsigned int nb_ace_lite;
+};
+
+enum cci_ace_port_type {
+       ACE_INVALID_PORT = 0x0,
+       ACE_PORT,
+       ACE_LITE_PORT,
+};
+
+struct cci_ace_port {
+       void __iomem *base;
+       enum cci_ace_port_type type;
+       struct device_node *dn;
+};
+
+static struct cci_ace_port *ports;
+static unsigned int nb_cci_ports;
+
+static void __iomem *cci_ctrl_base;
+
+struct cpu_port {
+       u64 mpidr;
+       u32 port;
+};
+/*
+ * Use the port MSB as valid flag, shift can be made dynamic

drivers/bus/arm-cci.c line 72

Quoted 10 lines
drivers/bus/arm-cci.c new file +10 -0
+ * by computing number of bits required for port indexes.
+ * Code disabling CCI cpu ports runs with D-cache invalidated
+ * and SCTLR bit clear so data accesses must be kept to a minimum
+ * to improve performance; for now shift is left static to
+ * avoid one more data access while disabling the CCI port.
+ */
+#define PORT_VALID_SHIFT       31
+#define PORT_VALID             (0x1 << PORT_VALID_SHIFT)
+
+static inline void init_cpu_port(struct cpu_port *port, u32 index, u32 mpidr)

The mpidr should be u64.

drivers/bus/arm-cci.c line 83

Quoted 11 lines
drivers/bus/arm-cci.c new file +11 -0
+{
+       port->port = PORT_VALID | index;
+       port->mpidr = mpidr;
+}
+
+static inline bool cpu_port_is_valid(struct cpu_port *port)
+{
+       return !!(port->port & PORT_VALID);
+}
+
+static inline bool cpu_port_match(struct cpu_port *port, u32 mpidr)

Same here.

drivers/bus/arm-cci.c line 87

Quoted 4 lines
drivers/bus/arm-cci.c new file +4 -0
+{
+       return port->mpidr == (mpidr & MPIDR_HWID_BITMASK);
+}
+

Cheers,
Javi

#3 lore raw

RFC PATCH v3 1/1 drivers: bus: add ARM CCI support

Quoted 5 lines · Javi Merino
Quoted 1 line · Lorenzo Pieralisi

+static inline void init_cpu_port(struct cpu_port *port, u32 index, u32 mpidr)

The mpidr should be u64.

Why?

Nicolas

#4 lore raw

RFC PATCH v3 1/1 drivers: bus: add ARM CCI support

Quoted 9 lines · Nicolas Pitre
Quoted 5 lines · Javi Merino
Quoted 1 line · Lorenzo Pieralisi

+static inline void init_cpu_port(struct cpu_port *port, u32 index, u32 mpidr)

The mpidr should be u64.

Why?

Since on arm64 MPIDR_EL1 size is 64 bits, we need the driver to work on both arm and arm64.

Lorenzo

#5 lore raw

RFC PATCH v3 1/1 drivers: bus: add ARM CCI support

Quoted 5 lines · Lorenzo Pieralisi
Quoted 4 lines · Nicolas Pitre
Quoted 2 lines · Javi Merino

On Thu, May 09, 2013 at 11:34:00AM +0100, Lorenzo Pieralisi wrote:

Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
Quoted 1 line

+static inline void init_cpu_port(struct cpu_port *port, u32 index, u32 mpidr)

The mpidr should be u64.

Why?

Since on arm64 MPIDR_EL1 size is 64 bits, we need the driver to work on both arm and arm64.

OK.

Are you OK for me to carry this patch in the series for MCPM support on RTSM? I would like to send it to arm-soc ASAP.

Nicolas

#6 lore raw

RFC PATCH v3 1/1 drivers: bus: add ARM CCI support

Quoted 13 lines · Nicolas Pitre
Quoted 11 lines · Lorenzo Pieralisi
Quoted 9 lines · Nicolas Pitre
Quoted 5 lines · Javi Merino
Quoted 1 line · Lorenzo Pieralisi

+static inline void init_cpu_port(struct cpu_port *port, u32 index, u32 mpidr)

The mpidr should be u64.

Why?

Quoted 7 lines
Quoted 2 lines

Since on arm64 MPIDR_EL1 size is 64 bits, we need the driver to work on both arm and arm64.

OK.

Are you OK for me to carry this patch in the series for MCPM support on RTSM? I would like to send it to arm-soc ASAP.

Absolutely, I will apply the latest fixes and send the updated version to you for arm-soc inclusion today.

Lorenzo

#7 lore raw

RFC PATCH v3 1/1 drivers: bus: add ARM CCI support

In reply to Lorenzo Pieralisi

On Thu, 2013-05-09 at 11:34 +0100, Lorenzo Pieralisi wrote:
[...]

Quoted 25 lines
drivers/bus/arm-cci.c new file +25 -0
+static int __init cci_probe(void)
+{
+	struct cci_nb_ports const *cci_config;
+	int ret, i, nb_ace = 0, nb_ace_lite = 0;
+	struct device_node *np, *cp;
+	const char *match_str;
+	bool is_ace;
+
+	np = of_find_matching_node(NULL, arm_cci_matches);
+	if (!np)
+		return -ENODEV;
+
+	cci_config = of_match_node(arm_cci_matches, np)->data;
+	if (!cci_config)
+		return -ENODEV;
+
+	nb_cci_ports = cci_config->nb_ace + cci_config->nb_ace_lite;
+
+	ports = kcalloc(sizeof(*ports), nb_cci_ports, GFP_KERNEL);
+	if (!ports)
+		return -ENOMEM;
+
+	cci_ctrl_base = of_iomap(np, 0);
+
+	if (!cci_ctrl_base) {

drivers/bus/arm-cci.c line 335

Quoted 10 lines
drivers/bus/arm-cci.c new file +10 -0
+		WARN(1, "unable to ioremap CCI ctrl\n");
+		ret = -ENXIO;
+		goto memalloc_err;
+	}
+
+	for_each_child_of_node(np, cp) {
+		i = nb_ace + nb_ace_lite;
+
+		if (i >= nb_cci_ports)
+			break;

I know this is a bit late, but...

Would it not be best to check here that the node type is in fact "slave-if", that way if we add support in later versions for other node types (e.g. for the PMU) then we don't cause backward compatibility issues. I'm thinking here of the same concerns that Rob raised with "ARM: kernel: fix arm_dt_init_cpu_maps() to skip non-cpu nodes"

Quoted 7 lines
drivers/bus/arm-cci.c new file +7 -0
+		if (of_property_read_string(cp, "interface-type",
+					&match_str)) {
+			WARN(1, "node %s missing interface-type property\n",
+				  cp->full_name);
+			continue;
+		}
+

[..]

Signature · 1 line
Tixy
#8 lore raw

RFC PATCH v3 1/1 drivers: bus: add ARM CCI support

Quoted 40 lines · Jon Medhurst (Tixy)

On Thu, 2013-05-09 at 11:34 +0100, Lorenzo Pieralisi wrote:
[...]

Quoted 25 lines
drivers/bus/arm-cci.c new file +25 -0
+static int __init cci_probe(void)
+{
+	struct cci_nb_ports const *cci_config;
+	int ret, i, nb_ace = 0, nb_ace_lite = 0;
+	struct device_node *np, *cp;
+	const char *match_str;
+	bool is_ace;
+
+	np = of_find_matching_node(NULL, arm_cci_matches);
+	if (!np)
+		return -ENODEV;
+
+	cci_config = of_match_node(arm_cci_matches, np)->data;
+	if (!cci_config)
+		return -ENODEV;
+
+	nb_cci_ports = cci_config->nb_ace + cci_config->nb_ace_lite;
+
+	ports = kcalloc(sizeof(*ports), nb_cci_ports, GFP_KERNEL);
+	if (!ports)
+		return -ENOMEM;
+
+	cci_ctrl_base = of_iomap(np, 0);
+
+	if (!cci_ctrl_base) {

drivers/bus/arm-cci.c line 335

Quoted 10 lines
drivers/bus/arm-cci.c new file +10 -0
+		WARN(1, "unable to ioremap CCI ctrl\n");
+		ret = -ENXIO;
+		goto memalloc_err;
+	}
+
+	for_each_child_of_node(np, cp) {
+		i = nb_ace + nb_ace_lite;
+
+		if (i >= nb_cci_ports)
+			break;

I know this is a bit late, but...

Quoted 5 lines

Would it not be best to check here that the node type is in fact "slave-if", that way if we add support in later versions for other node types (e.g. for the PMU) then we don't cause backward compatibility issues. I'm thinking here of the same concerns that Rob raised with "ARM: kernel: fix arm_dt_init_cpu_maps() to skip non-cpu nodes"

The node name, not type, but point is taken. I prefer to add a compatible property to slave-if nodes (eg "arm,cci-400-control-if"), I do not think that checking the node name is the standard way of doing things in DT world.

Thoughts ?

Thanks,
Lorenzo

Quoted 13 lines
Quoted 7 lines
drivers/bus/arm-cci.c new file +7 -0
+		if (of_property_read_string(cp, "interface-type",
+					&match_str)) {
+			WARN(1, "node %s missing interface-type property\n",
+				  cp->full_name);
+			continue;
+		}
+

[..]

Signature · 1 line
Tixy
#9 lore raw

RFC PATCH v3 1/1 drivers: bus: add ARM CCI support

Quoted 7 lines · Lorenzo Pieralisi

The node name, not type, but point is taken. I prefer to add a compatible property to slave-if nodes (eg "arm,cci-400-control-if"), I do not think that checking the node name is the standard way of doing
things in DT world.

Thoughts ?

Agreed.

Cheers,
Ben.