iommu/arm-smmu: Add support for adding masters/clocks using generic bindings

This series adds support for xlate callback to add master devices configs using generic bindings and clocks/regulators required to access smmu.

OF_IOMMU_DECLARE is used to register and probe the smmu controller devices before the masters are added in of_platform_populate. Here, we are registering only the smmu driver and iommu_ops. The smmu device registration is not done in OF_IOMMU_DECLARE to avoid early smmu probe, because the clocks if any required for the controller are not yet ready.

So the idea here is to get the right direction in adding this support.

Mitchel Humpherys (2):
iommu/arm-smmu: add support for specifying clocks iommu/arm-smmu: Add support for specifying regulators

Sricharan R (2):
iommu/arm-smmu: Init driver using IOMMU_OF_DECLARE iommu/arm-smmu: Add xlate callback for initializing master devices
from dt

 .../devicetree/bindings/iommu/arm,smmu.txt         |  14 ++
 drivers/iommu/arm-smmu.c                           | 210 +++++++++++++++++++--
 2 files changed, 208 insertions(+), 16 deletions(-)

QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation

This patch uses IOMMU_OF_DECLARE to register the driver and the iommu_ops. So when master devices of the iommu are registered, of_xlate callback can be used to add the master configurations to the smmu driver.

---
 drivers/iommu/arm-smmu.c | 36 +++++++++++++++++++++++-------------
 1 file changed, 23 insertions(+), 13 deletions(-)

index fc13dd5..8c4eb43 100644
@@ -41,6 +41,8 @@
 #include <linux/platform_device.h>
 #include <linux/slab.h>
 #include <linux/spinlock.h>
+#include <linux/of_iommu.h>
+#include <linux/of_platform.h>
 
 #include <linux/amba/bus.h>
 

@@ -254,6 +256,8 @@
 #define FSYNR0_WNR			(1 << 4)
 
 static int force_stage;
+static bool init_done;
+
 module_param_named(force_stage, force_stage, int, S_IRUGO | S_IWUSR);
 MODULE_PARM_DESC(force_stage,
 	"Force SMMU mappings to be installed@a particular stage of translation. A value of '1' or '2' forces the corresponding stage. All other values are ignored (i.e. no stage is forced). Note that selecting a specific stage will disable support for nested translation.");

@@ -1848,20 +1852,8 @@ static struct platform_driver arm_smmu_driver = {
 
 static int __init arm_smmu_init(void)
 {
-	struct device_node *np;
 	int ret;
 
-	/*
-	 * Play nice with systems that don't have an ARM SMMU by checking that
-	 * an ARM SMMU exists in the system before proceeding with the driver
-	 * and IOMMU bus operation registration.
-	 */
-	np = of_find_matching_node(NULL, arm_smmu_of_match);
-	if (!np)
-		return 0;
-
-	of_node_put(np);
-
 	ret = platform_driver_register(&arm_smmu_driver);
 	if (ret)
 		return ret;

@@ -1880,15 +1872,33 @@ static int __init arm_smmu_init(void)
 		bus_set_iommu(&pci_bus_type, &arm_smmu_ops);
 #endif
 
+	init_done = true;
+
 	return 0;
 }
 
+static int __init arm_smmu_of_setup(struct device_node *np)
+{
+
+	if (!init_done)
+		arm_smmu_init();
+
+	of_iommu_set_ops(np, &arm_smmu_ops);
+
+	return 0;
+}
+
+IOMMU_OF_DECLARE(arm_smmu_v1, "arm,smmu-v1", arm_smmu_of_setup);
+IOMMU_OF_DECLARE(arm_smmu_v2, "arm,smmu-v2", arm_smmu_of_setup);
+IOMMU_OF_DECLARE(arm_smmu_400, "arm,mmu-400", arm_smmu_of_setup);
+IOMMU_OF_DECLARE(arm_smmu_401, "arm,mmu-401", arm_smmu_of_setup);
+IOMMU_OF_DECLARE(arm_smmu_500, "arm,mmu-500", arm_smmu_of_setup);
+
 static void __exit arm_smmu_exit(void)
 {
 	return platform_driver_unregister(&arm_smmu_driver);
 }
 
-subsys_initcall(arm_smmu_init);
 module_exit(arm_smmu_exit);
 
 MODULE_DESCRIPTION("IOMMU API for ARM architected SMMU implementations");

QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation

[adding Robin]

I'd really prefer to do this on top of Laurent's series reworking some of the of_xlate code and I think Robin [CC'd] is working on that. That will also pave the way for specifying per-master SMR configuration in the device-tree.

Will

Hi Will,

iommu at lists.linux-foundation.org;

the

pave

Ok, understand it. By Rebasing on Laurent's series this patch should
not
be needed at all, dependencies with clock devices should also be
solved.
I will rebase this series over there.

Regards,
Sricharan

This adds of_xlate callback to arm-smmu driver. xlate callback is called during device registration from DT for those master devices attached to iommus using generic iommu bindings.

---
 drivers/iommu/arm-smmu.c | 36 ++++++++++++++++++++++++++++++++++++
 1 file changed, 36 insertions(+)

index 8c4eb43..2cf65ab 100644
@@ -321,6 +321,7 @@ struct arm_smmu_device {
 	unsigned int			*irqs;
 
 	struct list_head		list;
+	struct device_node		*node;
 	struct rb_root			masters;
 };
 

@@ -1434,6 +1435,38 @@ out_unlock:
 	return ret;
 }
 
+static int arm_smmu_of_xlate(struct device *dev,
+			     struct of_phandle_args *spec)
+{
+	struct arm_smmu_device *smmu;
+	struct arm_smmu_master *master;
+	int streamid;
+
+	spin_lock(&arm_smmu_devices_lock);
+	list_for_each_entry(smmu, &arm_smmu_devices, list) {
+		if (smmu->node == spec->np)
+			break;
+	}
+	spin_unlock(&arm_smmu_devices_lock);
+
+	if (!smmu || (smmu->node != spec->np))
+		return -ENODEV;
+
+	spec->np = dev->of_node;
+
+	master = find_smmu_master(smmu, spec->np);
+	if (!master) {
+		if (register_smmu_master(smmu, smmu->dev, spec))
+			return -ENODEV;
+	} else {
+		streamid = master->cfg.num_streamids;
+		master->cfg.streamids[streamid] = spec->args[0];
+		master->cfg.num_streamids++;
+	}
+
+	return 0;
+}
+
 static struct iommu_ops arm_smmu_ops = {
 	.capable		= arm_smmu_capable,
 	.domain_init		= arm_smmu_domain_init,

@@ -1449,6 +1482,7 @@ static struct iommu_ops arm_smmu_ops = {
 	.domain_get_attr	= arm_smmu_domain_get_attr,
 	.domain_set_attr	= arm_smmu_domain_set_attr,
 	.pgsize_bitmap		= -1UL, /* Restricted during device attach */
+	.of_xlate		= arm_smmu_of_xlate,
 };
 
 static void arm_smmu_device_reset(struct arm_smmu_device *smmu)

@@ -1782,6 +1816,8 @@ static int arm_smmu_device_dt_probe(struct platform_device *pdev)
 		}
 	}
 
+	smmu->node = dev->of_node;
+
 	INIT_LIST_HEAD(&smmu->list);
 	spin_lock(&arm_smmu_devices_lock);
 	list_add(&smmu->list, &arm_smmu_devices);

QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation

From: Mitchel Humpherys <[email protected]>

On some platforms with tight power constraints it is polite to only leave your clocks on for as long as you absolutely need them. Currently we assume that all clocks necessary for SMMU register access are always on.

Add some optional device tree properties to specify any clocks that are necessary for SMMU register access and turn them on and off as needed.

If no clocks are specified in the device tree things continue to work the way they always have: we assume all necessary clocks are always turned on. The clocks are turned 'on' during the SMMU device probe and remains 'on' till the device exists.

---
 .../devicetree/bindings/iommu/arm,smmu.txt         | 11 +++
 drivers/iommu/arm-smmu.c                           | 86 +++++++++++++++++++++-
 2 files changed, 94 insertions(+), 3 deletions(-)

index 0676050..c2cf4fe 100644
@@ -49,6 +49,17 @@ conditions.
                   aliases of secure registers have to be used during
                   SMMU configuration.
 
+- clocks	: List of clocks to be used during SMMU register access. See
+		  Documentation/devicetree/bindings/clock/clock-bindings.txt
+		  for information about the format. For each clock specified
+		  here, there must be a corresponding entery in clock-names
+		  (see below).
+
+- clock-names	: List of clock names corresponding to the clocks specified in
+		  the "clocks" property (above). See
+		  Documentation/devicetree/bindings/clock/clock-bindings.txt
+		  for more info.
+
 Example:
 
         smmu {

QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation

You've borrowed this commit message from Mitch's previous version of this patch, but now you leave the clocks enabled most of the time so it doesn't make much sense anymore.

Anyway, I'm OK with this kind of clock management in the driver, but I think that anything more fine-grained needs to be designed into the IOMMU core.

Will

Hi Will,

iommu at lists.linux-foundation.org;

Sorry, I should have changed that to make it clear.

think

Ok. Tried this to get the right direction.
I will check for the power savings and if there are not much then I would
use
the above approach.

Regards,
Sricharan

From: Mitchel Humpherys <[email protected]>

This adds the support to turn on the regulators required for SMMUs. It is turned on during the SMMU probe and remains 'on' till the device exists.

---
 .../devicetree/bindings/iommu/arm,smmu.txt         |  3 ++
 drivers/iommu/arm-smmu.c                           | 48 +++++++++++++++++++++-
 2 files changed, 50 insertions(+), 1 deletion(-)

index c2cf4fe..433d778 100644
@@ -60,6 +60,9 @@ conditions.
 		  Documentation/devicetree/bindings/clock/clock-bindings.txt
 		  for more info.
 
+- vdd-supply    : Phandle of the regulator that should be powered on during
+                  SMMU register access.
+
 Example:
 
         smmu {

QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation

The device always exists. Until the driver is removed perhaps?

We won't be using regulators. We'll be using power domains instead so please rewrite this patch accordingly.

Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project

Hi Stephen,

Till the device exists.

Ok. Thanks for the direction.

Regards,
Sricharan