Soon to be released versions of RPi4's firmware will take of care
passing their bootloader's configuration[1] to the OS by copying it into
memory and creating a reserved memory node in the board's DT. By
modeling this reserved memory node as an nvmem device using
'nvmem-rmem', which this series introduces, user-space applications will
be able to query this information through nvmem's sysfs interface.
An alternative approach, less nice IMO, would be to create a
platform-specific 'soc' driver.
Regards,
Nicolas
[1] https://www.raspberrypi.org/documentation/hardware/raspberrypi/bcm2711_bootloader_config.md
---
Changes since v1:
- Avoid the reserved-memory indirection by integrating the nvmem driver
into the reserved memory node.
Nicolas Saenz Julienne (5):
dt-bindings: nvmem: Add bindings for rmem driver
nvmem: Add driver to expose reserved memory as nvmem
ARM: dts: bcm2711: Add reserved memory template to hold firmware
configuration
arm64: defconfig: Enable nvmem's rmem driver
ARM: multi_v7_defconfig: Enable nvmem's rmem driver
.../devicetree/bindings/nvmem/rmem.yaml | 49 ++++++++++
arch/arm/boot/dts/bcm2711-rpi-4-b.dts | 17 ++++
arch/arm/configs/multi_v7_defconfig | 1 +
arch/arm64/configs/defconfig | 1 +
drivers/nvmem/Kconfig | 8 ++
drivers/nvmem/Makefile | 2 +
drivers/nvmem/rmem.c | 97 +++++++++++++++++++
drivers/of/platform.c | 1 +
8 files changed, 176 insertions(+)
create mode 100644 Documentation/devicetree/bindings/nvmem/rmem.yaml
create mode 100644 drivers/nvmem/rmem.c
--
2.29.2
Firmware/co-processors might use reserved memory areas in order to pass
data stemming from an nvmem device otherwise non accessible to Linux.
For example an EEPROM memory only physically accessible to firmware, or
data only accessible early at boot time.
Introduce the dt-bindings to nvmem's rmem.
Signed-off-by: Nicolas Saenz Julienne <[email protected]>
---
Changes since v1:
- Update schema to new driver design
.../devicetree/bindings/nvmem/rmem.yaml | 49 +++++++++++++++++++
1 file changed, 49 insertions(+)
create mode 100644 Documentation/devicetree/bindings/nvmem/rmem.yaml
diff --git a/Documentation/devicetree/bindings/nvmem/rmem.yaml b/Documentation/devicetree/bindings/nvmem/rmem.yaml
new file mode 100644
index 000000000000..29b53871aa02
--- /dev/null
+++ b/Documentation/devicetree/bindings/nvmem/rmem.yaml
@@ -0,0 +1,49 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/nvmem/rmem.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Reserved Memory Based nvmem Device
+
+maintainers:
+ - Nicolas Saenz Julienne <[email protected]>
+
+allOf:
+ - $ref: "nvmem.yaml#"
+
+properties:
+ compatible:
+ items:
+ - enum:
+ - raspberrypi,bootloader-config
+ - const: nvmem-rmem
+
+ no-map:
+ $ref: /schemas/types.yaml#/definitions/flag
+ description:
+ Avoid creating a virtual mapping of the region as part of the OS'
+ standard mapping of system memory.
+
+required:
+ - compatible
+ - no-map
+
+unevaluatedProperties: false
+
+examples:
+ - |
+ reserved-memory {
+ #address-cells = <1>;
+ #size-cells = <1>;
+
+ blconfig: nvram@10000000 {
+ compatible = "raspberrypi,bootloader-config", "nvmem-rmem";
+ #address-cells = <1>;
+ #size-cells = <1>;
+ reg = <0x10000000 0x1000>;
+ no-map;
+ };
+ };
+
+...
--
2.29.2
It'll be used by the RPi4 family of boards to access its bootloader
configuration.
Signed-off-by: Nicolas Saenz Julienne <[email protected]>
---
arch/arm/configs/multi_v7_defconfig | 1 +
1 file changed, 1 insertion(+)
diff --git a/arch/arm/configs/multi_v7_defconfig b/arch/arm/configs/multi_v7_defconfig
index c5f25710fedc..7a326c5eff7a 100644
--- a/arch/arm/configs/multi_v7_defconfig
+++ b/arch/arm/configs/multi_v7_defconfig
@@ -1107,6 +1107,7 @@ CONFIG_ROCKCHIP_EFUSE=m
CONFIG_NVMEM_SUNXI_SID=y
CONFIG_NVMEM_VF610_OCOTP=y
CONFIG_MESON_MX_EFUSE=m
+CONFIG_NVMEM_RMEM=m
CONFIG_FSI=m
CONFIG_FSI_MASTER_GPIO=m
CONFIG_FSI_MASTER_HUB=m
--
2.29.2
Firmware/co-processors might use reserved memory areas in order to pass
data stemming from an nvmem device otherwise non accessible to Linux.
For example an EEPROM memory only physically accessible to firmware, or
data only accessible early at boot time.
In order to expose this data to other drivers and user-space, the driver
models the reserved memory area as an nvmem device.
Signed-off-by: Nicolas Saenz Julienne <[email protected]>
---
Changes since v1:
- Remove reserved memory phandle indirection by directly creating a
platform device from the reserved memory DT node
- Only map memory upon reading it to avoid corruption
- Small cosmetic cleanups
drivers/nvmem/Kconfig | 8 ++++
drivers/nvmem/Makefile | 2 +
drivers/nvmem/rmem.c | 97 ++++++++++++++++++++++++++++++++++++++++++
drivers/of/platform.c | 1 +
4 files changed, 108 insertions(+)
create mode 100644 drivers/nvmem/rmem.c
diff --git a/drivers/nvmem/Kconfig b/drivers/nvmem/Kconfig
index 954d3b4a52ab..fecc19b884bf 100644
--- a/drivers/nvmem/Kconfig
+++ b/drivers/nvmem/Kconfig
@@ -270,4 +270,12 @@ config SPRD_EFUSE
This driver can also be built as a module. If so, the module
will be called nvmem-sprd-efuse.
+config NVMEM_RMEM
+ tristate "Reserved Memory Based Driver Support"
+ help
+ This drivers maps reserved memory into an nvmem device. It might be
+ useful to expose information left by firmware in memory.
+
+ This driver can also be built as a module. If so, the module
+ will be called nvmem-rmem.
endif
diff --git a/drivers/nvmem/Makefile b/drivers/nvmem/Makefile
index a7c377218341..5376b8e0dae5 100644
--- a/drivers/nvmem/Makefile
+++ b/drivers/nvmem/Makefile
@@ -55,3 +55,5 @@ obj-$(CONFIG_NVMEM_ZYNQMP) += nvmem_zynqmp_nvmem.o
nvmem_zynqmp_nvmem-y := zynqmp_nvmem.o
obj-$(CONFIG_SPRD_EFUSE) += nvmem_sprd_efuse.o
nvmem_sprd_efuse-y := sprd-efuse.o
+obj-$(CONFIG_NVMEM_RMEM) += nvmem-rmem.o
+nvmem-rmem-y := rmem.o
diff --git a/drivers/nvmem/rmem.c b/drivers/nvmem/rmem.c
new file mode 100644
index 000000000000..b11c3c974b3d
--- /dev/null
+++ b/drivers/nvmem/rmem.c
@@ -0,0 +1,97 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Copyright (C) 2020 Nicolas Saenz Julienne <[email protected]>
+ */
+
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/nvmem-provider.h>
+#include <linux/of_reserved_mem.h>
+#include <linux/platform_device.h>
+
+struct rmem {
+ struct device *dev;
+ struct nvmem_device *nvmem;
+ struct reserved_mem *mem;
+
+ phys_addr_t size;
+};
+
+static int rmem_read(void *context, unsigned int offset,
+ void *val, size_t bytes)
+{
+ struct rmem *priv = context;
+ size_t available = priv->mem->size;
+ loff_t off = offset;
+ void *addr;
+ int count;
+
+ /*
+ * Only map the reserved memory at this point to avoid potential rogue
+ * kernel threads inadvertently modifying it. Based on the current
+ * uses-cases for this driver, the performance hit isn't a concern.
+ * Nor is likely to be, given the nature of the subsystem. Most nvmem
+ * devices operate over slow buses to begin with.
+ *
+ * An alternative would be setting the memory as RO, set_memory_ro(),
+ * but as of Dec 2020 this isn't possible on arm64.
+ */
+ addr = memremap(priv->mem->base, available, MEMREMAP_WB);
+ if (IS_ERR(addr)) {
+ dev_err(priv->dev, "Failed to remap memory region\n");
+ return PTR_ERR(addr);
+ }
+
+ count = memory_read_from_buffer(val, bytes, &off, addr, available);
+
+ memunmap(addr);
+
+ return count;
+}
+
+static int rmem_probe(struct platform_device *pdev)
+{
+ struct nvmem_config config = { };
+ struct device *dev = &pdev->dev;
+ struct reserved_mem *mem;
+ struct rmem *priv;
+
+ priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL);
+ if (!priv)
+ return -ENOMEM;
+ priv->dev = dev;
+
+ mem = of_reserved_mem_lookup(dev->of_node);
+ if (!mem) {
+ dev_err(dev, "Failed to lookup reserved memory\n");
+ return -EINVAL;
+ }
+ priv->mem = mem;
+
+ config.dev = dev;
+ config.priv = priv;
+ config.name = "rmem";
+ config.size = mem->size;
+ config.reg_read = rmem_read;
+
+ return PTR_ERR_OR_ZERO(devm_nvmem_register(dev, &config));
+}
+
+static const struct of_device_id rmem_match[] = {
+ { .compatible = "nvmem-rmem", },
+ { /* sentinel */ },
+};
+MODULE_DEVICE_TABLE(of, rmem_match);
+
+static struct platform_driver rmem_driver = {
+ .probe = rmem_probe,
+ .driver = {
+ .name = "rmem",
+ .of_match_table = rmem_match,
+ },
+};
+module_platform_driver(rmem_driver);
+
+MODULE_AUTHOR("Nicolas Saenz Julienne <[email protected]>");
+MODULE_DESCRIPTION("Reserved Memory Based nvmem Driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/of/platform.c b/drivers/of/platform.c
index 79bd5f5a1bf1..6699cdbe58b6 100644
--- a/drivers/of/platform.c
+++ b/drivers/of/platform.c
@@ -511,6 +511,7 @@ static const struct of_device_id reserved_mem_matches[] = {
{ .compatible = "qcom,rmtfs-mem" },
{ .compatible = "qcom,cmd-db" },
{ .compatible = "ramoops" },
+ { .compatible = "nvmem-rmem" },
{}
};
--
2.29.2
On Fri, 18 Dec 2020 16:43:16 +0100, Nicolas Saenz Julienne wrote:
> Firmware/co-processors might use reserved memory areas in order to pass
> data stemming from an nvmem device otherwise non accessible to Linux.
> For example an EEPROM memory only physically accessible to firmware, or
> data only accessible early at boot time.
>
> Introduce the dt-bindings to nvmem's rmem.
>
> Signed-off-by: Nicolas Saenz Julienne <[email protected]>
>
> ---
>
> Changes since v1:
> - Update schema to new driver design
>
> .../devicetree/bindings/nvmem/rmem.yaml | 49 +++++++++++++++++++
> 1 file changed, 49 insertions(+)
> create mode 100644 Documentation/devicetree/bindings/nvmem/rmem.yaml
>
My bot found errors running 'make dt_binding_check' on your patch:
yamllint warnings/errors:
./Documentation/devicetree/bindings/nvmem/rmem.yaml:19:9: [warning] wrong indentation: expected 10 but found 8 (indentation)
dtschema/dtc warnings/errors:
See https://patchwork.ozlabs.org/patch/1418470
This check can fail if there are any dependencies. The base for a patch
series is generally the most recent rc1.
If you already ran 'make dt_binding_check' and didn't see the above
error(s), then make sure 'yamllint' is installed and dt-schema is up to
date:
pip3 install dtschema --upgrade
Please check and re-submit.
On Fri, 18 Dec 2020 16:43:17 +0100, Nicolas Saenz Julienne wrote:
> Firmware/co-processors might use reserved memory areas in order to pass
> data stemming from an nvmem device otherwise non accessible to Linux.
> For example an EEPROM memory only physically accessible to firmware, or
> data only accessible early at boot time.
>
> In order to expose this data to other drivers and user-space, the driver
> models the reserved memory area as an nvmem device.
>
> Signed-off-by: Nicolas Saenz Julienne <[email protected]>
>
> ---
>
> Changes since v1:
> - Remove reserved memory phandle indirection by directly creating a
> platform device from the reserved memory DT node
> - Only map memory upon reading it to avoid corruption
> - Small cosmetic cleanups
>
> drivers/nvmem/Kconfig | 8 ++++
> drivers/nvmem/Makefile | 2 +
> drivers/nvmem/rmem.c | 97 ++++++++++++++++++++++++++++++++++++++++++
> drivers/of/platform.c | 1 +
> 4 files changed, 108 insertions(+)
> create mode 100644 drivers/nvmem/rmem.c
>
Reviewed-by: Rob Herring <[email protected]>