Skip to contentSkip to conversation

remoteproc: qcom: Add PAS based SPSS PIL driver

PATCH In review Rishabh Bhatnagar 3 participants 5 messages Mar 21, 2020 – Mar 25, 2020

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/2 remoteproc: qcom: Add PAS based SPSS PIL driver

This adds Peripheral Authentication System (PAS) based SPSS PIL driver for loading SPSS firmware and booting. The firmware is verified and booted with the help of the PAS in TrustZone. This driver also adds functionality to handle late attach feature for a remote processor. The driver queries the initial state of remote processor and is capable of handling successful bootup and crash scenarios.

---
 drivers/remoteproc/Kconfig     |  17 ++
 drivers/remoteproc/Makefile    |   1 +
 drivers/remoteproc/qcom_spss.c | 500 +++++++++++++++++++++++++++++++++++++++++
 3 files changed, 518 insertions(+)
 create mode 100644 drivers/remoteproc/qcom_spss.c
drivers/remoteproc/Kconfig +17 -0
index de3862c..ab3e89a 100644
@@ -151,6 +151,23 @@ config QCOM_Q6V5_PAS
 	  for the Qualcomm Hexagon v5 based remote processors. This is commonly
 	  used to control subsystems such as ADSP, Compute and Sensor.
 
+config QCOM_SPSS
+	tristate "Qualcomm secure subsystem Peripheral loading support"
+	depends on OF && ARCH_QCOM
+	depends on QCOM_SMEM
+	depends on RPMSG_QCOM_SMD || (COMPILE_TEST && RPMSG_QCOM_SMD=n)
+	depends on RPMSG_QCOM_GLINK_SMEM || RPMSG_QCOM_GLINK_SMEM=n
+	depends on QCOM_SYSMON || QCOM_SYSMON=n
+	select MFD_SYSCON
+	select QCOM_MDT_LOADER
+	select QCOM_Q6V5_COMMON
+	select QCOM_RPROC_COMMON
+	select QCOM_SCM
+	help
+	  Say y here to support the TrustZone based Peripheral Image Loader
+	  for the Qualcomm secure subsystem remote processor. This also supports
+	  remote processors that are booted before kernel comes up.
+
 config QCOM_Q6V5_WCSS
 	tristate "Qualcomm Hexagon based WCSS Peripheral Image Loader"
 	depends on OF && ARCH_QCOM
drivers/remoteproc/Makefile +1 -0
index e30a1b1..e781552 100644
@@ -15,6 +15,7 @@ obj-$(CONFIG_OMAP_REMOTEPROC)		+= omap_remoteproc.o
 obj-$(CONFIG_WKUP_M3_RPROC)		+= wkup_m3_rproc.o
 obj-$(CONFIG_DA8XX_REMOTEPROC)		+= da8xx_remoteproc.o
 obj-$(CONFIG_KEYSTONE_REMOTEPROC)	+= keystone_remoteproc.o
+obj-$(CONFIG_QCOM_SPSS)			+= qcom_spss.o
 obj-$(CONFIG_QCOM_RPROC_COMMON)		+= qcom_common.o
 obj-$(CONFIG_QCOM_Q6V5_COMMON)		+= qcom_q6v5.o
 obj-$(CONFIG_QCOM_Q6V5_ADSP)		+= qcom_q6v5_adsp.o
drivers/remoteproc/qcom_spss.c new file +500 -0

500 lines not on this page · raw message

Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#1 lore raw

PATCH v2 2/2 dt-bindings: remoteproc: Add documentation for SPSS remoteproc

Add devicetree binding for Secure Subsystem remote processor support in remoteproc framework. This describes all the resources needed by SPSS to boot and handle crash and shutdown scenarios.

---
 .../devicetree/bindings/remoteproc/qcom,spss.yaml  | 125 +++++++++++++++++++++
 1 file changed, 125 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml
Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml new file +125 -0
new file mode 100644
index 0000000..9ca7947a9
@@ -0,0 +1,125 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/qcom,spss.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Qualcomm SPSS Peripheral Image Loader
+
+maintainers:
+  - Rishabh Bhatnagar <[email protected]>
+description: |
+  This document defines the binding for a component that loads and boots firmware
+  on the Qualcomm Secure Peripheral Processor. This processor is booted in the
+  bootloader stage and it attaches itself to linux later on in the boot process.
+
+properties:
+  compatible:
+    enum:
+      "qcom,sm8250-spss-pas"
+
+  reg:
+    items:
+      - description: IRQ status register
+      - description: IRQ clear register
+      - description: IRQ mask register
+      - description: Error register
+      - description: Error spare register
+
+  reg-names:
+    items:
+      - const: sp2soc_irq_status
+      - const: sp2soc_irq_clr
+      - const: sp2soc_irq_mask
+      - const: rmb_err
+      - const: rmb_err_spare2
+
+  interrupts:
+    maxitems: 1
+    items:
+      - description: rx interrupt
+
+  clocks:
+    items:
+      - description:
+                    reference to the xo clock to be held on behalf
+                    of the booting Hexagon core
+
+  clock-names:
+    items:
+      - const: xo
+
+  cx-supply: true
+
+  px-supply: true
+
+  memory-region: true
+    items:
+      - description: reference to the reserved-memory for the SPSS
+
+  qcom,spss-scsr-bits:
+    - description: Bits that are set by remote processor in the irq status
+                   register region to represent different states during
+                   boot process
+
+  child-node:
+    description: Subnode named either "smd-edge" or "glink-edge" that
+                 describes the communication edge, channels and devices
+                 related to the SPSS.
+
+required:
+  - compatible
+  - reg
+  - reg-names
+  - interrupts
+  - clocks
+  - clock-names
+  - cx-supply
+  - px-supply
+  - memory-region
+  - qcom,spss-scsr-bits
+
+
+examples:
+  - |
+    spss {
+        compatible = "qcom,sm8250-spss-pil";
+        reg = <0x188101c 0x4>,
+                <0x1881024 0x4>,
+                <0x1881028 0x4>,
+                <0x188103c 0x4>,
+                <0x1882014 0x4>;
+        reg-names = "sp2soc_irq_status", "sp2soc_irq_clr",
+                    "sp2soc_irq_mask", "rmb_err", "rmb_err_spare2";
+        interrupts = <0 352 1>;
+
+        cx-supply = <&VDD_CX_LEVEL>;
+        cx-uV-uA = <RPMH_REGULATOR_LEVEL_TURBO 100000>;
+        px-supply = <&VDD_MX_LEVEL>;
+        px-uV = <RPMH_REGULATOR_LEVEL_TURBO 100000>;
+
+        clocks = <&clock_rpmh RPMH_CXO_CLK>;
+        clock-names = "xo";
+        qcom,proxy-clock-names = "xo";
+        status = "ok";
+
+        memory-region = <&pil_spss_mem>;
+        qcom,spss-scsr-bits = <24 25>;
+
+        glink-edge {
+                qcom,remote-pid = <8>;
+                transport = "spss";
+                mboxes = <&sp_scsr 0>;
+                mbox-names = "spss_spss";
+                interrupt-parent = <&intsp>;
+                interrupts = <0 0 IRQ_TYPE_LEVEL_HIGH>;
+
+                reg = <0x1885008 0x8>,
+                      <0x1885010 0x4>;
+                reg-names = "qcom,spss-addr",
+                            "qcom,spss-size";
+
+                label = "spss";
+                qcom,glink-label = "spss";
+        };
+    };
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
maintainer
#2 lore raw

Re: [PATCH v2 2/2] dt-bindings: remoteproc: Add documentation for SPSS remoteproc

Quoted 10 lines · Rishabh Bhatnagar

Add devicetree binding for Secure Subsystem remote processor support in remoteproc framework. This describes all the resources needed by SPSS to boot and handle crash and shutdown scenarios.

---
 .../devicetree/bindings/remoteproc/qcom,spss.yaml  | 125 +++++++++++++++++++++
 1 file changed, 125 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml

My bot found errors running 'make dt_binding_check' on your patch:

Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml: mapping values are not allowed in this context
in "<unicode string>", line 57, column 10
Documentation/devicetree/bindings/Makefile:12: recipe for target 'Documentation/devicetree/bindings/remoteproc/qcom,spss.example.dts' failed make[1]: *** [Documentation/devicetree/bindings/remoteproc/qcom,spss.example.dts] Error 1
make[1]: *** Waiting for unfinished jobs....
Makefile:1262: recipe for target 'dt_binding_check' failed
make: *** [dt_binding_check] Error 2

See https://patchwork.ozlabs.org/patch/1259273

If you already ran 'make dt_binding_check' and didn't see the above error(s), then make sure dt-schema is up to date:

pip3 install git+https://github.com/devicetree-org/dt-schema.git@master --upgrade

Please check and re-submit.

#3 lore raw

Re: [PATCH v2 2/2] dt-bindings: remoteproc: Add documentation for SPSS remoteproc

In reply to Rishabh Bhatnagar
Quoted 91 lines · Rishabh Bhatnagar

Add devicetree binding for Secure Subsystem remote processor support in remoteproc framework. This describes all the resources needed by SPSS to boot and handle crash and shutdown scenarios.

---
 .../devicetree/bindings/remoteproc/qcom,spss.yaml  | 125 +++++++++++++++++++++
 1 file changed, 125 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml
Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml new file +75 -0
new file mode 100644
index 0000000..9ca7947a9
@@ -0,0 +1,125 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/qcom,spss.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Qualcomm SPSS Peripheral Image Loader
+
+maintainers:
+  - Rishabh Bhatnagar <[email protected]>
+description: |
+  This document defines the binding for a component that loads and boots firmware
+  on the Qualcomm Secure Peripheral Processor. This processor is booted in the
+  bootloader stage and it attaches itself to linux later on in the boot process.
+
+properties:
+  compatible:
+    enum:
+      "qcom,sm8250-spss-pas"
+
+  reg:
+    items:
+      - description: IRQ status register
+      - description: IRQ clear register
+      - description: IRQ mask register
+      - description: Error register
+      - description: Error spare register
+
+  reg-names:
+    items:
+      - const: sp2soc_irq_status
+      - const: sp2soc_irq_clr
+      - const: sp2soc_irq_mask
+      - const: rmb_err
+      - const: rmb_err_spare2
+
+  interrupts:
+    maxitems: 1
+    items:
+      - description: rx interrupt
+
+  clocks:
+    items:
+      - description:
+                    reference to the xo clock to be held on behalf
+                    of the booting Hexagon core
+
+  clock-names:
+    items:
+      - const: xo
+
+  cx-supply: true
+
+  px-supply: true
+
+  memory-region: true
+    items:
+      - description: reference to the reserved-memory for the SPSS
+
+  qcom,spss-scsr-bits:
+    - description: Bits that are set by remote processor in the irq status
+                   register region to represent different states during
+                   boot process
+
+  child-node:
+    description: Subnode named either "smd-edge" or "glink-edge" that
+                 describes the communication edge, channels and devices
+                 related to the SPSS.
+
+required:
+  - compatible
+  - reg
+  - reg-names
+  - interrupts
+  - clocks

Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml line 85

Quoted 10 lines
Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml new file +10 -0
+  - clock-names
+  - cx-supply
+  - px-supply
+  - memory-region
+  - qcom,spss-scsr-bits
+
+
+examples:
+  - |
+    spss {

remoteproc@188101c but probably rather remoteproc@1880000

Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml line 86

Quoted 1 line
Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml new file +1 -0
+        compatible = "qcom,sm8250-spss-pil";

s/pil/pas/

Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml line 91

Quoted 5 lines
Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml new file +5 -0
+        reg = <0x188101c 0x4>,
+                <0x1881024 0x4>,
+                <0x1881028 0x4>,
+                <0x188103c 0x4>,
+                <0x1882014 0x4>;

As noted in the driver review, these are all from the same block, map it once.

Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml line 94

Quoted 3 lines
Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml new file +3 -0
+        reg-names = "sp2soc_irq_status", "sp2soc_irq_clr",
+                    "sp2soc_irq_mask", "rmb_err", "rmb_err_spare2";
+        interrupts = <0 352 1>;

interrupts = <GIC_SPI 352 IRQ_TYPE_EDGE_RISING>;

Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml line 96

Quoted 2 lines
Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml new file +2 -0
+
+        cx-supply = <&VDD_CX_LEVEL>;

These are power domains.

Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml line 97

Quoted 1 line
Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml new file +1 -0
+        cx-uV-uA = <RPMH_REGULATOR_LEVEL_TURBO 100000>;

And we'll just vote for max.

Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml line 103

Quoted 6 lines
Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml new file +6 -0
+        px-supply = <&VDD_MX_LEVEL>;
+        px-uV = <RPMH_REGULATOR_LEVEL_TURBO 100000>;
+
+        clocks = <&clock_rpmh RPMH_CXO_CLK>;
+        clock-names = "xo";
+        qcom,proxy-clock-names = "xo";

We don't specify this in DT.

Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml line 104

Quoted 1 line
Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml new file +1 -0
+        status = "ok";

This can be omitted from the example.

Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml line 107

Quoted 3 lines
Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml new file +3 -0
+
+        memory-region = <&pil_spss_mem>;
+        qcom,spss-scsr-bits = <24 25>;

As requested, just hard code these in the driver instead.

Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml line 109

Quoted 2 lines
Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml new file +2 -0
+
+        glink-edge {

This depends on a separate binding, which we haven't yet discussed. Perhaps worth omitting it for now?

Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml line 111

Quoted 2 lines
Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml new file +2 -0
+                qcom,remote-pid = <8>;
+                transport = "spss";

The registered subdev should always be of "spss" type, shouldn't be a need to describe that here.

Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml line 115

Quoted 4 lines
Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml new file +4 -0
+                mboxes = <&sp_scsr 0>;
+                mbox-names = "spss_spss";
+                interrupt-parent = <&intsp>;
+                interrupts = <0 0 IRQ_TYPE_LEVEL_HIGH>;

As you map the entire scsr region in the remoteproc driver, this should reference the spss rproc node above.

Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml line 120

Quoted 5 lines
Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml new file +5 -0
+
+                reg = <0x1885008 0x8>,
+                      <0x1885010 0x4>;
+                reg-names = "qcom,spss-addr",
+                            "qcom,spss-size";

And it seems reasonable that we pass this information from the rproc when we create the subdev, rather than having the glink implementation dig it out from the scsr.

Regards,
Bjorn

Quoted 8 lines
Documentation/devicetree/bindings/remoteproc/qcom,spss.yaml new file +5 -0
+
+                label = "spss";
+                qcom,glink-label = "spss";
+        };
+    };
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#4 lore raw
In reply to Rishabh Bhatnagar
Quoted 112 lines · Rishabh Bhatnagar

This adds Peripheral Authentication System (PAS) based SPSS PIL driver for loading SPSS firmware and booting. The firmware is verified and booted with the help of the PAS in TrustZone. This driver also adds functionality to handle late attach feature for a remote processor. The driver queries the initial state of remote processor and is capable of handling successful bootup and crash scenarios.

---
 drivers/remoteproc/Kconfig     |  17 ++
 drivers/remoteproc/Makefile    |   1 +
 drivers/remoteproc/qcom_spss.c | 500 +++++++++++++++++++++++++++++++++++++++++
 3 files changed, 518 insertions(+)
 create mode 100644 drivers/remoteproc/qcom_spss.c
drivers/remoteproc/Kconfig +17 -0
index de3862c..ab3e89a 100644
@@ -151,6 +151,23 @@ config QCOM_Q6V5_PAS
 	  for the Qualcomm Hexagon v5 based remote processors. This is commonly
 	  used to control subsystems such as ADSP, Compute and Sensor.
 
+config QCOM_SPSS
+	tristate "Qualcomm secure subsystem Peripheral loading support"
+	depends on OF && ARCH_QCOM
+	depends on QCOM_SMEM
+	depends on RPMSG_QCOM_SMD || (COMPILE_TEST && RPMSG_QCOM_SMD=n)
+	depends on RPMSG_QCOM_GLINK_SMEM || RPMSG_QCOM_GLINK_SMEM=n
+	depends on QCOM_SYSMON || QCOM_SYSMON=n
+	select MFD_SYSCON
+	select QCOM_MDT_LOADER
+	select QCOM_Q6V5_COMMON
+	select QCOM_RPROC_COMMON
+	select QCOM_SCM
+	help
+	  Say y here to support the TrustZone based Peripheral Image Loader
+	  for the Qualcomm secure subsystem remote processor. This also supports
+	  remote processors that are booted before kernel comes up.
+
 config QCOM_Q6V5_WCSS
 	tristate "Qualcomm Hexagon based WCSS Peripheral Image Loader"
 	depends on OF && ARCH_QCOM
drivers/remoteproc/Makefile +1 -0
index e30a1b1..e781552 100644
@@ -15,6 +15,7 @@ obj-$(CONFIG_OMAP_REMOTEPROC)		+= omap_remoteproc.o
 obj-$(CONFIG_WKUP_M3_RPROC)		+= wkup_m3_rproc.o
 obj-$(CONFIG_DA8XX_REMOTEPROC)		+= da8xx_remoteproc.o
 obj-$(CONFIG_KEYSTONE_REMOTEPROC)	+= keystone_remoteproc.o
+obj-$(CONFIG_QCOM_SPSS)			+= qcom_spss.o
 obj-$(CONFIG_QCOM_RPROC_COMMON)		+= qcom_common.o
 obj-$(CONFIG_QCOM_Q6V5_COMMON)		+= qcom_q6v5.o
 obj-$(CONFIG_QCOM_Q6V5_ADSP)		+= qcom_q6v5_adsp.o
drivers/remoteproc/qcom_spss.c new file +50 -0
new file mode 100644
index 0000000..7be7b2b
@@ -0,0 +1,500 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (c) 2020, The Linux Foundation. All rights reserved.
+ * Qualcomm SPSS Peripheral Image Loader for SM8250
+ *
+ */
+
+#include <linux/clk.h>
+#include <linux/firmware.h>
+#include <linux/interrupt.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/of_address.h>
+#include <linux/of_device.h>
+#include <linux/platform_device.h>
+#include <linux/qcom_scm.h>
+#include <linux/regulator/consumer.h>
+#include <linux/remoteproc.h>
+#include <linux/soc/qcom/mdt_loader.h>
+
+#include "qcom_common.h"
+#include "remoteproc_internal.h"
+
+#define ERR_READY	0
+#define PBL_DONE	1
+
+struct spss_data {
+	const char *firmware_name;
+	int pas_id;
+	const char *ssr_name;
+	bool skip_fw_load;
+};
+
+struct qcom_spss {
+	struct device *dev;
+	struct rproc *rproc;
+
+	struct clk *xo;
+
+	struct regulator *cx_supply;
+	struct regulator *px_supply;
+
+	int pas_id;
+
+	struct completion start_done;
+
+	phys_addr_t mem_phys;
+	phys_addr_t mem_reloc;
+	void *mem_region;
+	size_t mem_size;

drivers/remoteproc/qcom_spss.c line 60

Quoted 10 lines
drivers/remoteproc/qcom_spss.c new file +10 -0
+	int generic_irq;
+
+	struct qcom_rproc_glink glink_subdev;
+	struct qcom_rproc_ssr ssr_subdev;
+	void __iomem *irq_status;
+	void __iomem *irq_clr;
+	void __iomem *irq_mask;
+	void __iomem *err_status;
+	void __iomem *err_status_spare;
+	u32 bits_arr[2];

Rather than having this lookup table, just define PBL_DONE and ERR_READY as BIT(24) and BIT(25), respectively. Once we add support for a platform where these differs we can add it to the data associated with the compatible.

drivers/remoteproc/qcom_spss.c line 69

Quoted 9 lines
drivers/remoteproc/qcom_spss.c new file +9 -0
+};
+
+static void clear_pbl_done(struct qcom_spss *spss)
+{
+	uint32_t err_value, rmb_err_spare0, rmb_err_spare1, rmb_err_spare2;
+
+	err_value =  __raw_readl(spss->err_status);
+	rmb_err_spare2 =  __raw_readl(spss->err_status_spare);
+	rmb_err_spare1 =  __raw_readl(spss->err_status_spare-4);

err_status_spare is the returned value of a ioremap, so -4 here is invalid.

drivers/remoteproc/qcom_spss.c line 70

Quoted 1 line
drivers/remoteproc/qcom_spss.c new file +1 -0
+	rmb_err_spare0 =  __raw_readl(spss->err_status_spare-8);

Why bypass the endian handling here? Can we use readl() and writel() throughout instead?

drivers/remoteproc/qcom_spss.c line 79

Quoted 9 lines
drivers/remoteproc/qcom_spss.c new file +9 -0
+
+	if (err_value) {
+		dev_err(spss->dev, "PBL error status register: 0x%08x,\
+			spare0 register: 0x%08x, spare1 register: 0x%08x,\
+			spare2 register: 0x%08x\n", err_value, rmb_err_spare0,
+					rmb_err_spare1, rmb_err_spare2);
+	} else
+		dev_info(spss->dev, "PBL_DONE - 1st phase loading [%s] \
+					completed ok\n", spss->rproc->name);

Please omit any "noisy" prints in the logs. You may dev_dbg() this if you really want it.

Quoted 77 lines
drivers/remoteproc/qcom_spss.c new file +77 -0
+
+	__raw_writel(BIT(spss->bits_arr[PBL_DONE]), spss->irq_clr);
+}
+
+static void clear_err_ready(struct qcom_spss *spss)
+{
+	dev_info(spss->dev, "SW_INIT_DONE - 2nd phase loading [%s] \
+					completed ok\n", spss->rproc->name);
+
+	__raw_writel(BIT(spss->bits_arr[ERR_READY]), spss->irq_clr);
+	complete(&spss->start_done);
+}
+
+static void clear_sw_init_done_error(struct qcom_spss *spss, int err)
+{
+	uint32_t rmb_err_spare0;
+	uint32_t rmb_err_spare1;
+	uint32_t rmb_err_spare2;
+
+	dev_info(spss->dev, "SW_INIT_DONE - ERROR [%s] [0x%x].\n",
+						spss->rproc->name, err);
+
+	rmb_err_spare2 =  __raw_readl(spss->err_status_spare);
+	rmb_err_spare1 =  __raw_readl(spss->err_status_spare-4);
+	rmb_err_spare0 =  __raw_readl(spss->err_status_spare-8);
+
+	dev_err(spss->dev, "spare0 register: 0x%08x, spare1 register: 0x%08x,\
+		spare2 register: 0x%08x\n", rmb_err_spare0, rmb_err_spare1,
+							rmb_err_spare2);
+
+	/* Clear the interrupt source */
+	__raw_writel(BIT(spss->bits_arr[ERR_READY]), spss->irq_clr);
+}
+
+
+
+static void clear_wdog(struct qcom_spss *spss)
+{
+	/* Check crash status to know if device is restarting*/
+	if (spss->rproc->state == RPROC_RUNNING) {
+		dev_err(spss->dev, "wdog bite received from %s!\n",
+							spss->rproc->name);
+		__raw_writel(BIT(spss->bits_arr[ERR_READY]), spss->irq_clr);
+		rproc_report_crash(spss->rproc, RPROC_WATCHDOG);
+	}
+}
+
+static irqreturn_t spss_generic_handler(int irq, void *dev_id)
+{
+	struct qcom_spss *spss = dev_id;
+	uint32_t status_val, err_value;
+
+	err_value =  __raw_readl(spss->err_status_spare);
+	status_val = __raw_readl(spss->irq_status);
+
+	if (status_val & BIT(spss->bits_arr[ERR_READY])) {
+		if (!err_value)
+			clear_err_ready(spss);
+		else if (err_value == 0x44554d50)
+			clear_wdog(spss);
+		else
+			clear_sw_init_done_error(spss, err_value);
+	}
+
+	if (status_val & BIT(spss->bits_arr[PBL_DONE]))
+		clear_pbl_done(spss);
+
+	return IRQ_HANDLED;
+}
+
+static void mask_scsr_irqs(struct qcom_spss *spss)
+{
+	uint32_t mask_val;
+
+	/* Masking all interrupts */
+	mask_val = ~0;
+	__raw_writel(mask_val,  spss->irq_mask);

drivers/remoteproc/qcom_spss.c line 166

Quoted 10 lines
drivers/remoteproc/qcom_spss.c new file +10 -0
+}
+
+static void unmask_scsr_irqs(struct qcom_spss *spss)
+{
+	uint32_t mask_val;
+
+	/* unmasking interrupts handled by HLOS */
+	mask_val = ~0;
+	__raw_writel(mask_val & ~BIT(spss->bits_arr[ERR_READY]) &
+			~BIT(spss->bits_arr[PBL_DONE]), spss->irq_mask);

This looks undefined, do you mean to write ~(ERR_READY | PBL_DONE)?

drivers/remoteproc/qcom_spss.c line 169

Quoted 3 lines
drivers/remoteproc/qcom_spss.c new file +3 -0
+}
+
+static bool check_status(struct qcom_spss *spss)

s/check_status/check_if_crashed/

Quoted 43 lines
drivers/remoteproc/qcom_spss.c new file +43 -0
+{
+	uint32_t status_val, err_value;
+
+	err_value =  __raw_readl(spss->err_status_spare);
+	status_val = __raw_readl(spss->irq_status);
+
+	if ((status_val & BIT(spss->bits_arr[ERR_READY])) && err_value) {
+		dev_err(spss->dev, "SPSS crashed before booting\n");
+		return true;
+	}
+	return false;
+}
+
+static int spss_load(struct rproc *rproc, const struct firmware *fw)
+{
+	struct qcom_spss *spss = (struct qcom_spss *)rproc->priv;
+
+	if (rproc->skip_fw_load)
+		return 0;
+
+	return qcom_mdt_load(spss->dev, fw, rproc->firmware, spss->pas_id,
+			     spss->mem_region, spss->mem_phys, spss->mem_size,
+			     &spss->mem_reloc);
+
+}
+
+static int spss_stop(struct rproc *rproc)
+{
+	struct qcom_spss *spss = (struct qcom_spss *)rproc->priv;
+	int ret;
+
+	ret = qcom_scm_pas_shutdown(spss->pas_id);
+	if (ret)
+		dev_err(spss->dev, "failed to shutdown: %d\n", ret);
+
+	mask_scsr_irqs(spss);
+	/* Negate skip fw load if already set */
+	if (rproc->skip_fw_load)
+		rproc->skip_fw_load = false;
+
+	return ret;
+}
+

drivers/remoteproc/qcom_spss.c line 222

Quoted 10 lines
drivers/remoteproc/qcom_spss.c new file +10 -0
+static int spss_start(struct rproc *rproc)
+{
+	struct qcom_spss *spss = (struct qcom_spss *)rproc->priv;
+	int ret = 0;
+
+	if (rproc->skip_fw_load) {
+		/* If rproc already crashed stop it and retry boot*/
+		if (check_status(spss)) {
+			spss_stop(rproc);
+			return -EAGAIN;

Who will attempt to start the remoteproc again? I presume you do this because if you get this far you don't have a way to load the firmware again?

Seems like something we need to consider in when adding the support for synchronizing the MCUs - as Mathieu is working on.

drivers/remoteproc/qcom_spss.c line 233

Quoted 11 lines
drivers/remoteproc/qcom_spss.c new file +11 -0
+		}
+		/* If booted successfully then wait for init_done*/
+		else
+			goto wait_for_init;
+	}
+
+	ret = clk_prepare_enable(spss->xo);
+	if (ret)
+		return ret;
+
+	ret = regulator_enable(spss->cx_supply);

Would be nicer with a regulator_bulk_enable(), but these are power domains upstream, so this needs to be changed.

drivers/remoteproc/qcom_spss.c line 240

Quoted 7 lines
drivers/remoteproc/qcom_spss.c new file +7 -0
+	if (ret)
+		goto disable_xo_clk;
+
+	ret = regulator_enable(spss->px_supply);
+	if (ret)
+		goto disable_cx_supply;
+

You want to reinit start_done here.

Quoted 47 lines
drivers/remoteproc/qcom_spss.c new file +47 -0
+	ret = qcom_scm_pas_auth_and_reset(spss->pas_id);
+	if (ret) {
+		dev_err(spss->dev,
+			"failed to authenticate image and release reset\n");
+		goto disable_px_supply;
+	}
+
+wait_for_init:
+	unmask_scsr_irqs(spss);
+
+	ret = wait_for_completion_timeout(&spss->start_done,
+						msecs_to_jiffies(5000));
+	if (!ret) {
+		dev_err(spss->dev, "start timed out\n");
+		spss_stop(rproc);
+	}
+	ret = ret ? 0 : -ETIMEDOUT;
+
+disable_px_supply:
+	regulator_disable(spss->px_supply);
+disable_cx_supply:
+	regulator_disable(spss->cx_supply);
+disable_xo_clk:
+	clk_disable_unprepare(spss->xo);
+	return ret;
+}
+
+static void *spss_da_to_va(struct rproc *rproc, u64 da, int len)
+{
+	struct qcom_spss *spss = (struct qcom_spss *)rproc->priv;
+	int offset;
+
+	offset = da - spss->mem_reloc;
+	if (offset < 0 || offset + len > spss->mem_size)
+		return NULL;
+
+	return spss->mem_region + offset;
+}
+
+static const struct rproc_ops spss_ops = {
+	.start = spss_start,
+	.stop = spss_stop,
+	.da_to_va = spss_da_to_va,
+	.load = spss_load,
+};
+
+static int spss_init_clock(struct qcom_spss *spss)

drivers/remoteproc/qcom_spss.c line 297

Quoted 10 lines
drivers/remoteproc/qcom_spss.c new file +10 -0
+{
+	int ret;
+
+	spss->xo = devm_clk_get(spss->dev, "xo");
+	if (IS_ERR(spss->xo)) {
+		ret = PTR_ERR(spss->xo);
+		if (ret != -EPROBE_DEFER)
+			dev_err(spss->dev, "failed to get xo clock");
+		return ret;
+	}
	if (IS_ERR(spss->xo) && PTR_ERR(spss->xo) != -EPROBE_DEFER)
		dev_err(...);

	return PTR_ERR_OR_ZERO(spss->xo);

drivers/remoteproc/qcom_spss.c line 302

Quoted 5 lines
drivers/remoteproc/qcom_spss.c new file +5 -0
+
+	return 0;
+}
+
+static int spss_init_regulator(struct qcom_spss *spss)

These are power domains upstream.

Quoted 40 lines
drivers/remoteproc/qcom_spss.c new file +40 -0
+{
+	spss->cx_supply = devm_regulator_get(spss->dev, "cx");
+	if (IS_ERR(spss->cx_supply))
+		return PTR_ERR(spss->cx_supply);
+	regulator_set_load(spss->cx_supply, 100000);
+
+	spss->px_supply = devm_regulator_get(spss->dev, "px");
+	if (IS_ERR(spss->px_supply))
+		return PTR_ERR(spss->px_supply);
+	regulator_set_load(spss->px_supply, 100000);
+
+	return 0;
+}
+
+static int spss_alloc_memory_region(struct qcom_spss *spss)
+{
+	struct device_node *node;
+	struct resource r;
+	int ret;
+
+	node = of_parse_phandle(spss->dev->of_node, "memory-region", 0);
+	if (!node) {
+		dev_err(spss->dev, "no memory-region specified\n");
+		return -EINVAL;
+	}
+
+	ret = of_address_to_resource(node, 0, &r);
+	if (ret)
+		return ret;
+
+	spss->mem_phys = spss->mem_reloc = r.start;
+	spss->mem_size = resource_size(&r);
+	spss->mem_region = devm_ioremap_wc(spss->dev, spss->mem_phys,
+								spss->mem_size);
+	if (!spss->mem_region) {
+		dev_err(spss->dev, "unable to map memory region: %pa+%zx\n",
+			&r.start, spss->mem_size);
+		return -EBUSY;
+	}
+

drivers/remoteproc/qcom_spss.c line 352

Quoted 10 lines
drivers/remoteproc/qcom_spss.c new file +10 -0
+	return 0;
+}
+
+static int qcom_spss_init_mmio(struct platform_device *pdev, struct qcom_spss *spss)
+{
+	struct resource *res;
+	int ret;
+
+	res = platform_get_resource_byname(pdev, IORESOURCE_MEM,
+						"sp2soc_irq_status");

These are all individual registers in the same memory region, please just map it once.

And yes, this driver then becomes the natural place to implement what lives in qsee_ipc_irq.c as well.

drivers/remoteproc/qcom_spss.c line 353

Quoted 1 line
drivers/remoteproc/qcom_spss.c new file +1 -0
+	spss->irq_status = devm_ioremap_resource(&pdev->dev, res);

Use devm_platform_ioremap_resource_byname() - or now that you only have a single region use devm_platform_ioremap_resource(pdev, 0);

Quoted 20 lines
drivers/remoteproc/qcom_spss.c new file +20 -0
+	if (IS_ERR(spss->irq_status))
+		return PTR_ERR(spss->irq_status);
+
+	res = platform_get_resource_byname(pdev, IORESOURCE_MEM,
+						"sp2soc_irq_clr");
+	spss->irq_clr = devm_ioremap_resource(&pdev->dev, res);
+	if (IS_ERR(spss->irq_clr))
+		return PTR_ERR(spss->irq_clr);
+
+	res = platform_get_resource_byname(pdev, IORESOURCE_MEM,
+						"sp2soc_irq_mask");
+	spss->irq_mask = devm_ioremap_resource(&pdev->dev, res);
+	if (IS_ERR(spss->irq_mask))
+		return PTR_ERR(spss->irq_mask);
+
+	res = platform_get_resource_byname(pdev, IORESOURCE_MEM,
+						"rmb_err");
+	spss->err_status = devm_ioremap_resource(&pdev->dev, res);
+	if (IS_ERR(spss->err_status))
+		return PTR_ERR(spss->err_status);

drivers/remoteproc/qcom_spss.c line 383

Quoted 10 lines
drivers/remoteproc/qcom_spss.c new file +10 -0
+
+	res = platform_get_resource_byname(pdev, IORESOURCE_MEM,
+						"rmb_err_spare2");
+	spss->err_status_spare = devm_ioremap_resource(&pdev->dev, res);
+	if (IS_ERR(spss->err_status_spare))
+		return PTR_ERR(spss->err_status_spare);
+
+	ret = of_property_read_u32_array(pdev->dev.of_node,
+			       "qcom,spss-scsr-bits", spss->bits_arr,
+				ARRAY_SIZE(spss->bits_arr));

As requested above, please just hard code these in the driver for now.

Quoted 13 lines
drivers/remoteproc/qcom_spss.c new file +13 -0
+	if (ret)
+		return ret;
+
+	return 0;
+
+}
+
+static int qcom_spss_probe(struct platform_device *pdev)
+{
+	const struct spss_data *desc;
+	struct qcom_spss *spss;
+	struct rproc *rproc;
+	const char *fw_name;

drivers/remoteproc/qcom_spss.c line 406

Quoted 10 lines
drivers/remoteproc/qcom_spss.c new file +10 -0
+
+	int ret;
+
+	desc = of_device_get_match_data(&pdev->dev);
+	if (!desc)
+		return -EINVAL;
+
+	if (!qcom_scm_is_available())
+		return -EPROBE_DEFER;
+

Please query firmware-name from DT, before falling back to "spss.mdt" as default file name - as in the other drivers.

drivers/remoteproc/qcom_spss.c line 414

Quoted 8 lines
drivers/remoteproc/qcom_spss.c new file +8 -0
+	rproc = rproc_alloc(&pdev->dev, pdev->name, &spss_ops,
+			    fw_name, sizeof(*spss));
+	if (!rproc) {
+		dev_err(&pdev->dev, "unable to allocate remoteproc\n");
+		return -ENOMEM;
+	}
+
+	rproc->skip_fw_load = desc->skip_fw_load;

Rather than inherting this from desc you should be able to just read it back from the hardware here (and this will still be pending conclusion on the core support for "skip_fw_load").

drivers/remoteproc/qcom_spss.c line 416

Quoted 2 lines
drivers/remoteproc/qcom_spss.c new file +2 -0
+
+	spss = (struct qcom_spss *)rproc->priv;

No need to explicitly typecast a void *.

Quoted 24 lines
drivers/remoteproc/qcom_spss.c new file +24 -0
+	spss->dev = &pdev->dev;
+	spss->rproc = rproc;
+	spss->pas_id = desc->pas_id;
+	init_completion(&spss->start_done);
+	platform_set_drvdata(pdev, spss);
+
+	ret = qcom_spss_init_mmio(pdev, spss);
+	if (ret)
+		goto free_rproc;
+
+	ret = spss_alloc_memory_region(spss);
+	if (ret)
+		goto free_rproc;
+
+	ret = spss_init_clock(spss);
+	if (ret)
+		goto free_rproc;
+
+	ret = spss_init_regulator(spss);
+	if (ret)
+		goto free_rproc;
+
+	qcom_add_glink_subdev(rproc, &spss->glink_subdev);
+	qcom_add_ssr_subdev(rproc, &spss->ssr_subdev, desc->ssr_name);

drivers/remoteproc/qcom_spss.c line 450

Quoted 10 lines
drivers/remoteproc/qcom_spss.c new file +10 -0
+
+	ret = rproc_add(rproc);
+	if (ret)
+		goto free_rproc;
+
+	mask_scsr_irqs(spss);
+	spss->generic_irq = platform_get_irq(pdev, 0);
+	ret = devm_request_threaded_irq(&pdev->dev, spss->generic_irq, NULL,
+		spss_generic_handler, IRQF_TRIGGER_RISING | IRQF_ONESHOT,
+							"generic-irq", spss);

Please skip IRQF_TRIGGER_RISING and let this come from DT, give the irq a better name (spss-generic-irq?) and please indent broken lines to line up with the open parenthesis.

drivers/remoteproc/qcom_spss.c line 453

Quoted 3 lines
drivers/remoteproc/qcom_spss.c new file +3 -0
+	if (ret) {
+		dev_err(&pdev->dev, "failed to acquire generic IRQ\n");
+		goto free_rproc;

This is past rproc_add(), so you need to rproc_del() as well, unless you do this before adding the remoteproc - which would make sense, in case the core finds the firmware and "boots" the remoteproc before returning from rproc_add().

Quoted 16 lines
drivers/remoteproc/qcom_spss.c new file +16 -0
+	}
+
+	return 0;
+
+free_rproc:
+	rproc_free(rproc);
+
+	return ret;
+}
+
+static int qcom_spss_remove(struct platform_device *pdev)
+{
+	struct qcom_spss *spss = platform_get_drvdata(pdev);
+
+	rproc_del(spss->rproc);
+	qcom_remove_glink_subdev(spss->rproc, &spss->glink_subdev);

drivers/remoteproc/qcom_spss.c line 479

Quoted 10 lines
drivers/remoteproc/qcom_spss.c new file +10 -0
+	qcom_remove_ssr_subdev(spss->rproc, &spss->ssr_subdev);
+	rproc_free(spss->rproc);
+
+	return 0;
+}
+
+static const struct spss_data spss_resource_init = {
+		.firmware_name = "spss.mdt",
+		.pas_id = 14,
+		.ssr_name = "spss",

I don't see that these are going to differ between different platforms, so just hard code them where needed.

drivers/remoteproc/qcom_spss.c line 480

Quoted 1 line
drivers/remoteproc/qcom_spss.c new file +1 -0
+		.skip_fw_load = true,

Can we not derive this from the hardware state during initialization?

Regards,
Bjorn

Quoted 23 lines
drivers/remoteproc/qcom_spss.c new file +20 -0
+};
+
+static const struct of_device_id spss_of_match[] = {
+	{ .compatible = "qcom,sm8250-spss-pas", .data = &spss_resource_init},
+	{ },
+};
+MODULE_DEVICE_TABLE(of, spss_of_match);
+
+static struct platform_driver spss_driver = {
+	.probe = qcom_spss_probe,
+	.remove = qcom_spss_remove,
+	.driver = {
+		.name = "qcom_spss",
+		.of_match_table = spss_of_match,
+	},
+};
+
+module_platform_driver(spss_driver);
+MODULE_DESCRIPTION("Qualcomm Peripheral Image Loader for secure subsystem");
+MODULE_LICENSE("GPL v2");
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project