Skip to contentSkip to conversation

MIPS: cm: Probe GCR address from devicetree

PATCH In review Jiaxun Yang 4 participants 20 messages Jun 12, 2024 – Sep 10, 2024

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

v2 cover MIPS: cm: Probe GCR address from devicetree

Hi all,

This series enabled mips-cm code to probe GCR address from devicetree.

This feature has been implemented in MIPS's out-of-tree kernel for a while, and MIPS's u-boot fork on boston will generate required "mti,mips-cm" node as well.

Please review.
Thanks

---
Changes in v2:
- Fix probe order on malta (Serge)
- dt binding improvements (Conor)
- Build warning fix
- Link to v1: https://lore.kernel.org/r/[email protected]

---
Jiaxun Yang (6):
MIPS: generic: Do __dt_setup_arch in prom_init
MIPS: malta: Move SMP initialisation to device_tree_init
MIPS: cm: Prefix probe functions with __init
MIPS: Move mips_cm_probe after prom_init
dt-bindings: mips: Document mti,mips-cm
MIPS: cm: Probe GCR address from DeviceTree

 .../devicetree/bindings/mips/mti,mips-cm.yaml      | 38 ++++++++++++
 arch/mips/generic/init.c                           |  9 ++-
 arch/mips/include/asm/mips-cm.h                    |  4 +-
 arch/mips/kernel/mips-cm.c                         | 69 ++++++++++++++++++----
 arch/mips/kernel/setup.c                           |  2 +-
 arch/mips/mti-malta/malta-init.c                   |  8 ++-
 6 files changed, 111 insertions(+), 19 deletions(-)

---
base-commit: 2b84edefcad14934796fad37b16512b6a2ca467e
change-id: 20240506-cm_probe-0c667c8b63bf

Best regards,

Signature · 1 line
Jiaxun Yang <[email protected]>
#1 lore raw

PATCH v2 1/6 MIPS: generic: Do __dt_setup_arch in prom_init

We want fdt parse functions to be available as early as possible, thus do __dt_setup_arch immediately after we get fdt address in prom_init.

---
 arch/mips/generic/init.c | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)
arch/mips/generic/init.c +4 -5
index 1d712eac1617..9fd09061de78 100644
@@ -26,8 +26,12 @@ static __initconst const void *mach_match_data;
 
 void __init prom_init(void)
 {
+	fw_init_cmdline();
 	plat_get_fdt();
 	BUG_ON(!fdt);
+	if (mach && mach->fixup_fdt)
+		fdt = mach->fixup_fdt(fdt, mach_match_data);
+	__dt_setup_arch((void *)fdt);
 }
 
 void __init *plat_get_fdt(void)
@@ -101,11 +105,6 @@ void __init plat_fdt_relocated(void *new_location)
 
 void __init plat_mem_setup(void)
 {
-	if (mach && mach->fixup_fdt)
-		fdt = mach->fixup_fdt(fdt, mach_match_data);
-
-	fw_init_cmdline();
-	__dt_setup_arch((void *)fdt);
 }
 
 void __init device_tree_init(void)
Signature · 1 line
2.43.0
#2 lore raw

PATCH v2 2/6 MIPS: malta: Move SMP initialisation to device_tree_init

In reply to Jiaxun Yang

Move SMP initialisation to device_tree_init as what generic platform did, this allows us to move mips_cm_probe to a later point, as CPC needs to be probed after CM.

---
v2: New patch

---
 arch/mips/mti-malta/malta-init.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
arch/mips/mti-malta/malta-init.c +7 -1
index 000d6d50520a..f67247d7fc69 100644
@@ -284,12 +284,18 @@ void __init prom_init(void)
 #ifdef CONFIG_SERIAL_8250_CONSOLE
 	console_config();
 #endif
-	/* Early detection of CMP support */
+}
+
+void __init device_tree_init(void)
+{
+	unflatten_and_copy_device_tree();
 	mips_cpc_probe();
 
 	if (!register_cps_smp_ops())
 		return;
 	if (!register_vsmp_smp_ops())
 		return;
+
 	register_up_smp_ops();
 }
+
Signature · 1 line
2.43.0
#3 lore raw

PATCH v2 3/6 MIPS: cm: Prefix probe functions with __init

In reply to Jiaxun Yang

Those functions are only used at boot time.
Prefix them with __init so they can be discarded after boot.

---
 arch/mips/include/asm/mips-cm.h | 4 ++--
 arch/mips/kernel/mips-cm.c      | 8 ++++----
 2 files changed, 6 insertions(+), 6 deletions(-)
arch/mips/include/asm/mips-cm.h +2 -2
index c2930a75b7e4..5292b516d60b 100644
@@ -30,7 +30,7 @@ extern void __iomem *mips_cm_l2sync_base;
  * and may be overridden by platforms which determine this address in a
  * different way by defining a function with the same prototype.
  */
-extern phys_addr_t mips_cm_phys_base(void);
+extern phys_addr_t __init mips_cm_phys_base(void);
 
 /**
  * mips_cm_l2sync_phys_base - retrieve the physical base address of the CM
@@ -43,7 +43,7 @@ extern phys_addr_t mips_cm_phys_base(void);
  * determine this address in a different way by defining a function with the
  * same prototype.
  */
-extern phys_addr_t mips_cm_l2sync_phys_base(void);
+extern phys_addr_t  __init mips_cm_l2sync_phys_base(void);
 
 /*
  * mips_cm_is64 - determine CM register width
arch/mips/kernel/mips-cm.c +4 -4
index 3a115fab5573..dddc9428fe58 100644
@@ -179,7 +179,7 @@ static char *cm3_causes[32] = {
 static DEFINE_PER_CPU_ALIGNED(spinlock_t, cm_core_lock);
 static DEFINE_PER_CPU_ALIGNED(unsigned long, cm_core_lock_flags);
 
-phys_addr_t __weak mips_cm_phys_base(void)
+phys_addr_t __init __weak mips_cm_phys_base(void)
 {
 	unsigned long cmgcr;
 
@@ -198,7 +198,7 @@ phys_addr_t __weak mips_cm_phys_base(void)
 	return (cmgcr & MIPS_CMGCRF_BASE) << (36 - 32);
 }
 
-phys_addr_t __weak mips_cm_l2sync_phys_base(void)
+phys_addr_t __init __weak mips_cm_l2sync_phys_base(void)
 {
 	u32 base_reg;
 
@@ -214,7 +214,7 @@ phys_addr_t __weak mips_cm_l2sync_phys_base(void)
 	return mips_cm_phys_base() + MIPS_CM_GCR_SIZE;
 }
 
-static void mips_cm_probe_l2sync(void)
+static void __init mips_cm_probe_l2sync(void)
 {
 	unsigned major_rev;
 	phys_addr_t addr;
@@ -237,7 +237,7 @@ static void mips_cm_probe_l2sync(void)
 	mips_cm_l2sync_base = ioremap(addr, MIPS_CM_L2SYNC_SIZE);
 }
 
-int mips_cm_probe(void)
+int __init mips_cm_probe(void)
 {
 	phys_addr_t addr;
 	u32 base_reg;
Signature · 1 line
2.43.0
#4 lore raw

PATCH v2 4/6 MIPS: Move mips_cm_probe after prom_init

In reply to Jiaxun Yang

Move mips_cm_probe after prom_init so we can use fdt functions in mips_cm_probe to obtain essential information.

Impat for all systems that may have CM in system:
- geneirc: Adjusted code to accommodate this change
- Lantiq: No impact, CM configuration won't be changed at all
- ralink: Called mips_cm_probe on it's own, in prom_init->prom_soc_init
- malta: Addressed in previous patch

---
 arch/mips/kernel/setup.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/mips/kernel/setup.c +1 -1
index 12a1a4ffb602..732579c8f4f8 100644
@@ -773,8 +773,8 @@ static void __init setup_rng_seed(void)
 void __init setup_arch(char **cmdline_p)
 {
 	cpu_probe();
-	mips_cm_probe();
 	prom_init();
+	mips_cm_probe();
 
 	setup_early_fdc_console();
 #ifdef CONFIG_EARLY_PRINTK
Signature · 1 line
2.43.0
#5 lore raw

PATCH v2 5/6 dt-bindings: mips: Document mti,mips-cm

In reply to Jiaxun Yang

Add devicetree binding documentation for MIPS Coherence Manager.

---
v2:
- Better wording for register desc
- cm -> coherency-manager
- schema matches compatible

---
 .../devicetree/bindings/mips/mti,mips-cm.yaml      | 38 ++++++++++++++++++++++
 1 file changed, 38 insertions(+)
Documentation/devicetree/bindings/mips/mti,mips-cm.yaml new file +38 -0
new file mode 100644
index 000000000000..9f500804737d
@@ -0,0 +1,38 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/mips/mti,mips-cm.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: MIPS Coherence Manager
+
+description: |
+  Defines a location of the MIPS Coherence Manager registers.
+
+maintainers:
+  - Jiaxun Yang <[email protected]>
+
+properties:
+  compatible:
+    const: mti,mips-cm
+
+  reg:
+    description:
+      Base address and size of an unoccupied region in system's MMIO address
+      space, which will be used to map the MIPS CM global control registers
+      block. It is conventionally decided by the system integrator.
+    maxItems: 1
+
+required:
+  - compatible
+  - reg
+
+additionalProperties: false
+
+examples:
+  - |
+    coherency-manager@1fbf8000 {
+      compatible = "mti,mips-cm";
+      reg = <0x1bde8000 0x8000>;
+    };
+...
Signature · 1 line
2.43.0
Reviewed-by on v2 5/6
#6 lore raw

Re: [PATCH v2 5/6] dt-bindings: mips: Document mti,mips-cm

Quoted 18 lines · Jiaxun Yang

Add devicetree binding documentation for MIPS Coherence Manager.

---
v2:
- Better wording for register desc
- cm -> coherency-manager
- schema matches compatible

---
 .../devicetree/bindings/mips/mti,mips-cm.yaml      | 38 ++++++++++++++++++++++
 1 file changed, 38 insertions(+)
Documentation/devicetree/bindings/mips/mti,mips-cm.yaml new file +0 -0
new file mode 100644
index 000000000000..9f500804737d

Documentation/devicetree/bindings/mips/mti,mips-cm.yaml line 9

Quoted 9 lines
Documentation/devicetree/bindings/mips/mti,mips-cm.yaml new file +9 -0
@@ -0,0 +1,38 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/mips/mti,mips-cm.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: MIPS Coherence Manager
+
+description: |

The | is not needed here.

Thanks,
Conor.

#7 lore raw

Re: [PATCH v2 5/6] dt-bindings: mips: Document mti,mips-cm

In reply to Jiaxun Yang
Quoted 33 lines · Jiaxun Yang

Add devicetree binding documentation for MIPS Coherence Manager.

---
v2:
- Better wording for register desc
- cm -> coherency-manager
- schema matches compatible

---
 .../devicetree/bindings/mips/mti,mips-cm.yaml      | 38 ++++++++++++++++++++++
 1 file changed, 38 insertions(+)
Documentation/devicetree/bindings/mips/mti,mips-cm.yaml new file +15 -0
new file mode 100644
index 000000000000..9f500804737d
@@ -0,0 +1,38 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/mips/mti,mips-cm.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: MIPS Coherence Manager
+
+description: |
+  Defines a location of the MIPS Coherence Manager registers.
+
+maintainers:
+  - Jiaxun Yang <[email protected]>
+
+properties:

Documentation/devicetree/bindings/mips/mti,mips-cm.yaml line 24

Quoted 10 lines
Documentation/devicetree/bindings/mips/mti,mips-cm.yaml new file +9 -0
+  compatible:
+    const: mti,mips-cm
+
 
+  reg:
+    description:
+      Base address and size of an unoccupied region in system's MMIO address
+      space, which will be used to map the MIPS CM global control registers
+      block. It is conventionally decided by the system integrator.
+    maxItems: 1

Could you please extend the reg array to containing two values: gcr and l2sync? The later is the L2-cache-only sync region which can be customized by the CM means.

It's better to define the reg-names property too, so the node would look like this:

        cm2: cm2@1fbf8000 {
                compatible = "mti,mips-cm";
                reg = <0 0x1fbf8000 0 0x8000>,
                      <0 0x1fbf0000 0 0x1000>;
                reg-names = "gcr", "l2sync";
        };

-Serge(y)

Quoted 18 lines
Documentation/devicetree/bindings/mips/mti,mips-cm.yaml new file +14 -0
+
+required:
+  - compatible
+  - reg
+
+additionalProperties: false
+
+examples:
+  - |
+    coherency-manager@1fbf8000 {
+      compatible = "mti,mips-cm";
+      reg = <0x1bde8000 0x8000>;
+    };
+...
Signature · 2 lines
-- 
2.43.0
#8 lore raw

PATCH v2 6/6 MIPS: cm: Probe GCR address from DeviceTree

In reply to Jiaxun Yang

Traditionally, CM GCR address can be probed from CP0_CMGCRBase.

However there are chips in wild that do have CM GCR but CP0_CMGCRBase is not available from CPU point of view. Thus we need to be able to probe GCR address from DeviceTree.

It is implemented as:
- If only CP0_CMGCRBase present, trust CP0_CMGCRBase
- If only mti,mips-cm node present, trust mti,mips-cm reg prop
- If both present, remap address space to address specified in dt

---
v2: Fix build warning (test bot)

---
 arch/mips/kernel/mips-cm.c | 61 +++++++++++++++++++++++++++++++++++++++++-----
 1 file changed, 55 insertions(+), 6 deletions(-)
arch/mips/kernel/mips-cm.c +55 -6
index dddc9428fe58..02afc795ba8a 100644
@@ -5,6 +5,8 @@
  */
 
 #include <linux/errno.h>
+#include <linux/of_address.h>
+#include <linux/of_fdt.h>
 #include <linux/percpu.h>
 #include <linux/spinlock.h>
 
@@ -179,23 +181,70 @@ static char *cm3_causes[32] = {
 static DEFINE_PER_CPU_ALIGNED(spinlock_t, cm_core_lock);
 static DEFINE_PER_CPU_ALIGNED(unsigned long, cm_core_lock_flags);
 
+static int __init mips_cm_fdt_scan(unsigned long node, const char *uname,
+		int depth, void *data)
+{
+	u64 addr;
+	unsigned long *cmgcr = data;
+
+	if (!of_flat_dt_is_compatible(node, "mti,mips-cm"))
+		return 0;
+
+	addr = of_flat_dt_translate_address(node);
+	if (addr == OF_BAD_ADDR || addr >= ULONG_MAX)
+		*cmgcr = 0;
+	else
+		*cmgcr = addr;
+
+	return 0;
+}
+
 phys_addr_t __init __weak mips_cm_phys_base(void)
 {
-	unsigned long cmgcr;
+	unsigned long gcr_reg = 0, gcr_dt = 0;
+
+	if (of_have_populated_dt()) {
+		int err;
+		struct resource res;
+		struct device_node *cm_node;
+
+		cm_node = of_find_compatible_node(of_root, NULL, "mti,mips-cm");
+		if (cm_node) {
+			err = of_address_to_resource(cm_node, 0, &res);
+			of_node_put(cm_node);
+			if (!err)
+				gcr_dt = res.start;
+		}
+	} else {
+		of_scan_flat_dt(mips_cm_fdt_scan, &gcr_dt);
+	}
 
 	/* Check the CMGCRBase register is implemented */
 	if (!(read_c0_config() & MIPS_CONF_M))
-		return 0;
+		return gcr_dt;
 
 	if (!(read_c0_config2() & MIPS_CONF_M))
-		return 0;
+		return gcr_dt;
 
 	if (!(read_c0_config3() & MIPS_CONF3_CMGCR))
-		return 0;
+		return gcr_dt;
 
 	/* Read the address from CMGCRBase */
-	cmgcr = read_c0_cmgcrbase();
-	return (cmgcr & MIPS_CMGCRF_BASE) << (36 - 32);
+	gcr_reg = read_c0_cmgcrbase();
+	gcr_reg = (gcr_reg & MIPS_CMGCRF_BASE) << (36 - 32);
+
+	/* If no of node, return straight away */
+	if (!gcr_dt)
+		return gcr_reg;
+
+	/* If the CMGCRBase mismatches with dt, remap it */
+	if (gcr_reg != gcr_dt) {
+		pr_info("Remapping CMGCRBase from 0x%08lx to 0x%08lx\n",
+			gcr_reg, gcr_dt);
+		change_gcr_base(CM_GCR_BASE_GCRBASE, gcr_dt);
+	}
+
+	return gcr_dt;
 }
 
 phys_addr_t __init __weak mips_cm_l2sync_phys_base(void)
Signature · 1 line
2.43.0
#9 lore raw

Re: [PATCH v2 6/6] MIPS: cm: Probe GCR address from DeviceTree

Hi Jiaxun, Thomas

Quoted 93 lines · Jiaxun Yang

Traditionally, CM GCR address can be probed from CP0_CMGCRBase.

However there are chips in wild that do have CM GCR but CP0_CMGCRBase is not available from CPU point of view. Thus we need to be able to probe GCR address from DeviceTree.

It is implemented as:
- If only CP0_CMGCRBase present, trust CP0_CMGCRBase
- If only mti,mips-cm node present, trust mti,mips-cm reg prop
- If both present, remap address space to address specified in dt

---
v2: Fix build warning (test bot)

---
 arch/mips/kernel/mips-cm.c | 61 +++++++++++++++++++++++++++++++++++++++++-----
 1 file changed, 55 insertions(+), 6 deletions(-)
arch/mips/kernel/mips-cm.c +43 -6
index dddc9428fe58..02afc795ba8a 100644
@@ -5,6 +5,8 @@
  */
 
 #include <linux/errno.h>
+#include <linux/of_address.h>
+#include <linux/of_fdt.h>
 #include <linux/percpu.h>
 #include <linux/spinlock.h>
 
@@ -179,23 +181,70 @@ static char *cm3_causes[32] = {
 static DEFINE_PER_CPU_ALIGNED(spinlock_t, cm_core_lock);
 static DEFINE_PER_CPU_ALIGNED(unsigned long, cm_core_lock_flags);
 
+static int __init mips_cm_fdt_scan(unsigned long node, const char *uname,
+		int depth, void *data)
+{
+	u64 addr;
+	unsigned long *cmgcr = data;
+
+	if (!of_flat_dt_is_compatible(node, "mti,mips-cm"))
+		return 0;
+
+	addr = of_flat_dt_translate_address(node);
+	if (addr == OF_BAD_ADDR || addr >= ULONG_MAX)
+		*cmgcr = 0;
+	else
+		*cmgcr = addr;
+
+	return 0;
+}
+
 phys_addr_t __init __weak mips_cm_phys_base(void)
 {
-	unsigned long cmgcr;
+	unsigned long gcr_reg = 0, gcr_dt = 0;
+
+	if (of_have_populated_dt()) {
+		int err;
+		struct resource res;
+		struct device_node *cm_node;
+
+		cm_node = of_find_compatible_node(of_root, NULL, "mti,mips-cm");
+		if (cm_node) {
+			err = of_address_to_resource(cm_node, 0, &res);
+			of_node_put(cm_node);
+			if (!err)
+				gcr_dt = res.start;
+		}
+	} else {
+		of_scan_flat_dt(mips_cm_fdt_scan, &gcr_dt);
+	}
 
 	/* Check the CMGCRBase register is implemented */
 	if (!(read_c0_config() & MIPS_CONF_M))
-		return 0;
+		return gcr_dt;
 
 	if (!(read_c0_config2() & MIPS_CONF_M))
-		return 0;
+		return gcr_dt;
 
 	if (!(read_c0_config3() & MIPS_CONF3_CMGCR))
-		return 0;
+		return gcr_dt;
 
 	/* Read the address from CMGCRBase */
-	cmgcr = read_c0_cmgcrbase();
-	return (cmgcr & MIPS_CMGCRF_BASE) << (36 - 32);
+	gcr_reg = read_c0_cmgcrbase();
+	gcr_reg = (gcr_reg & MIPS_CMGCRF_BASE) << (36 - 32);
+

arch/mips/kernel/mips-cm.c static char *cm3_causes[32] = { · line 245

Quoted 10 lines
arch/mips/kernel/mips-cm.c +9 -0
+	/* If no of node, return straight away */
+	if (!gcr_dt)
+		return gcr_reg;
+
+	/* If the CMGCRBase mismatches with dt, remap it */
+	if (gcr_reg != gcr_dt) {
+		pr_info("Remapping CMGCRBase from 0x%08lx to 0x%08lx\n",
+			gcr_reg, gcr_dt);
 
+		change_gcr_base(CM_GCR_BASE_GCRBASE, gcr_dt);

This causes the kernel boot-up procedure to crash/hang-up because the CM GCR base address is supposed to be redefined by means of the already mapped CM GCR address space by accessing the CM_GCR_BASE_GCRBASE register:
change_gcr_base()

+-> read_gcr_base()
    +-> addr_gcr_base()
        +-> return mips_gcr_base + CM_GCR_BASE_GCRBASE

By the time of the change_gcr_base() call in mips_cm_phys_base(), the mips_gcr_base variable hasn't been defined. So the IO operations performed in the change_gcr_base() method would be accessing the NULL-based memory space. That's why the kernel crash/hanging-up.

In order to fix this we have to first map the CM GCR block at the default base-address, then update the CM GCR-base CSR and after that remap the CM GCR-space.

Please also note, the GCR_BASE field might be RO. It depends on the IP-core configuration. So it's possible that the CM_GCR_BASE_GCRBASE field update won't work. Although that will be detected a bit later in the mips_cm_probe() method by comparing the address returned from mips_cm_phys_base() and retrieved from the CM GCR-base CSR.

-Serge(y)

Quoted 10 lines
arch/mips/kernel/mips-cm.c +3 -0
+	}
+
+	return gcr_dt;
 }
 

phys_addr_t __init __weak mips_cm_l2sync_phys_base(void)

Signature · 1 line
2.43.0
#10 lore raw

Re: [PATCH v2 6/6] MIPS: cm: Probe GCR address from DeviceTree

在2024年9月10日九月 下午1:36,Serge Semin写道:
[...]

Quoted 6 lines

This causes the kernel boot-up procedure to crash/hang-up because the CM GCR base address is supposed to be redefined by means of the already mapped CM GCR address space by accessing the CM_GCR_BASE_GCRBASE register:
change_gcr_base()

Quoted 8 lines
+-> read_gcr_base()
    +-> addr_gcr_base()
        +-> return mips_gcr_base + CM_GCR_BASE_GCRBASE

By the time of the change_gcr_base() call in mips_cm_phys_base(), the mips_gcr_base variable hasn't been defined. So the IO operations performed in the change_gcr_base() method would be accessing the NULL-based memory space. That's why the kernel crash/hanging-up.

Thanks for the analysis!
This path was not taken on my audience hardware, so I didn't catch this, will fix in next version.

Quoted 10 lines

In order to fix this we have to first map the CM GCR block at the default base-address, then update the CM GCR-base CSR and after that remap the CM GCR-space.

Please also note, the GCR_BASE field might be RO. It depends on the IP-core configuration. So it's possible that the CM_GCR_BASE_GCRBASE field update won't work. Although that will be detected a bit later in the mips_cm_probe() method by comparing the address returned from mips_cm_phys_base() and retrieved from the CM GCR-base CSR.

Hmm, I just checked RTL and RDL for CM2 and CM3 and I didn't see it as a configurable option. It's possible to change hardware reset value but not make it RO.

Maybe it was possible on earlier IP release, in this case it's always user's responsibility to write correct address in DeviceTree :-)

Thanks

Quoted 2 lines

-Serge(y)

Signature · 1 line
- Jiaxun
#11 lore raw

Re: [PATCH v2 6/6] MIPS: cm: Probe GCR address from DeviceTree

Quoted 35 lines · Jiaxun Yang

在2024年9月10日九月 下午1:36,Serge Semin写道:
[...]

Quoted 6 lines

This causes the kernel boot-up procedure to crash/hang-up because the CM GCR base address is supposed to be redefined by means of the already mapped CM GCR address space by accessing the CM_GCR_BASE_GCRBASE register:
change_gcr_base()

Quoted 8 lines
+-> read_gcr_base()
    +-> addr_gcr_base()
        +-> return mips_gcr_base + CM_GCR_BASE_GCRBASE

By the time of the change_gcr_base() call in mips_cm_phys_base(), the mips_gcr_base variable hasn't been defined. So the IO operations performed in the change_gcr_base() method would be accessing the NULL-based memory space. That's why the kernel crash/hanging-up.

Thanks for the analysis!
This path was not taken on my audience hardware, so I didn't catch this, will fix in next version.

Quoted 10 lines

In order to fix this we have to first map the CM GCR block at the default base-address, then update the CM GCR-base CSR and after that remap the CM GCR-space.

Please also note, the GCR_BASE field might be RO. It depends on the IP-core configuration. So it's possible that the CM_GCR_BASE_GCRBASE field update won't work. Although that will be detected a bit later in the mips_cm_probe() method by comparing the address returned from mips_cm_phys_base() and retrieved from the CM GCR-base CSR.

Quoted 2 lines

Hmm, I just checked RTL and RDL for CM2 and CM3 and I didn't see it as a configurable option. It's possible to change hardware reset value but not make it RO.

Both MIPS P5600 and P6600 databooks define the GCR_BASE field as optionally R/W:

GCR_BASE 31:15 This field sets the base address of the 32KB R or R/W
GCR block of the P5600 MPS. (IP Config-
This register has a fixed value after reset if uration)
configured as Read-Only (an IP Configuration Option).

Quoted 2 lines

Maybe it was possible on earlier IP release,

I found the text above in the latest MIPS Warrior P-class software manuals downloaded from the ImagTech site. Not sure why your RTL code doesn't have such configs.

Quoted 2 lines

in this case it's always
user's responsibility to write correct address in DeviceTree :-)

Right. The system just won't work if the CM GCR base address couldn't be updated.

-Serge(y)

Quoted 8 lines

Thanks

Quoted 2 lines

-Serge(y)

Signature · 1 line
- Jiaxun
#12 lore raw

Re: [PATCH v2 6/6] MIPS: cm: Probe GCR address from DeviceTree

在2024年9月10日九月 下午9:07,Serge Semin写道:
[...]

Quoted 8 lines

Both MIPS P5600 and P6600 databooks define the GCR_BASE field as optionally R/W:

GCR_BASE 31:15 This field sets the base address of the 32KB R or R/W
GCR block of the P5600 MPS. (IP Config-
This register has a fixed value after reset if uration)
configured as Read-Only (an IP Configuration Option).

Thanks for the pointer, I traced code history and it seems like MIPS decided to not expose this functionality at some point, but documents were not updated.

Maybe I should add a read back check here.

Thanks

Signature · 1 line
- Jiaxun
#13 lore raw

Re: [PATCH v2 6/6] MIPS: cm: Probe GCR address from DeviceTree

Quoted 14 lines · Jiaxun Yang

在2024年9月10日九月 下午9:07,Serge Semin写道:
[...]

Quoted 8 lines

Both MIPS P5600 and P6600 databooks define the GCR_BASE field as optionally R/W:

GCR_BASE 31:15 This field sets the base address of the 32KB R or R/W
GCR block of the P5600 MPS. (IP Config-
This register has a fixed value after reset if uration)
configured as Read-Only (an IP Configuration Option).

Quoted 2 lines

Thanks for the pointer, I traced code history and it seems like MIPS decided to not expose this functionality at some point, but documents were not updated.

Got it. Thanks for clarification.

Quoted 2 lines

Maybe I should add a read back check here.

The check is already implemented in the mips_cm_probe() method. Just 10 lines below the mips_cm_phys_base() method call.

-Serge(y)

Quoted 4 lines

Thanks

Signature · 1 line
- Jiaxun
#14 lore raw
In reply to Jiaxun Yang
Quoted 7 lines · 在2024年6月12日六月 下午6:08,Jiaxun Yang写道:

Hi all,

This series enabled mips-cm code to probe GCR address from devicetree.

This feature has been implemented in MIPS's out-of-tree kernel for a while, and MIPS's u-boot fork on boston will generate required "mti,mips-cm" node as well.

Folks, any comments on this series?

Thanks

Quoted 36 lines

Please review.
Thanks

---
Changes in v2:
- Fix probe order on malta (Serge)
- dt binding improvements (Conor)
- Build warning fix
- Link to v1:
https://lore.kernel.org/r/[email protected]

---
Jiaxun Yang (6):
MIPS: generic: Do __dt_setup_arch in prom_init
MIPS: malta: Move SMP initialisation to device_tree_init
MIPS: cm: Prefix probe functions with __init
MIPS: Move mips_cm_probe after prom_init
dt-bindings: mips: Document mti,mips-cm
MIPS: cm: Probe GCR address from DeviceTree

 .../devicetree/bindings/mips/mti,mips-cm.yaml      | 38 ++++++++++++
 arch/mips/generic/init.c                           |  9 ++-
 arch/mips/include/asm/mips-cm.h                    |  4 +-
 arch/mips/kernel/mips-cm.c                         | 69 ++++++++++++++++++----
 arch/mips/kernel/setup.c                           |  2 +-
 arch/mips/mti-malta/malta-init.c                   |  8 ++-
 6 files changed, 111 insertions(+), 19 deletions(-)

---
base-commit: 2b84edefcad14934796fad37b16512b6a2ca467e
change-id: 20240506-cm_probe-0c667c8b63bf

Best regards,

Signature · 1 line
Jiaxun Yang <[email protected]>
Signature · 1 line
- Jiaxun
#15 lore raw
In reply to Jiaxun Yang
Quoted 10 lines · 在2024年6月12日六月 下午6:08,Jiaxun Yang写道:

Hi all,

This series enabled mips-cm code to probe GCR address from devicetree.

This feature has been implemented in MIPS's out-of-tree kernel for a while, and MIPS's u-boot fork on boston will generate required "mti,mips-cm" node as well.

Please review.
Thanks

Ping on this?

Thanks

Quoted 33 lines

---
Changes in v2:
- Fix probe order on malta (Serge)
- dt binding improvements (Conor)
- Build warning fix
- Link to v1:
https://lore.kernel.org/r/[email protected]

---
Jiaxun Yang (6):
MIPS: generic: Do __dt_setup_arch in prom_init
MIPS: malta: Move SMP initialisation to device_tree_init
MIPS: cm: Prefix probe functions with __init
MIPS: Move mips_cm_probe after prom_init
dt-bindings: mips: Document mti,mips-cm
MIPS: cm: Probe GCR address from DeviceTree

 .../devicetree/bindings/mips/mti,mips-cm.yaml      | 38 ++++++++++++
 arch/mips/generic/init.c                           |  9 ++-
 arch/mips/include/asm/mips-cm.h                    |  4 +-
 arch/mips/kernel/mips-cm.c                         | 69 ++++++++++++++++++----
 arch/mips/kernel/setup.c                           |  2 +-
 arch/mips/mti-malta/malta-init.c                   |  8 ++-
 6 files changed, 111 insertions(+), 19 deletions(-)

---
base-commit: 2b84edefcad14934796fad37b16512b6a2ca467e
change-id: 20240506-cm_probe-0c667c8b63bf

Best regards,

Signature · 1 line
Jiaxun Yang <[email protected]>
Signature · 1 line
- Jiaxun
#16 lore raw
In reply to Jiaxun Yang

Hi Jiaxun

Quoted 10 lines · Jiaxun Yang

Hi all,

This series enabled mips-cm code to probe GCR address from devicetree.

This feature has been implemented in MIPS's out-of-tree kernel for a while, and MIPS's u-boot fork on boston will generate required "mti,mips-cm" node as well.

Please review.
Thanks

Got this tested on my P5600-based SoC implemented as non-generic platform. Alas the system hangs up on the early boot-up stage with no even a single char printed to the console. I'll be able to get back to the problem debugging on the next week.

-Serge(y)

Quoted 33 lines

---
Changes in v2:
- Fix probe order on malta (Serge)
- dt binding improvements (Conor)
- Build warning fix
- Link to v1: https://lore.kernel.org/r/[email protected]

---
Jiaxun Yang (6):
MIPS: generic: Do __dt_setup_arch in prom_init
MIPS: malta: Move SMP initialisation to device_tree_init
MIPS: cm: Prefix probe functions with __init
MIPS: Move mips_cm_probe after prom_init
dt-bindings: mips: Document mti,mips-cm
MIPS: cm: Probe GCR address from DeviceTree

 .../devicetree/bindings/mips/mti,mips-cm.yaml      | 38 ++++++++++++
 arch/mips/generic/init.c                           |  9 ++-
 arch/mips/include/asm/mips-cm.h                    |  4 +-
 arch/mips/kernel/mips-cm.c                         | 69 ++++++++++++++++++----
 arch/mips/kernel/setup.c                           |  2 +-
 arch/mips/mti-malta/malta-init.c                   |  8 ++-
 6 files changed, 111 insertions(+), 19 deletions(-)

---
base-commit: 2b84edefcad14934796fad37b16512b6a2ca467e
change-id: 20240506-cm_probe-0c667c8b63bf

Best regards,

Signature · 1 line
Jiaxun Yang <[email protected]>
#17 lore raw
Quoted 11 lines · Serge Semin

Hi Jiaxun

Quoted 8 lines · Jiaxun Yang

Hi all,

This series enabled mips-cm code to probe GCR address from devicetree.

This feature has been implemented in MIPS's out-of-tree kernel for a while, and MIPS's u-boot fork on boston will generate required "mti,mips-cm" node as well.

Quoted 7 lines
Quoted 2 lines

Please review.
Thanks

Got this tested on my P5600-based SoC implemented as non-generic platform. Alas the system hangs up on the early boot-up stage with no even a single char printed to the console. I'll be able to get back to the problem debugging on the next week.

any news about that ?

Thomas.

Signature · 2 lines
Crap can work. Given enough thrust pigs will fly, but it's not necessarily a
good idea.                                                [ RFC1925, 2.3 ]
#18 lore raw

Hi Thomas

Quoted 15 lines · Thomas Bogendoerfer
Quoted 14 lines · Serge Semin

Hi Jiaxun

Quoted 10 lines · Jiaxun Yang

Hi all,

This series enabled mips-cm code to probe GCR address from devicetree.

This feature has been implemented in MIPS's out-of-tree kernel for a while, and MIPS's u-boot fork on boston will generate required "mti,mips-cm" node as well.

Please review.
Thanks

Quoted 6 lines
Quoted 4 lines

Got this tested on my P5600-based SoC implemented as non-generic platform. Alas the system hangs up on the early boot-up stage with no even a single char printed to the console. I'll be able to get back to the problem debugging on the next week.

any news about that ?

Oops. This patch set has absolutely slipped out of my mind. I am getting back to it immediately and will submit the debug status shortly after I dig out the reason of the hanging up. Sorry for the inconvenience.

-Serge(y)

Quoted 6 lines

Thomas.

Signature · 2 lines
Crap can work. Given enough thrust pigs will fly, but it's not necessarily a
good idea.                                                [ RFC1925, 2.3 ]
#19 lore raw
Quoted 23 lines · Serge Semin

Hi Thomas

Quoted 20 lines · Thomas Bogendoerfer
Quoted 18 lines · Serge Semin

Hi Jiaxun

Quoted 10 lines · Jiaxun Yang

Hi all,

This series enabled mips-cm code to probe GCR address from devicetree.

This feature has been implemented in MIPS's out-of-tree kernel for a while, and MIPS's u-boot fork on boston will generate required "mti,mips-cm" node as well.

Please review.
Thanks

Got this tested on my P5600-based SoC implemented as non-generic platform. Alas the system hangs up on the early boot-up stage with no even a single char printed to the console. I'll be able to get back to the problem debugging on the next week.

Quoted 6 lines
Quoted 1 line

any news about that ?

Oops. This patch set has absolutely slipped out of my mind. I am getting back to it immediately and will submit the debug status shortly after I dig out the reason of the hanging up. Sorry for the inconvenience.

Found the reason of the problem on my platform. It was due to the too early change_gcs_control() invocation. Since mips_cm_probe() is now called after the prom_init() method the later function can't access any CM-register. So for the system to boot up I had to move the GCR controler register update to the plat_mem_setup() method in my platform code. After that the system booted up successfully. Double checked the rest of the platforms in the vanilla kernel repo for having the similar issue. It seems to me there is no platform left in the kernel with such potential problem presented.

But then I decided to test out the actual GCR-base address setup procedure implemented in this patch set, and found another problem unrelated to my platform. I'll submit the problem summary in reply to the respective patch in this series.

-Serge(y)

Quoted 9 lines

-Serge(y)

Quoted 6 lines

Thomas.

Signature · 2 lines
Crap can work. Given enough thrust pigs will fly, but it's not necessarily a
good idea.                                                [ RFC1925, 2.3 ]