Skip to contentSkip to conversation

acpi/apei: Add devm_ghes_register_vendor_record_notifier()

PATCH Superseded Kai-Heng Feng 5 participants 17 messages Mar 19 – Mar 25

Subsystems (5)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/3 acpi/apei: Add devm_ghes_register_vendor_record_notifier()

Add a device-managed wrapper around ghes_register_vendor_record_notifier() so drivers can avoid manual cleanup on device removal or probe failure.

Cc Jonathan Cameron <[email protected]>

---
v2:
- New patch.

 drivers/acpi/apei/ghes.c | 25 +++++++++++++++++++++++++
 include/acpi/ghes.h      |  3 +++
 2 files changed, 28 insertions(+)
drivers/acpi/apei/ghes.c +25 -0
index 8acd2742bb27..d31a70a05538 100644
@@ -689,6 +689,31 @@ void ghes_unregister_vendor_record_notifier(struct notifier_block *nb)
 }
 EXPORT_SYMBOL_GPL(ghes_unregister_vendor_record_notifier);
 
+static void ghes_vendor_record_notifier_destroy(void *nb)
+{
+	ghes_unregister_vendor_record_notifier(nb);
+}
+
+/**
+ * devm_ghes_register_vendor_record_notifier - device-managed vendor record notifier registration
+ * @dev: device that owns the notifier lifetime
+ * @nb: pointer to the notifier_block structure of the vendor record handler
+ *
+ * Return: 0 on success, negative errno on failure.
+ */
+int devm_ghes_register_vendor_record_notifier(struct device *dev,
+					      struct notifier_block *nb)
+{
+	int ret;
+
+	ret = ghes_register_vendor_record_notifier(nb);
+	if (ret)
+		return ret;
+
+	return devm_add_action_or_reset(dev, ghes_vendor_record_notifier_destroy, nb);
+}
+EXPORT_SYMBOL_GPL(devm_ghes_register_vendor_record_notifier);
+
 static void ghes_vendor_record_work_func(struct work_struct *work)
 {
 	struct ghes_vendor_record_entry *entry;
include/acpi/ghes.h +3 -0
index 7bea522c0657..ca3ace828c1c 100644
@@ -71,6 +71,9 @@ int ghes_register_vendor_record_notifier(struct notifier_block *nb);
  */
 void ghes_unregister_vendor_record_notifier(struct notifier_block *nb);
 
+int devm_ghes_register_vendor_record_notifier(struct device *dev,
+					      struct notifier_block *nb);
+
 struct list_head *ghes_get_devices(void);
 
 void ghes_estatus_pool_region_free(unsigned long addr, u32 size);
Signature · 1 line
2.50.1 (Apple Git-155)
#1 lore raw

PATCH v2 2/3 PCI: hisi: Use devm_ghes_register_vendor_record_notifier()

Switch to the device-managed variant so the notifier is automatically unregistered on device removal, allowing the open-coded remove callback to be dropped entirely.

Cc Jonathan Cameron <[email protected]>

---
v2:
- New patch.

 drivers/pci/controller/pcie-hisi-error.c | 12 +-----------
 1 file changed, 1 insertion(+), 11 deletions(-)
drivers/pci/controller/pcie-hisi-error.c +1 -11
index aaf1ed2b6e59..36be86d827a8 100644
@@ -287,25 +287,16 @@ static int hisi_pcie_error_handler_probe(struct platform_device *pdev)
 
 	priv->nb.notifier_call = hisi_pcie_notify_error;
 	priv->dev = &pdev->dev;
-	ret = ghes_register_vendor_record_notifier(&priv->nb);
+	ret = devm_ghes_register_vendor_record_notifier(&pdev->dev, &priv->nb);
 	if (ret) {
 		dev_err(&pdev->dev,
 			"Failed to register hisi pcie controller error handler with apei\n");
 		return ret;
 	}
 
-	platform_set_drvdata(pdev, priv);
-
 	return 0;
 }
 
-static void hisi_pcie_error_handler_remove(struct platform_device *pdev)
-{
-	struct hisi_pcie_error_private *priv = platform_get_drvdata(pdev);
-
-	ghes_unregister_vendor_record_notifier(&priv->nb);
-}
-
 static const struct acpi_device_id hisi_pcie_acpi_match[] = {
 	{ "HISI0361", 0 },
 	{ }
@@ -317,7 +308,6 @@ static struct platform_driver hisi_pcie_error_handler_driver = {
 		.acpi_match_table = hisi_pcie_acpi_match,
 	},
 	.probe		= hisi_pcie_error_handler_probe,
-	.remove		= hisi_pcie_error_handler_remove,
 };
 module_platform_driver(hisi_pcie_error_handler_driver);
 
Signature · 1 line
2.50.1 (Apple Git-155)
Reviewed-by on v2 2/3
#2 lore raw

Re: [PATCH v2 2/3] PCI: hisi: Use devm_ghes_register_vendor_record_notifier()

Quoted 6 lines · On Thu, 19 Mar 2026 19:13:08 +0800 Kai-Heng Feng <[email protected]> wrote:

Switch to the device-managed variant so the notifier is automatically unregistered on device removal, allowing the open-coded remove callback to be dropped entirely.

Cc Jonathan Cameron <[email protected]>

+CC Shiju.

Quoted 46 lines

---
v2:
- New patch.

 drivers/pci/controller/pcie-hisi-error.c | 12 +-----------
 1 file changed, 1 insertion(+), 11 deletions(-)
drivers/pci/controller/pcie-hisi-error.c +1 -11
index aaf1ed2b6e59..36be86d827a8 100644
@@ -287,25 +287,16 @@ static int hisi_pcie_error_handler_probe(struct platform_device *pdev)
 
 	priv->nb.notifier_call = hisi_pcie_notify_error;
 	priv->dev = &pdev->dev;
-	ret = ghes_register_vendor_record_notifier(&priv->nb);
+	ret = devm_ghes_register_vendor_record_notifier(&pdev->dev, &priv->nb);
 	if (ret) {
 		dev_err(&pdev->dev,
 			"Failed to register hisi pcie controller error handler with apei\n");
 		return ret;
 	}
 
-	platform_set_drvdata(pdev, priv);
-
 	return 0;
 }
 
-static void hisi_pcie_error_handler_remove(struct platform_device *pdev)
-{
-	struct hisi_pcie_error_private *priv = platform_get_drvdata(pdev);
-
-	ghes_unregister_vendor_record_notifier(&priv->nb);
-}
-
 static const struct acpi_device_id hisi_pcie_acpi_match[] = {
 	{ "HISI0361", 0 },
 	{ }
@@ -317,7 +308,6 @@ static struct platform_driver hisi_pcie_error_handler_driver = {
 		.acpi_match_table = hisi_pcie_acpi_match,
 	},
 	.probe		= hisi_pcie_error_handler_probe,
-	.remove		= hisi_pcie_error_handler_remove,
 };
 module_platform_driver(hisi_pcie_error_handler_driver);
 
#3 lore raw

PATCH v2 3/3 acpi/apei: Add NVIDIA GHES vendor CPER record handler

In reply to Kai-Heng Feng

Add support for decoding NVIDIA-specific CPER sections delivered via the APEI GHES vendor record notifier chain. NVIDIA hardware generates vendor-specific CPER sections containing error signatures and diagnostic register dumps. This implementation registers a notifier_block with the GHES vendor record notifier and decodes these sections, printing error details via dev_info().

The driver binds to ACPI device NVDA2012, present on NVIDIA server platforms. The NVIDIA CPER section contains a fixed header with error metadata (signature, error type, severity, socket) followed by variable-length register address-value pairs for hardware diagnostics.

This work is based on libcper [0].

Example output:
nvidia-ghes NVDA2012:00: NVIDIA CPER section, error_data_length: 544 nvidia-ghes NVDA2012:00: signature: CMET-INFO
nvidia-ghes NVDA2012:00: error_type: 0
nvidia-ghes NVDA2012:00: error_instance: 0
nvidia-ghes NVDA2012:00: severity: 3
nvidia-ghes NVDA2012:00: socket: 0
nvidia-ghes NVDA2012:00: number_regs: 32
nvidia-ghes NVDA2012:00: instance_base: 0x0000000000000000
nvidia-ghes NVDA2012:00: register[0]: address=0x8000000100000000 value=0x0000000100000000

[0] https://github.com/openbmc/libcper/commit/683e055061ce

Cc Jonathan Cameron <[email protected]>
Cc Shiju Jose <[email protected]>

---
v2:
- Use right headers.
- Use embedded struct and __counted_by.
- Drop __packed.
- Remove unecessary casts.
- Use * in sizeof().
- Use devm_kmalloc() and struct assignment.
- Use dev_err_probe and new devm helper.

 MAINTAINERS                     |   6 ++
 drivers/acpi/apei/Kconfig       |  14 +++
 drivers/acpi/apei/Makefile      |   1 +
 drivers/acpi/apei/nvidia-ghes.c | 146 ++++++++++++++++++++++++++++++++
 4 files changed, 167 insertions(+)
 create mode 100644 drivers/acpi/apei/nvidia-ghes.c
MAINTAINERS +6 -0
index d7241695df96..a9be03fbf1a9 100644
@@ -18908,6 +18908,12 @@ S:	Maintained
 F:	drivers/video/fbdev/nvidia/
 F:	drivers/video/fbdev/riva/
 
+NVIDIA GHES VENDOR CPER RECORD HANDLER
+M:	Kai-Heng Feng <[email protected]>
+L:	[email protected]
+S:	Maintained
+F:	drivers/acpi/apei/nvidia-ghes.c
+
 NVIDIA VRS RTC DRIVER
 M:	Shubhi Garg <[email protected]>
 L:	[email protected]
drivers/acpi/apei/Kconfig +14 -0
index 070c07d68dfb..7dc49f14f223 100644
@@ -74,6 +74,20 @@ config ACPI_APEI_EINJ_CXL
 
 	  If unsure say 'n'
 
+config ACPI_APEI_NVIDIA_GHES
+	tristate "NVIDIA GHES vendor record handler"
+	depends on ACPI_APEI_GHES
+	help
+	  Support for decoding NVIDIA-specific CPER sections delivered via
+	  the APEI GHES vendor record notifier chain. Registers a handler
+	  for the NVIDIA section GUID and logs error signatures, severity,
+	  socket, and diagnostic register address-value pairs.
+
+	  Enable on NVIDIA server platforms (e.g. DGX, HGX) that expose
+	  ACPI device NVDA2012 in their firmware tables.
+
+	  If unsure, say N.
+
 config ACPI_APEI_ERST_DEBUG
 	tristate "APEI Error Record Serialization Table (ERST) Debug Support"
 	depends on ACPI_APEI
drivers/acpi/apei/Makefile +1 -0
index 1a0b85923cd4..4a883f67d698 100644
@@ -10,5 +10,6 @@ obj-$(CONFIG_ACPI_APEI_EINJ)	+= einj.o
 einj-y				:= einj-core.o
 einj-$(CONFIG_ACPI_APEI_EINJ_CXL) += einj-cxl.o
 obj-$(CONFIG_ACPI_APEI_ERST_DEBUG) += erst-dbg.o
+obj-$(CONFIG_ACPI_APEI_NVIDIA_GHES) += nvidia-ghes.o
 
 apei-y := apei-base.o hest.o erst.o bert.o
drivers/acpi/apei/nvidia-ghes.c new file +146 -0

146 lines not on this page · raw message

Signature · 1 line
2.50.1 (Apple Git-155)
Reviewed-by on v2 3/3
#4 lore raw

Re: [PATCH v2 3/3] acpi/apei: Add NVIDIA GHES vendor CPER record handler

Quoted 25 lines · On Thu, 19 Mar 2026 19:13:09 +0800 Kai-Heng Feng <[email protected]> wrote:

Add support for decoding NVIDIA-specific CPER sections delivered via the APEI GHES vendor record notifier chain. NVIDIA hardware generates vendor-specific CPER sections containing error signatures and diagnostic register dumps. This implementation registers a notifier_block with the GHES vendor record notifier and decodes these sections, printing error details via dev_info().

The driver binds to ACPI device NVDA2012, present on NVIDIA server platforms. The NVIDIA CPER section contains a fixed header with error metadata (signature, error type, severity, socket) followed by variable-length register address-value pairs for hardware diagnostics.

This work is based on libcper [0].

Example output:
nvidia-ghes NVDA2012:00: NVIDIA CPER section, error_data_length: 544 nvidia-ghes NVDA2012:00: signature: CMET-INFO
nvidia-ghes NVDA2012:00: error_type: 0
nvidia-ghes NVDA2012:00: error_instance: 0
nvidia-ghes NVDA2012:00: severity: 3
nvidia-ghes NVDA2012:00: socket: 0
nvidia-ghes NVDA2012:00: number_regs: 32
nvidia-ghes NVDA2012:00: instance_base: 0x0000000000000000
nvidia-ghes NVDA2012:00: register[0]: address=0x8000000100000000 value=0x0000000100000000

Quoted 4 lines

[0] https://github.com/openbmc/libcper/commit/683e055061ce

Cc Jonathan Cameron <[email protected]>
Cc Shiju Jose <[email protected]>

Only significant thing is around use of dev_err_probe().

I'm surprised that didn't give you error messages in the log even on success.

With that fixed (other stuff is all up to you).

Quoted 14 lines

apei-y := apei-base.o hest.o erst.o bert.o

drivers/acpi/apei/nvidia-ghes.c new file +7 -0
new file mode 100644
index 000000000000..aa2e3a387b49
 
+static void nvidia_ghes_print_error(struct device *dev,
+				    const struct cper_sec_nvidia *nvidia_err,
+				    size_t error_data_length, bool fatal)
+{
+	const char *level = fatal ? KERN_ERR : KERN_INFO;
+	size_t min_size;
+	int i;

...

drivers/acpi/apei/nvidia-ghes.c line 70

Quoted 11 lines
drivers/acpi/apei/nvidia-ghes.c new file +11 -0
+	 * Validate that all registers fit within error_data_length.
+	 * Each register pair is two little-endian u64s.
+	 */
+	min_size = struct_size(nvidia_err, regs, nvidia_err->number_regs);
+	if (error_data_length < min_size) {
+		dev_err(dev, "Invalid number_regs %u (section size %zu, need %zu)\n",
+			nvidia_err->number_regs, error_data_length, min_size);
+		return;
+	}
+
+	for (i = 0; i < nvidia_err->number_regs; i++)

Trivial but I'd take advantage of it now being acceptable (in general) to do

	for (int i = 0; i < ....)
Quoted 10 lines
drivers/acpi/apei/nvidia-ghes.c new file +9 -0
+		dev_printk(level, dev, "register[%d]: address=0x%016llx value=0x%016llx\n",
+			   i, le64_to_cpu(nvidia_err->regs[i].addr),
+			   le64_to_cpu(nvidia_err->regs[i].val));
+}
 
+static int nvidia_ghes_probe(struct platform_device *pdev)
+{
+	struct nvidia_ghes_private *priv;
+
+	priv = devm_kmalloc(&pdev->dev, sizeof(*priv), GFP_KERNEL);

drivers/acpi/apei/nvidia-ghes.c line 125

Quoted 10 lines
drivers/acpi/apei/nvidia-ghes.c new file +10 -0
+	if (!priv)
+		return -ENOMEM;
+
+	*priv = (struct nvidia_ghes_private) {
+		.nb.notifier_call = nvidia_ghes_notify,
+		.dev = &pdev->dev,
+	};
+
+	return dev_err_probe(&pdev->dev,
+			     devm_ghes_register_vendor_record_notifier(&pdev->dev, &priv->nb),

That's too not great for readability and dev_err_probe() should only be called on errors I'm fairly sure it doesn't have special handling for 0 so will call dev_err() or dev_warn() and print some stuff before saying 'no error'.

	int ret;
	...

	ret = devm_ghes_register_vendor_record_notifier(&pdev->dev, &priv->nb);
	if (ret)
		return dev_err_probe(&pdev->dev,
				      "Failed to register NVIDIA GHES vendor record notifier\n");

	return 0;

drivers/acpi/apei/nvidia-ghes.c line 130

Quoted 5 lines
drivers/acpi/apei/nvidia-ghes.c new file +5 -0
+			     "Failed to register NVIDIA GHES vendor record notifier\n");
+}
+
+static const struct acpi_device_id nvidia_ghes_acpi_match[] = {
+	{ "NVDA2012" },

London Olympics :)

drivers/acpi/apei/nvidia-ghes.c line 140

Quoted 10 lines
drivers/acpi/apei/nvidia-ghes.c new file +10 -0
+	{ }
+};
+MODULE_DEVICE_TABLE(acpi, nvidia_ghes_acpi_match);
+
+static struct platform_driver nvidia_ghes_driver = {
+	.driver = {
+		.name		= "nvidia-ghes",
+		.acpi_match_table = nvidia_ghes_acpi_match,
+	},
+	.probe	= nvidia_ghes_probe,

I'd just not attempt to align the =

static struct platform_driver nvidia_ghes_driver = {
	.driver = {
		.name = "nvidia-ghes",
		.acpi_match_table = nvidia_ghes_acpi_match,
	},
	.probe = nvidia_ghes_probe,

There aren't enough of them to make it much of a readability improvement and doing this often results in unnecessary churn as a driver evolves. Also it's already broken!

Quoted 6 lines
drivers/acpi/apei/nvidia-ghes.c new file +6 -0
+};
+module_platform_driver(nvidia_ghes_driver);
+
+MODULE_AUTHOR("Kai-Heng Feng <[email protected]>");
+MODULE_DESCRIPTION("NVIDIA GHES vendor CPER record handler");
+MODULE_LICENSE("GPL");
#5 lore raw

Re: [PATCH v2 3/3] acpi/apei: Add NVIDIA GHES vendor CPER record handler

Quoted 72 lines · Jonathan Cameron

External email: Use caution opening links or attachments

Quoted 25 lines · On Thu, 19 Mar 2026 19:13:09 +0800 Kai-Heng Feng <[email protected]> wrote:

Add support for decoding NVIDIA-specific CPER sections delivered via the APEI GHES vendor record notifier chain. NVIDIA hardware generates vendor-specific CPER sections containing error signatures and diagnostic register dumps. This implementation registers a notifier_block with the GHES vendor record notifier and decodes these sections, printing error details via dev_info().

The driver binds to ACPI device NVDA2012, present on NVIDIA server platforms. The NVIDIA CPER section contains a fixed header with error metadata (signature, error type, severity, socket) followed by variable-length register address-value pairs for hardware diagnostics.

This work is based on libcper [0].

Example output:
nvidia-ghes NVDA2012:00: NVIDIA CPER section, error_data_length: 544 nvidia-ghes NVDA2012:00: signature: CMET-INFO
nvidia-ghes NVDA2012:00: error_type: 0
nvidia-ghes NVDA2012:00: error_instance: 0
nvidia-ghes NVDA2012:00: severity: 3
nvidia-ghes NVDA2012:00: socket: 0
nvidia-ghes NVDA2012:00: number_regs: 32
nvidia-ghes NVDA2012:00: instance_base: 0x0000000000000000
nvidia-ghes NVDA2012:00: register[0]: address=0x8000000100000000 value=0x0000000100000000

Quoted 4 lines

[0] https://github.com/openbmc/libcper/commit/683e055061ce

Cc Jonathan Cameron <[email protected]>
Cc Shiju Jose <[email protected]>

Only significant thing is around use of dev_err_probe().

I'm surprised that didn't give you error messages in the log even on success.

With that fixed (other stuff is all up to you).

Quoted 14 lines

apei-y := apei-base.o hest.o erst.o bert.o

drivers/acpi/apei/nvidia-ghes.c new file +7 -0
new file mode 100644
index 000000000000..aa2e3a387b49
 
+static void nvidia_ghes_print_error(struct device *dev,
+                                 const struct cper_sec_nvidia *nvidia_err,
+                                 size_t error_data_length, bool fatal)
+{
+     const char *level = fatal ? KERN_ERR : KERN_INFO;
+     size_t min_size;
+     int i;

...

drivers/acpi/apei/nvidia-ghes.c line 70

Quoted 11 lines
drivers/acpi/apei/nvidia-ghes.c new file +11 -0
+      * Validate that all registers fit within error_data_length.
+      * Each register pair is two little-endian u64s.
+      */
+     min_size = struct_size(nvidia_err, regs, nvidia_err->number_regs);
+     if (error_data_length < min_size) {
+             dev_err(dev, "Invalid number_regs %u (section size %zu, need %zu)\n",
+                     nvidia_err->number_regs, error_data_length, min_size);
+             return;
+     }
+
+     for (i = 0; i < nvidia_err->number_regs; i++)
Quoted 2 lines

Trivial but I'd take advantage of it now being acceptable (in general) to do

        for (int i = 0; i < ....)

Didn't know it's acceptable now. Will change.

Quoted 25 lines
Quoted 11 lines
drivers/acpi/apei/nvidia-ghes.c new file +9 -0
+             dev_printk(level, dev, "register[%d]: address=0x%016llx value=0x%016llx\n",
+                        i, le64_to_cpu(nvidia_err->regs[i].addr),
+                        le64_to_cpu(nvidia_err->regs[i].val));
+}
 
+static int nvidia_ghes_probe(struct platform_device *pdev)
+{
+     struct nvidia_ghes_private *priv;
+
+     priv = devm_kmalloc(&pdev->dev, sizeof(*priv), GFP_KERNEL);

drivers/acpi/apei/nvidia-ghes.c line 125

Quoted 10 lines
drivers/acpi/apei/nvidia-ghes.c new file +10 -0
+     if (!priv)
+             return -ENOMEM;
+
+     *priv = (struct nvidia_ghes_private) {
+             .nb.notifier_call = nvidia_ghes_notify,
+             .dev = &pdev->dev,
+     };
+
+     return dev_err_probe(&pdev->dev,
+                          devm_ghes_register_vendor_record_notifier(&pdev->dev, &priv->nb),

That's too not great for readability and dev_err_probe() should only be called on errors I'm fairly sure it doesn't have special handling for 0 so will call dev_err() or dev_warn() and print some stuff before saying 'no error'.

Quoted 9 lines
        int ret;
        ...

        ret = devm_ghes_register_vendor_record_notifier(&pdev->dev, &priv->nb);
        if (ret)
                return dev_err_probe(&pdev->dev,
                                      "Failed to register NVIDIA GHES vendor record notifier\n");

        return 0;

OK, will change.

Quoted 10 lines

drivers/acpi/apei/nvidia-ghes.c line 130

Quoted 8 lines
drivers/acpi/apei/nvidia-ghes.c new file +5 -0
+                          "Failed to register NVIDIA GHES vendor record notifier\n");
+}
+
+static const struct acpi_device_id nvidia_ghes_acpi_match[] = {
+     { "NVDA2012" },

London Olympics :)

Michael Phelps did great :)

Quoted 20 lines

drivers/acpi/apei/nvidia-ghes.c line 140

Quoted 11 lines
drivers/acpi/apei/nvidia-ghes.c new file +10 -0
+     { }
+};
+MODULE_DEVICE_TABLE(acpi, nvidia_ghes_acpi_match);
+
+static struct platform_driver nvidia_ghes_driver = {
+     .driver = {
+             .name           = "nvidia-ghes",
+             .acpi_match_table = nvidia_ghes_acpi_match,
+     },
+     .probe  = nvidia_ghes_probe,

I'd just not attempt to align the =

static struct platform_driver nvidia_ghes_driver = {
        .driver = {
                .name = "nvidia-ghes",
                .acpi_match_table = nvidia_ghes_acpi_match,
        },
        .probe = nvidia_ghes_probe,
Quoted 3 lines

There aren't enough of them to make it much of a readability improvement and doing this often results in unnecessary churn as a driver evolves. Also it's already broken!

OK, will change too.

Quoted 8 lines
Quoted 7 lines
drivers/acpi/apei/nvidia-ghes.c new file +6 -0
+};
+module_platform_driver(nvidia_ghes_driver);
+
+MODULE_AUTHOR("Kai-Heng Feng <[email protected]>");
+MODULE_DESCRIPTION("NVIDIA GHES vendor CPER record handler");
+MODULE_LICENSE("GPL");
#6 lore raw

Re: [PATCH v2 3/3] acpi/apei: Add NVIDIA GHES vendor CPER record handler

In reply to Kai-Heng Feng
Quoted 14 lines · Kai-Heng Feng

Add support for decoding NVIDIA-specific CPER sections delivered via the APEI GHES vendor record notifier chain. NVIDIA hardware generates vendor-specific CPER sections containing error signatures and diagnostic register dumps. This implementation registers a notifier_block with the GHES vendor record notifier and decodes these sections, printing error details via dev_info().

The driver binds to ACPI device NVDA2012, present on NVIDIA server platforms. The NVIDIA CPER section contains a fixed header with error metadata (signature, error type, severity, socket) followed by variable-length register address-value pairs for hardware diagnostics.

This work is based on libcper [0].

Quoted 10 lines

Example output:
nvidia-ghes NVDA2012:00: NVIDIA CPER section, error_data_length: 544 nvidia-ghes NVDA2012:00: signature: CMET-INFO
nvidia-ghes NVDA2012:00: error_type: 0
nvidia-ghes NVDA2012:00: error_instance: 0
nvidia-ghes NVDA2012:00: severity: 3
nvidia-ghes NVDA2012:00: socket: 0
nvidia-ghes NVDA2012:00: number_regs: 32
nvidia-ghes NVDA2012:00: instance_base: 0x0000000000000000
nvidia-ghes NVDA2012:00: register[0]: address=0x8000000100000000 value=0x0000000100000000

Is there a convenient way to connect NVDA2012:00 with the actual device? I assume this is typically a PCIe device? How would we relate this with PCIe errors?

Consider a cover letter. Some of these comments apply to the series.

Wrap commit logs to fit in 75 columns. When indented by "git log", all of these overflow 80 columns by just a few characters.

Possibly reorder so the acpi/apei patches are together. I don't think the NVIDIA record handler depends on the PCI patch.

Typical subject line style in drivers/acpi/apei appears to be:

ACPI: APEI: GHES: Add ...

drivers/acpi/apei/Kconfig config ACPI_APEI_EINJ_CXL · line 79

Quoted 3 lines
drivers/acpi/apei/Kconfig +3 -0
+config ACPI_APEI_NVIDIA_GHES
+	tristate "NVIDIA GHES vendor record handler"
+	depends on ACPI_APEI_GHES

Maybe s/ACPI_APEI_NVIDIA_GHES/ACPI_APEI_GHES_NVIDIA/ since there will likely be more, and they'll sort nicely if the vendor is at the end.

drivers/acpi/apei/Kconfig config ACPI_APEI_EINJ_CXL · line 87

Quoted 8 lines
drivers/acpi/apei/Kconfig +8 -0
+	help
+	  Support for decoding NVIDIA-specific CPER sections delivered via
+	  the APEI GHES vendor record notifier chain. Registers a handler
+	  for the NVIDIA section GUID and logs error signatures, severity,
+	  socket, and diagnostic register address-value pairs.
+
+	  Enable on NVIDIA server platforms (e.g. DGX, HGX) that expose
+	  ACPI device NVDA2012 in their firmware tables.

Wrap to fit in 80 columns like the rest of this file.

Quoted 1 line

+++ b/drivers/acpi/apei/nvidia-ghes.c

Maybe rename to "ghes-nvidia.c" so future decoders for other vendors are grouped?

#7 lore raw

Re: [PATCH v2 3/3] acpi/apei: Add NVIDIA GHES vendor CPER record handler

Quoted 6 lines · Bjorn Helgaas
Quoted 1 line · Kai-Heng Feng

Add support for decoding NVIDIA-specific CPER sections delivered via

...

Wrap commit logs to fit in 75 columns. When indented by "git log", all of these overflow 80 columns by just a few characters.

I'm so sorry, I had accidentally resized my window narrower. These are all fine; please ignore this and the similar Kconfig comment.

#8 lore raw

Re: [PATCH v2 3/3] acpi/apei: Add NVIDIA GHES vendor CPER record handler

In reply to Bjorn Helgaas
Quoted 29 lines · Bjorn Helgaas

External email: Use caution opening links or attachments

Quoted 14 lines · Kai-Heng Feng

Add support for decoding NVIDIA-specific CPER sections delivered via the APEI GHES vendor record notifier chain. NVIDIA hardware generates vendor-specific CPER sections containing error signatures and diagnostic register dumps. This implementation registers a notifier_block with the GHES vendor record notifier and decodes these sections, printing error details via dev_info().

The driver binds to ACPI device NVDA2012, present on NVIDIA server platforms. The NVIDIA CPER section contains a fixed header with error metadata (signature, error type, severity, socket) followed by variable-length register address-value pairs for hardware diagnostics.

This work is based on libcper [0].

Quoted 10 lines

Example output:
nvidia-ghes NVDA2012:00: NVIDIA CPER section, error_data_length: 544 nvidia-ghes NVDA2012:00: signature: CMET-INFO
nvidia-ghes NVDA2012:00: error_type: 0
nvidia-ghes NVDA2012:00: error_instance: 0
nvidia-ghes NVDA2012:00: severity: 3
nvidia-ghes NVDA2012:00: socket: 0
nvidia-ghes NVDA2012:00: number_regs: 32
nvidia-ghes NVDA2012:00: instance_base: 0x0000000000000000
nvidia-ghes NVDA2012:00: register[0]: address=0x8000000100000000 value=0x0000000100000000

Quoted 3 lines

Is there a convenient way to connect NVDA2012:00 with the actual device? I assume this is typically a PCIe device? How would we relate this with PCIe errors?

The CPER report is from ARM RAS firmware and not neccessarily be related to a PCIe device.

Quoted 2 lines

Consider a cover letter. Some of these comments apply to the series.

Will do in next version.

Quoted 9 lines

Wrap commit logs to fit in 75 columns. When indented by "git log", all of these overflow 80 columns by just a few characters.

Possibly reorder so the acpi/apei patches are together. I don't think the NVIDIA record handler depends on the PCI patch.

Typical subject line style in drivers/acpi/apei appears to be:

Quoted 8 lines

ACPI: APEI: GHES: Add ...

drivers/acpi/apei/Kconfig config ACPI_APEI_EINJ_CXL · line 79

Quoted 3 lines
drivers/acpi/apei/Kconfig +3 -0
+config ACPI_APEI_NVIDIA_GHES
+     tristate "NVIDIA GHES vendor record handler"
+     depends on ACPI_APEI_GHES

Maybe s/ACPI_APEI_NVIDIA_GHES/ACPI_APEI_GHES_NVIDIA/ since there will likely be more, and they'll sort nicely if the vendor is at the end.

OK, will do.

Quoted 10 lines

drivers/acpi/apei/Kconfig config ACPI_APEI_EINJ_CXL · line 87

Quoted 9 lines
drivers/acpi/apei/Kconfig +8 -0
+     help
+       Support for decoding NVIDIA-specific CPER sections delivered via
+       the APEI GHES vendor record notifier chain. Registers a handler
+       for the NVIDIA section GUID and logs error signatures, severity,
+       socket, and diagnostic register address-value pairs.
+
+       Enable on NVIDIA server platforms (e.g. DGX, HGX) that expose
+       ACPI device NVDA2012 in their firmware tables.
Quoted 6 lines

Wrap to fit in 80 columns like the rest of this file.

Quoted 1 line

+++ b/drivers/acpi/apei/nvidia-ghes.c

Maybe rename to "ghes-nvidia.c" so future decoders for other vendors are grouped?

Will do.

#9 lore raw

Re: [PATCH v2 3/3] acpi/apei: Add NVIDIA GHES vendor CPER record handler

Quoted 27 lines · Kai-Heng Feng
Quoted 26 lines · Bjorn Helgaas
Quoted 24 lines · Kai-Heng Feng

Add support for decoding NVIDIA-specific CPER sections delivered via the APEI GHES vendor record notifier chain. NVIDIA hardware generates vendor-specific CPER sections containing error signatures and diagnostic register dumps. This implementation registers a notifier_block with the GHES vendor record notifier and decodes these sections, printing error details via dev_info().

The driver binds to ACPI device NVDA2012, present on NVIDIA server platforms. The NVIDIA CPER section contains a fixed header with error metadata (signature, error type, severity, socket) followed by variable-length register address-value pairs for hardware diagnostics.

This work is based on libcper [0].

Example output:
nvidia-ghes NVDA2012:00: NVIDIA CPER section, error_data_length: 544 nvidia-ghes NVDA2012:00: signature: CMET-INFO
nvidia-ghes NVDA2012:00: error_type: 0
nvidia-ghes NVDA2012:00: error_instance: 0
nvidia-ghes NVDA2012:00: severity: 3
nvidia-ghes NVDA2012:00: socket: 0
nvidia-ghes NVDA2012:00: number_regs: 32
nvidia-ghes NVDA2012:00: instance_base: 0x0000000000000000
nvidia-ghes NVDA2012:00: register[0]: address=0x8000000100000000 value=0x0000000100000000

Quoted 6 lines
Quoted 3 lines

Is there a convenient way to connect NVDA2012:00 with the actual device? I assume this is typically a PCIe device? How would we relate this with PCIe errors?

The CPER report is from ARM RAS firmware and not neccessarily be related to a PCIe device.

Right, I know CPER is more general than just PCI/PCIe.

But in this case, I think NVDA2012 probably *is* a PCIe device. How would we figure out which one? If we have to manually do an acpidump, figure out which NVDA2012 is :00, and look for an _ADR or something, that doesn't really seem convenient for multi-NVDA2012 situations.

#10 lore raw

Re: [PATCH v2 3/3] acpi/apei: Add NVIDIA GHES vendor CPER record handler

Quoted 38 lines · Bjorn Helgaas

External email: Use caution opening links or attachments

Quoted 27 lines · Kai-Heng Feng
Quoted 26 lines · Bjorn Helgaas
Quoted 24 lines · Kai-Heng Feng

Add support for decoding NVIDIA-specific CPER sections delivered via the APEI GHES vendor record notifier chain. NVIDIA hardware generates vendor-specific CPER sections containing error signatures and diagnostic register dumps. This implementation registers a notifier_block with the GHES vendor record notifier and decodes these sections, printing error details via dev_info().

The driver binds to ACPI device NVDA2012, present on NVIDIA server platforms. The NVIDIA CPER section contains a fixed header with error metadata (signature, error type, severity, socket) followed by variable-length register address-value pairs for hardware diagnostics.

This work is based on libcper [0].

Example output:
nvidia-ghes NVDA2012:00: NVIDIA CPER section, error_data_length: 544 nvidia-ghes NVDA2012:00: signature: CMET-INFO
nvidia-ghes NVDA2012:00: error_type: 0
nvidia-ghes NVDA2012:00: error_instance: 0
nvidia-ghes NVDA2012:00: severity: 3
nvidia-ghes NVDA2012:00: socket: 0
nvidia-ghes NVDA2012:00: number_regs: 32
nvidia-ghes NVDA2012:00: instance_base: 0x0000000000000000
nvidia-ghes NVDA2012:00: register[0]: address=0x8000000100000000 value=0x0000000100000000

Quoted 6 lines
Quoted 3 lines

Is there a convenient way to connect NVDA2012:00 with the actual device? I assume this is typically a PCIe device? How would we relate this with PCIe errors?

The CPER report is from ARM RAS firmware and not neccessarily be related to a PCIe device.

Quoted 6 lines

Right, I know CPER is more general than just PCI/PCIe.

But in this case, I think NVDA2012 probably *is* a PCIe device. How would we figure out which one? If we have to manually do an acpidump, figure out which NVDA2012 is :00, and look for an _ADR or something, that doesn't really seem convenient for multi-NVDA2012 situations.

It's actually just an ACPI device:
Device (CPER)
{
Name (_HID, "NVDA2012") // _HID: Hardware ID
Name (_UID, 0x00) // _UID: Unique ID
Method (_DSM, 4, Serialized) // _DSM: Device-Specific Method
}

And that's it.

#11 lore raw

Re: [PATCH v2 3/3] acpi/apei: Add NVIDIA GHES vendor CPER record handler

Quoted 51 lines · Kai-Heng Feng
Quoted 35 lines · Bjorn Helgaas
Quoted 27 lines · Kai-Heng Feng
Quoted 26 lines · Bjorn Helgaas
Quoted 24 lines · Kai-Heng Feng

Add support for decoding NVIDIA-specific CPER sections delivered via the APEI GHES vendor record notifier chain. NVIDIA hardware generates vendor-specific CPER sections containing error signatures and diagnostic register dumps. This implementation registers a notifier_block with the GHES vendor record notifier and decodes these sections, printing error details via dev_info().

The driver binds to ACPI device NVDA2012, present on NVIDIA server platforms. The NVIDIA CPER section contains a fixed header with error metadata (signature, error type, severity, socket) followed by variable-length register address-value pairs for hardware diagnostics.

This work is based on libcper [0].

Example output:
nvidia-ghes NVDA2012:00: NVIDIA CPER section, error_data_length: 544 nvidia-ghes NVDA2012:00: signature: CMET-INFO
nvidia-ghes NVDA2012:00: error_type: 0
nvidia-ghes NVDA2012:00: error_instance: 0
nvidia-ghes NVDA2012:00: severity: 3
nvidia-ghes NVDA2012:00: socket: 0
nvidia-ghes NVDA2012:00: number_regs: 32
nvidia-ghes NVDA2012:00: instance_base: 0x0000000000000000
nvidia-ghes NVDA2012:00: register[0]: address=0x8000000100000000 value=0x0000000100000000

Quoted 6 lines
Quoted 3 lines

Is there a convenient way to connect NVDA2012:00 with the actual device? I assume this is typically a PCIe device? How would we relate this with PCIe errors?

The CPER report is from ARM RAS firmware and not neccessarily be related to a PCIe device.

Quoted 6 lines

Right, I know CPER is more general than just PCI/PCIe.

But in this case, I think NVDA2012 probably *is* a PCIe device. How would we figure out which one? If we have to manually do an acpidump, figure out which NVDA2012 is :00, and look for an _ADR or something, that doesn't really seem convenient for multi-NVDA2012 situations.

It's actually just an ACPI device:
Device (CPER)
{
Name (_HID, "NVDA2012") // _HID: Hardware ID
Name (_UID, 0x00) // _UID: Unique ID
Method (_DSM, 4, Serialized) // _DSM: Device-Specific Method
}

Quoted 1 line

And that's it.

Weird. There's nothing for a driver to operate the device with except _DSM? The device doesn't need any MMIO resources? I would expect some resources described by a _CRS method or some native enumeration protocol like PCI BARs.

The _UID 0x00 matches the "00" in "NVDA2012:00", but I think that's a coincidence; I think the "00" in the device name came from the ida_alloc() in acpi_device_set_name(), not from _UID.

So I still don't know how you would identify the correct part in a system with multiple NVDA2012 devices. I do see the "socket" and "instance_base" in the output. Maybe that would help, but those seem to be device-specific, and it seems like we should have a generic mechanism.

#12 lore raw

Re: [PATCH v2 3/3] acpi/apei: Add NVIDIA GHES vendor CPER record handler

Quoted 59 lines · On Wed, 25 Mar 2026 10:36:28 -0500 Bjorn Helgaas <[email protected]> wrote:
Quoted 51 lines · Kai-Heng Feng
Quoted 35 lines · Bjorn Helgaas
Quoted 27 lines · Kai-Heng Feng
Quoted 26 lines · Bjorn Helgaas
Quoted 24 lines · Kai-Heng Feng

Add support for decoding NVIDIA-specific CPER sections delivered via the APEI GHES vendor record notifier chain. NVIDIA hardware generates vendor-specific CPER sections containing error signatures and diagnostic register dumps. This implementation registers a notifier_block with the GHES vendor record notifier and decodes these sections, printing error details via dev_info().

The driver binds to ACPI device NVDA2012, present on NVIDIA server platforms. The NVIDIA CPER section contains a fixed header with error metadata (signature, error type, severity, socket) followed by variable-length register address-value pairs for hardware diagnostics.

This work is based on libcper [0].

Example output:
nvidia-ghes NVDA2012:00: NVIDIA CPER section, error_data_length: 544 nvidia-ghes NVDA2012:00: signature: CMET-INFO
nvidia-ghes NVDA2012:00: error_type: 0
nvidia-ghes NVDA2012:00: error_instance: 0
nvidia-ghes NVDA2012:00: severity: 3
nvidia-ghes NVDA2012:00: socket: 0
nvidia-ghes NVDA2012:00: number_regs: 32
nvidia-ghes NVDA2012:00: instance_base: 0x0000000000000000
nvidia-ghes NVDA2012:00: register[0]: address=0x8000000100000000 value=0x0000000100000000

Quoted 6 lines
Quoted 3 lines

Is there a convenient way to connect NVDA2012:00 with the actual device? I assume this is typically a PCIe device? How would we relate this with PCIe errors?

The CPER report is from ARM RAS firmware and not neccessarily be related to a PCIe device.

Quoted 6 lines

Right, I know CPER is more general than just PCI/PCIe.

But in this case, I think NVDA2012 probably *is* a PCIe device. How would we figure out which one? If we have to manually do an acpidump, figure out which NVDA2012 is :00, and look for an _ADR or something, that doesn't really seem convenient for multi-NVDA2012 situations.

It's actually just an ACPI device:
Device (CPER)
{
Name (_HID, "NVDA2012") // _HID: Hardware ID
Name (_UID, 0x00) // _UID: Unique ID
Method (_DSM, 4, Serialized) // _DSM: Device-Specific Method
}

Quoted 1 line

And that's it.

Weird. There's nothing for a driver to operate the device with except _DSM? The device doesn't need any MMIO resources? I would expect some resources described by a _CRS method or some native enumeration protocol like PCI BARs.

Quoted 8 lines

The _UID 0x00 matches the "00" in "NVDA2012:00", but I think that's a coincidence; I think the "00" in the device name came from the ida_alloc() in acpi_device_set_name(), not from _UID.

So I still don't know how you would identify the correct part in a system with multiple NVDA2012 devices. I do see the "socket" and "instance_base" in the output. Maybe that would help, but those seem to be device-specific, and it seems like we should have a generic mechanism.

It's not unique in ACPI terms. There are a few cases even in the ACPI spec of IDs that exist just to say some feature is there.

ACPI0017 is an example. Simply says, there be CXL here, go look for the tables.

Here this device is used to indicate that a platform should be ready to handle a particular type of error record. If it happened to expose any other interfaces, then I agree it would need resources or a _DSM etc.

Basically it's a workaround for the lack of discoverability in APEI / ACPI error reporting. Could use an _OSC bit for the same job but then we'd run out of those fast. Device IDs are near free.

Jonathan

Quoted 1 line
#13 lore raw

Re: [PATCH v2 3/3] acpi/apei: Add NVIDIA GHES vendor CPER record handler

Quoted 78 lines · Jonathan Cameron
Quoted 59 lines · On Wed, 25 Mar 2026 10:36:28 -0500 Bjorn Helgaas <[email protected]> wrote:
Quoted 51 lines · Kai-Heng Feng
Quoted 35 lines · Bjorn Helgaas
Quoted 27 lines · Kai-Heng Feng
Quoted 26 lines · Bjorn Helgaas
Quoted 24 lines · Kai-Heng Feng

Add support for decoding NVIDIA-specific CPER sections delivered via the APEI GHES vendor record notifier chain. NVIDIA hardware generates vendor-specific CPER sections containing error signatures and diagnostic register dumps. This implementation registers a notifier_block with the GHES vendor record notifier and decodes these sections, printing error details via dev_info().

The driver binds to ACPI device NVDA2012, present on NVIDIA server platforms. The NVIDIA CPER section contains a fixed header with error metadata (signature, error type, severity, socket) followed by variable-length register address-value pairs for hardware diagnostics.

This work is based on libcper [0].

Example output:
nvidia-ghes NVDA2012:00: NVIDIA CPER section, error_data_length: 544 nvidia-ghes NVDA2012:00: signature: CMET-INFO
nvidia-ghes NVDA2012:00: error_type: 0
nvidia-ghes NVDA2012:00: error_instance: 0
nvidia-ghes NVDA2012:00: severity: 3
nvidia-ghes NVDA2012:00: socket: 0
nvidia-ghes NVDA2012:00: number_regs: 32
nvidia-ghes NVDA2012:00: instance_base: 0x0000000000000000
nvidia-ghes NVDA2012:00: register[0]: address=0x8000000100000000 value=0x0000000100000000

Quoted 6 lines
Quoted 3 lines

Is there a convenient way to connect NVDA2012:00 with the actual device? I assume this is typically a PCIe device? How would we relate this with PCIe errors?

The CPER report is from ARM RAS firmware and not neccessarily be related to a PCIe device.

Quoted 6 lines

Right, I know CPER is more general than just PCI/PCIe.

But in this case, I think NVDA2012 probably *is* a PCIe device. How would we figure out which one? If we have to manually do an acpidump, figure out which NVDA2012 is :00, and look for an _ADR or something, that doesn't really seem convenient for multi-NVDA2012 situations.

It's actually just an ACPI device:
Device (CPER)
{
Name (_HID, "NVDA2012") // _HID: Hardware ID
Name (_UID, 0x00) // _UID: Unique ID
Method (_DSM, 4, Serialized) // _DSM: Device-Specific Method
}

Quoted 1 line

And that's it.

Weird. There's nothing for a driver to operate the device with except _DSM? The device doesn't need any MMIO resources? I would expect some resources described by a _CRS method or some native enumeration protocol like PCI BARs.

Quoted 8 lines

The _UID 0x00 matches the "00" in "NVDA2012:00", but I think that's a coincidence; I think the "00" in the device name came from the ida_alloc() in acpi_device_set_name(), not from _UID.

So I still don't know how you would identify the correct part in a system with multiple NVDA2012 devices. I do see the "socket" and "instance_base" in the output. Maybe that would help, but those seem to be device-specific, and it seems like we should have a generic mechanism.

It's not unique in ACPI terms. There are a few cases even in the ACPI spec of IDs that exist just to say some feature is there.

ACPI0017 is an example. Simply says, there be CXL here, go look for the tables.

Quoted 7 lines

Here this device is used to indicate that a platform should be ready to handle a particular type of error record. If it happened to expose any other interfaces, then I agree it would need resources or a _DSM etc.

Basically it's a workaround for the lack of discoverability in APEI / ACPI error reporting. Could use an _OSC bit for the same job but then we'd run out of those fast. Device IDs are near free.

Well, in principle, an auxiliary device could be registered when a given ACPI table was present. It would then trigger a driver load and the driver would probe against the table in question.

#14 lore raw
In reply to Kai-Heng Feng
Quoted 5 lines · On Thu, 19 Mar 2026 19:13:07 +0800 Kai-Heng Feng <[email protected]> wrote:

Add a device-managed wrapper around ghes_register_vendor_record_notifier() so drivers can avoid manual cleanup on device removal or probe failure.

Cc Jonathan Cameron <[email protected]>

Hi,

My only comment is about following local style. I think that means moving the docs to the header. Unfortunately whether things are in the header or the c file is a subsystem specific thing.

My preference is in the c file, but local style overrides that! Better to have all the docs in the same place.

Jonathan

Quoted 13 lines

---
v2:
- New patch.

 drivers/acpi/apei/ghes.c | 25 +++++++++++++++++++++++++
 include/acpi/ghes.h      |  3 +++
 2 files changed, 28 insertions(+)
b/ renamed from drivers/acpi/apei/ghes.c +0 -0
index 8acd2742bb27..d31a70a05538 100644

b/ void ghes_unregister_vendor_record_notifier(struct notifier_block *nb) · line 698

Quoted 10 lines
b/ renamed from drivers/acpi/apei/ghes.c +7 -0
@@ -689,6 +689,31 @@ void ghes_unregister_vendor_record_notifier(struct notifier_block *nb)
 }
 EXPORT_SYMBOL_GPL(ghes_unregister_vendor_record_notifier);
 
+static void ghes_vendor_record_notifier_destroy(void *nb)
+{
+	ghes_unregister_vendor_record_notifier(nb);
+}
+
+/**
+ * devm_ghes_register_vendor_record_notifier - device-managed vendor record notifier registration

There is also quite a bit of kernel doc in header. So I guess local convention is put it there not in the C code?

Hence I would move the docs there.

Quoted 35 lines
b/ renamed from drivers/acpi/apei/ghes.c +18 -0
+ * @dev: device that owns the notifier lifetime
+ * @nb: pointer to the notifier_block structure of the vendor record handler
+ *
+ * Return: 0 on success, negative errno on failure.
+ */
+int devm_ghes_register_vendor_record_notifier(struct device *ev,
+					      struct notifier_block *nb)
+{
+	int ret;
+
+	ret = ghes_register_vendor_record_notifier(nb);
+	if (ret)
+		return ret;
+
+	return devm_add_action_or_reset(dev, ghes_vendor_record_notifier_destroy, nb);
+}
+EXPORT_SYMBOL_GPL(devm_ghes_register_vendor_record_notifier);
+
 static void ghes_vendor_record_work_func(struct work_struct *work)
 {
 	struct ghes_vendor_record_entry *entry;
include/acpi/ghes.h +3 -0
index 7bea522c0657..ca3ace828c1c 100644
@@ -71,6 +71,9 @@ int ghes_register_vendor_record_notifier(struct notifier_block *nb);
  */
 void ghes_unregister_vendor_record_notifier(struct notifier_block *nb);
 
+int devm_ghes_register_vendor_record_notifier(struct device *dev,
+					      struct notifier_block *nb);
+
 struct list_head *ghes_get_devices(void);
 
 void ghes_estatus_pool_region_free(unsigned long addr, u32 size);
#15 lore raw
Quoted 11 lines · Jonathan Cameron

External email: Use caution opening links or attachments

Quoted 5 lines · On Thu, 19 Mar 2026 19:13:07 +0800 Kai-Heng Feng <[email protected]> wrote:

Add a device-managed wrapper around ghes_register_vendor_record_notifier() so drivers can avoid manual cleanup on device removal or probe failure.

Cc Jonathan Cameron <[email protected]>
Quoted 8 lines

Hi,

My only comment is about following local style. I think that means moving the docs to the header. Unfortunately whether things are in the header or the c file is a subsystem specific thing.

My preference is in the c file, but local style overrides that! Better to have all the docs in the same place.

You are right, I didn't notice it.

Quoted 30 lines

Jonathan

Quoted 13 lines

---
v2:
- New patch.

 drivers/acpi/apei/ghes.c | 25 +++++++++++++++++++++++++
 include/acpi/ghes.h      |  3 +++
 2 files changed, 28 insertions(+)
b/ renamed from drivers/acpi/apei/ghes.c +0 -0
index 8acd2742bb27..d31a70a05538 100644

b/ void ghes_unregister_vendor_record_notifier(struct notifier_block *nb) · line 698

Quoted 10 lines
b/ renamed from drivers/acpi/apei/ghes.c +7 -0
@@ -689,6 +689,31 @@ void ghes_unregister_vendor_record_notifier(struct notifier_block *nb)
 }
 EXPORT_SYMBOL_GPL(ghes_unregister_vendor_record_notifier);
 
+static void ghes_vendor_record_notifier_destroy(void *nb)
+{
+     ghes_unregister_vendor_record_notifier(nb);
+}
+
+/**
+ * devm_ghes_register_vendor_record_notifier - device-managed vendor record notifier registration

There is also quite a bit of kernel doc in header. So I guess local convention is put it there not in the C code?

Quoted 1 line

Hence I would move the docs there.

Sure, will do in next version.

Quoted 38 lines
Quoted 37 lines
b/ renamed from drivers/acpi/apei/ghes.c +18 -0
+ * @dev: device that owns the notifier lifetime
+ * @nb: pointer to the notifier_block structure of the vendor record handler
+ *
+ * Return: 0 on success, negative errno on failure.
+ */
+int devm_ghes_register_vendor_record_notifier(struct device *ev,
+                                           struct notifier_block *nb)
+{
+     int ret;
+
+     ret = ghes_register_vendor_record_notifier(nb);
+     if (ret)
+             return ret;
+
+     return devm_add_action_or_reset(dev, ghes_vendor_record_notifier_destroy, nb);
+}
+EXPORT_SYMBOL_GPL(devm_ghes_register_vendor_record_notifier);
+
 static void ghes_vendor_record_work_func(struct work_struct *work)
 {
      struct ghes_vendor_record_entry *entry;
include/acpi/ghes.h +3 -0
index 7bea522c0657..ca3ace828c1c 100644
@@ -71,6 +71,9 @@ int ghes_register_vendor_record_notifier(struct notifier_block *nb);
  */
 void ghes_unregister_vendor_record_notifier(struct notifier_block *nb);
 
+int devm_ghes_register_vendor_record_notifier(struct device *dev,
+                                           struct notifier_block *nb);
+
 struct list_head *ghes_get_devices(void);
 
 void ghes_estatus_pool_region_free(unsigned long addr, u32 size);
reviewer
#16 lore raw
In reply to Kai-Heng Feng

Hi Kai-Heng Feng,

Sorry, where is [PATCH 2/3]? did I miss something?

I reviewed the patch set, seems we don't need another patch?

Thanks
Hanjun

Quoted 64 lines · Kai-Heng Feng

Add a device-managed wrapper around ghes_register_vendor_record_notifier() so drivers can avoid manual cleanup on device removal or probe failure.

Cc Jonathan Cameron <[email protected]>

---
v2:
- New patch.

  drivers/acpi/apei/ghes.c | 25 +++++++++++++++++++++++++
  include/acpi/ghes.h      |  3 +++
  2 files changed, 28 insertions(+)
drivers/acpi/apei/ghes.c +25 -0
index 8acd2742bb27..d31a70a05538 100644
@@ -689,6 +689,31 @@ void ghes_unregister_vendor_record_notifier(struct notifier_block *nb)
  }
  EXPORT_SYMBOL_GPL(ghes_unregister_vendor_record_notifier);
 
+static void ghes_vendor_record_notifier_destroy(void *nb)
+{
+	ghes_unregister_vendor_record_notifier(nb);
+}
+
+/**
+ * devm_ghes_register_vendor_record_notifier - device-managed vendor record notifier registration
+ * @dev: device that owns the notifier lifetime
+ * @nb: pointer to the notifier_block structure of the vendor record handler
+ *
+ * Return: 0 on success, negative errno on failure.
+ */
+int devm_ghes_register_vendor_record_notifier(struct device *dev,
+					      struct notifier_block *nb)
+{
+	int ret;
+
+	ret = ghes_register_vendor_record_notifier(nb);
+	if (ret)
+		return ret;
+
+	return devm_add_action_or_reset(dev, ghes_vendor_record_notifier_destroy, nb);
+}
+EXPORT_SYMBOL_GPL(devm_ghes_register_vendor_record_notifier);
+
  static void ghes_vendor_record_work_func(struct work_struct *work)
  {
  	struct ghes_vendor_record_entry *entry;
include/acpi/ghes.h +3 -0
index 7bea522c0657..ca3ace828c1c 100644
@@ -71,6 +71,9 @@ int ghes_register_vendor_record_notifier(struct notifier_block *nb);
   */
  void ghes_unregister_vendor_record_notifier(struct notifier_block *nb);
 
+int devm_ghes_register_vendor_record_notifier(struct device *dev,
+					      struct notifier_block *nb);
+
  struct list_head *ghes_get_devices(void);
 
  void ghes_estatus_pool_region_free(unsigned long addr, u32 size);