Skip to contentSkip to conversation

Make ELOG log and trace consistently with GHES

PATCH Superseded Fabio M. De Francesco 3 participants 11 messages Apr 29, 2025 – Jun 2, 2025

Subsystems (6)
ACPI PCI EDAC-core CXL +2 more subsystems

v2 of 9 Superseded Latest is v9 →

lore
Share & tools
lore
#0 lore raw

v2 cover Make ELOG log and trace consistently with GHES

When Firmware First is enabled, BIOS handles errors first and then it makes them available to the kernel via the Common Platform Error Record (CPER) sections (UEFI 2.10 Appendix N). Linux parses the CPER sections via one of two similar paths, either ELOG or GHES.

Currently, ELOG and GHES show some inconsistencies in how they print to the kernel log as well as in how they report to userspace via trace events.

Make the two mentioned paths act similarly for what relates to logging and tracing.

--- Changes for v2 ---

- Add a patch to pass log levels to pci_print_aer() (Dan)
- Add a patch to trace CPER CXL Protocol Errors
- Rework commit messages (Dan)
- Use log_non_standard_event() (Bjorn)

--- Changes for v1 ---

- Drop the RFC prefix and restart from PATCH v1
- Drop patch 3/3 because a discussion on it has not yet been settled
- Drop namespacing in export of pci_print_aer while() (Dan)
- Don't use '#ifdef' in *.c files (Dan)
- Drop a reference on pdev after operation is complete (Dan)
- Don't log an error message if pdev is NULL (Dan)

--- Changes for RFC v2 ---

- 0/3: rework the subject line and the letter.
- 1/3: no changes.
- 2/3: trace CPER PCIe Section only if CONFIG_ACPI_APEI_PCIEAER is defined; the kernel test robot reported the use of two undefined symbols because the test for the config option was missing; rewrite the subject line and part of commit message.
- 3/3: no changes.

Fabio M. De Francesco (4):
ACPI: extlog: Trace CPER Non-standard Section Body PCI/AER: Modify pci_print_aer() to take log level
ACPI: extlog: Trace CPER PCI Express Error Section
ACPI: extlog: Trace CPER CXL Protocol Errors

 drivers/acpi/acpi_extlog.c | 96 ++++++++++++++++++++++++++++++++++++++
 drivers/cxl/core/pci.c     |  2 +-
 drivers/cxl/core/ras.c     |  6 +++
 drivers/pci/pcie/aer.c     | 18 +++----
 drivers/ras/ras.c          |  1 +
 include/cxl/event.h        |  2 +
 include/linux/aer.h        | 13 +++++-
 7 files changed, 126 insertions(+), 12 deletions(-)
Signature · 1 line
2.48.1
#1 lore raw

PATCH v2 1/4 ACPI: extlog: Trace CPER Non-standard Section Body

ghes_do_proc() has a catch-all for unknown or unhandled CPER formats (UEFI v2.10 Appendix N 2.3), extlog_print() does not. This gap was noticed by a RAS test that injected CXL protocol errors which were notified to extlog_print() via the IOMCA (I/O Machine Check Architecture) mechanism. Bring parity to the extlog_print() path by including a similar log_non_standard_event().

---
 drivers/acpi/acpi_extlog.c | 6 ++++++
 drivers/ras/ras.c          | 1 +
 2 files changed, 7 insertions(+)
drivers/acpi/acpi_extlog.c +6 -0
index f7fb7205028d..caca6ccd6e99 100644
@@ -182,6 +182,12 @@ static int extlog_print(struct notifier_block *nb, unsigned long val,
 			if (gdata->error_data_length >= sizeof(*mem))
 				trace_extlog_mem_event(mem, err_seq, fru_id, fru_text,
 						       (u8)gdata->error_severity);
+		} else {
+			void *err = acpi_hest_get_payload(gdata);
+
+			log_non_standard_event(sec_type, fru_id, fru_text,
+					       gdata->error_severity, err,
+					       gdata->error_data_length);
 		}
 	}
 
drivers/ras/ras.c +1 -0
index a6e4792a1b2e..ac0e132ccc3e 100644
@@ -51,6 +51,7 @@ void log_non_standard_event(const guid_t *sec_type, const guid_t *fru_id,
 {
 	trace_non_standard_event(sec_type, fru_id, fru_text, sev, err, len);
 }
+EXPORT_SYMBOL_GPL(log_non_standard_event);
 
 void log_arm_hw_error(struct cper_sec_proc_arm *err)
 {
Signature · 1 line
2.48.1
Reviewed-by on v2 1/4
#2 lore raw

Re: [PATCH 1/4 v2] ACPI: extlog: Trace CPER Non-standard Section Body

Quoted 10 lines · On Tue, 29 Apr 2025 19:21:06 +0200 "Fabio M. De Francesco"…

ghes_do_proc() has a catch-all for unknown or unhandled CPER formats (UEFI v2.10 Appendix N 2.3), extlog_print() does not. This gap was noticed by a RAS test that injected CXL protocol errors which were notified to extlog_print() via the IOMCA (I/O Machine Check Architecture) mechanism. Bring parity to the extlog_print() path by including a similar log_non_standard_event().

Makes sense to me.

Quoted 34 lines
---
 drivers/acpi/acpi_extlog.c | 6 ++++++
 drivers/ras/ras.c          | 1 +
 2 files changed, 7 insertions(+)
drivers/acpi/acpi_extlog.c +6 -0
index f7fb7205028d..caca6ccd6e99 100644
@@ -182,6 +182,12 @@ static int extlog_print(struct notifier_block *nb, unsigned long val,
 			if (gdata->error_data_length >= sizeof(*mem))
 				trace_extlog_mem_event(mem, err_seq, fru_id, fru_text,
 						       (u8)gdata->error_severity);
+		} else {
+			void *err = acpi_hest_get_payload(gdata);
+
+			log_non_standard_event(sec_type, fru_id, fru_text,
+					       gdata->error_severity, err,
+					       gdata->error_data_length);
 		}
 	}
 
drivers/ras/ras.c +1 -0
index a6e4792a1b2e..ac0e132ccc3e 100644
@@ -51,6 +51,7 @@ void log_non_standard_event(const guid_t *sec_type, const guid_t *fru_id,
 {
 	trace_non_standard_event(sec_type, fru_id, fru_text, sev, err, len);
 }
+EXPORT_SYMBOL_GPL(log_non_standard_event);
 
 void log_arm_hw_error(struct cper_sec_proc_arm *err)
 {
#3 lore raw

PATCH v2 2/4 PCI/AER: Modify pci_print_aer() to take log level

In reply to Fabio M. De Francesco

Modify pci_print_aer() to take a printk() log level in preparation of a patch that logs PCIe Components and Link errors from ELOG.

---
 drivers/cxl/core/pci.c |  2 +-
 drivers/pci/pcie/aer.c | 16 ++++++++--------
 include/linux/aer.h    |  4 ++--
 3 files changed, 11 insertions(+), 11 deletions(-)
drivers/cxl/core/pci.c +1 -1
index 3b80e9a76ba8..ad8d7939c2e1 100644
@@ -885,7 +885,7 @@ static void cxl_handle_rdport_errors(struct cxl_dev_state *cxlds)
 	if (!cxl_rch_get_aer_severity(&aer_regs, &severity))
 		return;
 
-	pci_print_aer(pdev, severity, &aer_regs);
+	pci_print_aer(KERN_ERR, pdev, severity, &aer_regs);
 
 	if (severity == AER_CORRECTABLE)
 		cxl_handle_rdport_cor_ras(cxlds, dport);
drivers/pci/pcie/aer.c +8 -8
index a1cf8c7ef628..d0ebf7c15afa 100644
@@ -760,7 +760,7 @@ int cper_severity_to_aer(int cper_severity)
 EXPORT_SYMBOL_GPL(cper_severity_to_aer);
 #endif
 
-void pci_print_aer(struct pci_dev *dev, int aer_severity,
+void pci_print_aer(char *level, struct pci_dev *dev, int aer_severity,
 		   struct aer_capability_regs *aer)
 {
 	int layer, agent, tlp_header_valid = 0;
@@ -785,14 +785,15 @@ void pci_print_aer(struct pci_dev *dev, int aer_severity,
 	info.mask = mask;
 	info.first_error = PCI_ERR_CAP_FEP(aer->cap_control);
 
-	pci_err(dev, "aer_status: 0x%08x, aer_mask: 0x%08x\n", status, mask);
+	pci_printk(level, dev, "aer_status: 0x%08x, aer_mask: 0x%08x\n",
+		   status, mask);
 	__aer_print_error(dev, &info);
-	pci_err(dev, "aer_layer=%s, aer_agent=%s\n",
-		aer_error_layer[layer], aer_agent_string[agent]);
+	pci_printk(level, dev, "aer_layer=%s, aer_agent=%s\n",
+		   aer_error_layer[layer], aer_agent_string[agent]);
 
 	if (aer_severity != AER_CORRECTABLE)
-		pci_err(dev, "aer_uncor_severity: 0x%08x\n",
-			aer->uncor_severity);
+		pci_printk(level, dev, "aer_uncor_severity: 0x%08x\n",
+			   aer->uncor_severity);
 
 	if (tlp_header_valid)
 		pcie_print_tlp_log(dev, &aer->header_log, dev_fmt("  "));
@@ -1146,8 +1147,7 @@ static void aer_recover_work_func(struct work_struct *work)
 			       PCI_SLOT(entry.devfn), PCI_FUNC(entry.devfn));
 			continue;
 		}
-		pci_print_aer(pdev, entry.severity, entry.regs);
-
+		pci_print_aer(KERN_ERR, pdev, entry.severity, entry.regs);
 		/*
 		 * Memory for aer_capability_regs(entry.regs) is being
 		 * allocated from the ghes_estatus_pool to protect it from
include/linux/aer.h +2 -2
index 02940be66324..45d0fb2e2e75 100644
@@ -64,8 +64,8 @@ static inline int pci_aer_clear_nonfatal_status(struct pci_dev *dev)
 static inline int pcie_aer_is_native(struct pci_dev *dev) { return 0; }
 #endif
 
-void pci_print_aer(struct pci_dev *dev, int aer_severity,
-		    struct aer_capability_regs *aer);
+void pci_print_aer(char *level, struct pci_dev *dev, int aer_severity,
+		   struct aer_capability_regs *aer);
 int cper_severity_to_aer(int cper_severity);
 void aer_recover_queue(int domain, unsigned int bus, unsigned int devfn,
 		       int severity, struct aer_capability_regs *aer_regs);
Signature · 1 line
2.48.1
Reviewed-by on v2 2/4
#4 lore raw

Re: [PATCH 2/4 v2] PCI/AER: Modify pci_print_aer() to take log level

Quoted 6 lines · On Tue, 29 Apr 2025 19:21:07 +0200 "Fabio M. De Francesco"…

Modify pci_print_aer() to take a printk() log level in preparation of a patch that logs PCIe Components and Link errors from ELOG.

Fair enough on reasoning and it does what you describe so

#5 lore raw

PATCH v2 3/4 ACPI: extlog: Trace CPER PCI Express Error Section

In reply to Fabio M. De Francesco

I/O Machine Check Arcitecture events may signal failing PCIe components or links. The AER event contains details on what was happening on the wire when the error was signaled.

Trace the CPER PCIe Error section (UEFI v2.10, Appendix N.2.7) reported by the I/O MCA.

---
 drivers/acpi/acpi_extlog.c | 30 ++++++++++++++++++++++++++++++
 drivers/pci/pcie/aer.c     |  2 +-
 include/linux/aer.h        | 13 +++++++++++--
 3 files changed, 42 insertions(+), 3 deletions(-)
drivers/acpi/acpi_extlog.c +30 -0
index caca6ccd6e99..7d7a813169f1 100644
@@ -131,6 +131,32 @@ static int print_extlog_rcd(const char *pfx,
 	return 1;
 }
 
+static void extlog_print_pcie(struct cper_sec_pcie *pcie_err,
+			      int severity)
+{
+	struct aer_capability_regs *aer;
+	struct pci_dev *pdev;
+	unsigned int devfn;
+	unsigned int bus;
+	int aer_severity;
+	int domain;
+
+	if (pcie_err->validation_bits & CPER_PCIE_VALID_DEVICE_ID &&
+	    pcie_err->validation_bits & CPER_PCIE_VALID_AER_INFO) {
+		aer_severity = cper_severity_to_aer(severity);
+		aer = (struct aer_capability_regs *)pcie_err->aer_info;
+		domain = pcie_err->device_id.segment;
+		bus = pcie_err->device_id.bus;
+		devfn = PCI_DEVFN(pcie_err->device_id.device,
+				  pcie_err->device_id.function);
+		pdev = pci_get_domain_bus_and_slot(domain, bus, devfn);
+		if (!pdev)
+			return;
+		pci_print_aer(KERN_DEBUG, pdev, aer_severity, aer);
+		pci_dev_put(pdev);
+	}
+}
+
 static int extlog_print(struct notifier_block *nb, unsigned long val,
 			void *data)
 {
@@ -182,6 +208,10 @@ static int extlog_print(struct notifier_block *nb, unsigned long val,
 			if (gdata->error_data_length >= sizeof(*mem))
 				trace_extlog_mem_event(mem, err_seq, fru_id, fru_text,
 						       (u8)gdata->error_severity);
+		} else if (guid_equal(sec_type, &CPER_SEC_PCIE)) {
+			struct cper_sec_pcie *pcie_err = acpi_hest_get_payload(gdata);
+
+			extlog_print_pcie(pcie_err, gdata->error_severity);
 		} else {
 			void *err = acpi_hest_get_payload(gdata);
 
drivers/pci/pcie/aer.c +1 -1
index d0ebf7c15afa..627fcf434698 100644
@@ -801,7 +801,7 @@ void pci_print_aer(char *level, struct pci_dev *dev, int aer_severity,
 	trace_aer_event(dev_name(&dev->dev), (status & ~mask),
 			aer_severity, tlp_header_valid, &aer->header_log);
 }
-EXPORT_SYMBOL_NS_GPL(pci_print_aer, "CXL");
+EXPORT_SYMBOL_GPL(pci_print_aer);
 
 /**
  * add_error_device - list device to be handled
include/linux/aer.h +11 -2
index 45d0fb2e2e75..737db92e6570 100644
@@ -56,17 +56,26 @@ struct aer_capability_regs {
 #if defined(CONFIG_PCIEAER)
 int pci_aer_clear_nonfatal_status(struct pci_dev *dev);
 int pcie_aer_is_native(struct pci_dev *dev);
+void pci_print_aer(char *level, struct pci_dev *dev, int aer_severity,
+		   struct aer_capability_regs *aer);
 #else
 static inline int pci_aer_clear_nonfatal_status(struct pci_dev *dev)
 {
 	return -EINVAL;
 }
 static inline int pcie_aer_is_native(struct pci_dev *dev) { return 0; }
+static inline void pci_print_aer(char *level, struct pci_dev *dev,
+				 int aer_severity,
+				 struct aer_capability_regs *aer)
+{ }
 #endif
 
-void pci_print_aer(char *level, struct pci_dev *dev, int aer_severity,
-		   struct aer_capability_regs *aer);
+#if defined(CONFIG_ACPI_APEI_PCIEAER)
 int cper_severity_to_aer(int cper_severity);
+#else
+static inline int cper_severity_to_aer(int cper_severity) { return 0; }
+#endif
+
 void aer_recover_queue(int domain, unsigned int bus, unsigned int devfn,
 		       int severity, struct aer_capability_regs *aer_regs);
 #endif //_AER_H_
Signature · 1 line
2.48.1
#6 lore raw

Re: [PATCH 3/4 v2] ACPI: extlog: Trace CPER PCI Express Error Section

Quoted 25 lines · Fabio M. De Francesco

I/O Machine Check Arcitecture events may signal failing PCIe components or links. The AER event contains details on what was happening on the wire when the error was signaled.

Trace the CPER PCIe Error section (UEFI v2.10, Appendix N.2.7) reported by the I/O MCA.

---
 drivers/acpi/acpi_extlog.c | 30 ++++++++++++++++++++++++++++++
 drivers/pci/pcie/aer.c     |  2 +-
 include/linux/aer.h        | 13 +++++++++++--
 3 files changed, 42 insertions(+), 3 deletions(-)
drivers/acpi/acpi_extlog.c +2 -0
index caca6ccd6e99..7d7a813169f1 100644
@@ -131,6 +131,32 @@ static int print_extlog_rcd(const char *pfx,
 	return 1;
 }
 
+static void extlog_print_pcie(struct cper_sec_pcie *pcie_err,
+			      int severity)

drivers/acpi/acpi_extlog.c static int print_extlog_rcd(const char *pfx, · line 145

Quoted 10 lines
drivers/acpi/acpi_extlog.c +10 -0
+{
+	struct aer_capability_regs *aer;
+	struct pci_dev *pdev;
+	unsigned int devfn;
+	unsigned int bus;
+	int aer_severity;
+	int domain;
+
+	if (pcie_err->validation_bits & CPER_PCIE_VALID_DEVICE_ID &&
+	    pcie_err->validation_bits & CPER_PCIE_VALID_AER_INFO) {

You can save an indentation level by inverting this check and returning early.

drivers/acpi/acpi_extlog.c static int print_extlog_rcd(const char *pfx, · line 146

Quoted 1 line
drivers/acpi/acpi_extlog.c +1 -0
+		aer_severity = cper_severity_to_aer(severity);

I think it would help with clarity if all these lines were aligned on the "=".

drivers/acpi/acpi_extlog.c static int print_extlog_rcd(const char *pfx, · line 149

Quoted 3 lines
drivers/acpi/acpi_extlog.c +3 -0
+		aer = (struct aer_capability_regs *)pcie_err->aer_info;
+		domain = pcie_err->device_id.segment;
+		bus = pcie_err->device_id.bus;

Many of these variables are passed unchanged to a single function below.

Why not pass them directly to the function?

Even if you split the function parameters across multiple lines, you will still have fewer lines. Plus you will not need to allocate the variables.

drivers/acpi/acpi_extlog.c static int print_extlog_rcd(const char *pfx, · line 154

Quoted 5 lines
drivers/acpi/acpi_extlog.c +5 -0
+		devfn = PCI_DEVFN(pcie_err->device_id.device,
+				  pcie_err->device_id.function);
+		pdev = pci_get_domain_bus_and_slot(domain, bus, devfn);
+		if (!pdev)
+			return;

Newline here, please.

drivers/acpi/acpi_extlog.c static int print_extlog_rcd(const char *pfx, · line 155

Quoted 1 line
drivers/acpi/acpi_extlog.c +1 -0
+		pci_print_aer(KERN_DEBUG, pdev, aer_severity, aer);

Why use a debug log level?

Quoted 41 lines
drivers/acpi/acpi_extlog.c +8 -0
+		pci_dev_put(pdev);
+	}
+}
+
 static int extlog_print(struct notifier_block *nb, unsigned long val,
 			void *data)
 {
@@ -182,6 +208,10 @@ static int extlog_print(struct notifier_block *nb, unsigned long val,
 			if (gdata->error_data_length >= sizeof(*mem))
 				trace_extlog_mem_event(mem, err_seq, fru_id, fru_text,
 						       (u8)gdata->error_severity);
+		} else if (guid_equal(sec_type, &CPER_SEC_PCIE)) {
+			struct cper_sec_pcie *pcie_err = acpi_hest_get_payload(gdata);
+
+			extlog_print_pcie(pcie_err, gdata->error_severity);
 		} else {
 			void *err = acpi_hest_get_payload(gdata);
 
drivers/pci/pcie/aer.c +1 -1
index d0ebf7c15afa..627fcf434698 100644
@@ -801,7 +801,7 @@ void pci_print_aer(char *level, struct pci_dev *dev, int aer_severity,
 	trace_aer_event(dev_name(&dev->dev), (status & ~mask),
 			aer_severity, tlp_header_valid, &aer->header_log);
 }
-EXPORT_SYMBOL_NS_GPL(pci_print_aer, "CXL");
+EXPORT_SYMBOL_GPL(pci_print_aer);
 
 /**
  * add_error_device - list device to be handled
include/linux/aer.h +2 -0
index 45d0fb2e2e75..737db92e6570 100644
@@ -56,17 +56,26 @@ struct aer_capability_regs {
 #if defined(CONFIG_PCIEAER)
 int pci_aer_clear_nonfatal_status(struct pci_dev *dev);
 int pcie_aer_is_native(struct pci_dev *dev);
+void pci_print_aer(char *level, struct pci_dev *dev, int aer_severity,
+		   struct aer_capability_regs *aer);

include/linux/aer.h struct aer_capability_regs { · line 70

Quoted 10 lines
include/linux/aer.h +4 -0
 #else
 static inline int pci_aer_clear_nonfatal_status(struct pci_dev *dev)
 {
 	return -EINVAL;
 }
 static inline int pcie_aer_is_native(struct pci_dev *dev) { return 0; }
+static inline void pci_print_aer(char *level, struct pci_dev *dev,
+				 int aer_severity,
+				 struct aer_capability_regs *aer)
+{ }

I think the "{ }" can just go at the end of the parameters.

include/linux/aer.h struct aer_capability_regs { · line 76

Quoted 8 lines
include/linux/aer.h +3 -2
 #endif
 
-void pci_print_aer(char *level, struct pci_dev *dev, int aer_severity,
-		   struct aer_capability_regs *aer);
+#if defined(CONFIG_ACPI_APEI_PCIEAER)
 int cper_severity_to_aer(int cper_severity);
+#else
+static inline int cper_severity_to_aer(int cper_severity) { return 0; }

This may have an unintentional side effect.

'0' means AER_NONFATAL.

So the function will return that the error is an uncorrectable AER error that is potentially recoverable. At a minimum, this will incorrectly classify the error for data collection, and it could cause incorrect handling.

I guess the risk is minimal, since CONFIG_ACPI_APEI_PCIEAER will likely be enabled on systems that would use this.

Thanks,
Yazen

#7 lore raw

Re: [PATCH 3/4 v2] ACPI: extlog: Trace CPER PCI Express Error Section

Quoted 24 lines · Central European Summer Time Yazen Ghannam
Quoted 23 lines · Fabio M. De Francesco

I/O Machine Check Arcitecture events may signal failing PCIe components or links. The AER event contains details on what was happening on the wire when the error was signaled.

Trace the CPER PCIe Error section (UEFI v2.10, Appendix N.2.7) reported by the I/O MCA.

---
 drivers/acpi/acpi_extlog.c | 30 ++++++++++++++++++++++++++++++
 drivers/pci/pcie/aer.c     |  2 +-
 include/linux/aer.h        | 13 +++++++++++--
 3 files changed, 42 insertions(+), 3 deletions(-)
drivers/acpi/acpi_extlog.c +0 -0
index caca6ccd6e99..7d7a813169f1 100644
@@ -131,6 +131,32 @@ static int print_extlog_rcd(const char *pfx,
 	return 1;
 }
 
Quoted 16 lines

drivers/acpi/acpi_extlog.c static int print_extlog_rcd(const char *pfx, · line 145

Quoted 12 lines
drivers/acpi/acpi_extlog.c +12 -0
+static void extlog_print_pcie(struct cper_sec_pcie *pcie_err,
+			      int severity)
+{
+	struct aer_capability_regs *aer;
+	struct pci_dev *pdev;
+	unsigned int devfn;
+	unsigned int bus;
+	int aer_severity;
+	int domain;
+
+	if (pcie_err->validation_bits & CPER_PCIE_VALID_DEVICE_ID &&
+	    pcie_err->validation_bits & CPER_PCIE_VALID_AER_INFO) {

You can save an indentation level by inverting this check and returning early.

Nice idea, I'll do it.

Quoted 12 lines

drivers/acpi/acpi_extlog.c static int print_extlog_rcd(const char *pfx, · line 146

Quoted 2 lines
drivers/acpi/acpi_extlog.c +1 -0
+		aer_severity = cper_severity_to_aer(severity);

I think it would help with clarity if all these lines were aligned on the "=".

drivers/acpi/acpi_extlog.c static int print_extlog_rcd(const char *pfx, · line 149

Quoted 3 lines
drivers/acpi/acpi_extlog.c +3 -0
+		aer = (struct aer_capability_regs *)pcie_err->aer_info;
+		domain = pcie_err->device_id.segment;
+		bus = pcie_err->device_id.bus;

Many of these variables are passed unchanged to a single function below.

Quoted 6 lines

Why not pass them directly to the function?

Even if you split the function parameters across multiple lines, you will still have fewer lines. Plus you will not need to allocate the variables.

I think that the cost is minimal and readability is improved.

Quoted 9 lines

drivers/acpi/acpi_extlog.c static int print_extlog_rcd(const char *pfx, · line 154

Quoted 6 lines
drivers/acpi/acpi_extlog.c +5 -0
+		devfn = PCI_DEVFN(pcie_err->device_id.device,
+				  pcie_err->device_id.function);
+		pdev = pci_get_domain_bus_and_slot(domain, bus, devfn);
+		if (!pdev)
+			return;

Newline here, please.

Sure.

Quoted 5 lines

drivers/acpi/acpi_extlog.c static int print_extlog_rcd(const char *pfx, · line 155

Quoted 2 lines
drivers/acpi/acpi_extlog.c +1 -0
+		pci_print_aer(KERN_DEBUG, pdev, aer_severity, aer);

Why use a debug log level?

Dan Williams suggested the debug log level commenting v1.

Quoted 68 lines
Quoted 42 lines
drivers/acpi/acpi_extlog.c +8 -0
+		pci_dev_put(pdev);
+	}
+}
+
 static int extlog_print(struct notifier_block *nb, unsigned long val,
 			void *data)
 {
@@ -182,6 +208,10 @@ static int extlog_print(struct notifier_block *nb, unsigned long val,
 			if (gdata->error_data_length >= sizeof(*mem))
 				trace_extlog_mem_event(mem, err_seq, fru_id, fru_text,
 						       (u8)gdata->error_severity);
+		} else if (guid_equal(sec_type, &CPER_SEC_PCIE)) {
+			struct cper_sec_pcie *pcie_err = acpi_hest_get_payload(gdata);
+
+			extlog_print_pcie(pcie_err, gdata->error_severity);
 		} else {
 			void *err = acpi_hest_get_payload(gdata);
 
drivers/pci/pcie/aer.c +1 -1
index d0ebf7c15afa..627fcf434698 100644
@@ -801,7 +801,7 @@ void pci_print_aer(char *level, struct pci_dev *dev, int aer_severity,
 	trace_aer_event(dev_name(&dev->dev), (status & ~mask),
 			aer_severity, tlp_header_valid, &aer->header_log);
 }
-EXPORT_SYMBOL_NS_GPL(pci_print_aer, "CXL");
+EXPORT_SYMBOL_GPL(pci_print_aer);
 
 /**
  * add_error_device - list device to be handled
include/linux/aer.h +2 -0
index 45d0fb2e2e75..737db92e6570 100644
@@ -56,17 +56,26 @@ struct aer_capability_regs {
 #if defined(CONFIG_PCIEAER)
 int pci_aer_clear_nonfatal_status(struct pci_dev *dev);
 int pcie_aer_is_native(struct pci_dev *dev);
+void pci_print_aer(char *level, struct pci_dev *dev, int aer_severity,
+		   struct aer_capability_regs *aer);

include/linux/aer.h struct aer_capability_regs { · line 70

Quoted 10 lines
include/linux/aer.h +4 -0
 #else
 static inline int pci_aer_clear_nonfatal_status(struct pci_dev *dev)
 {
 	return -EINVAL;
 }
 static inline int pcie_aer_is_native(struct pci_dev *dev) { return 0; }
+static inline void pci_print_aer(char *level, struct pci_dev *dev,
+				 int aer_severity,
+				 struct aer_capability_regs *aer)
+{ }

I think the "{ }" can just go at the end of the parameters.

include/linux/aer.h struct aer_capability_regs { · line 76

Quoted 8 lines
include/linux/aer.h +3 -2
 #endif
 
-void pci_print_aer(char *level, struct pci_dev *dev, int aer_severity,
-		   struct aer_capability_regs *aer);
+#if defined(CONFIG_ACPI_APEI_PCIEAER)
 int cper_severity_to_aer(int cper_severity);
+#else
+static inline int cper_severity_to_aer(int cper_severity) { return 0; }

This may have an unintentional side effect.

'0' means AER_NONFATAL.

Quoted 8 lines

So the function will return that the error is an uncorrectable AER error that is potentially recoverable. At a minimum, this will incorrectly classify the error for data collection, and it could cause incorrect handling.

I guess the risk is minimal, since CONFIG_ACPI_APEI_PCIEAER will likely be enabled on systems that would use this.

Noted. Kconfig will select CONFIG_ACPI_APEI_PCIEAER.

Quoted 4 lines

Thanks,
Yazen

Thanks,

Fabio

#8 lore raw

PATCH v2 4/4 ACPI: extlog: Trace CPER CXL Protocol Errors

In reply to Fabio M. De Francesco

When Firmware First is enabled, BIOS handles errors first and then it makes them available to the kernel via the Common Platform Error Record (CPER) sections (UEFI 2.10 Appendix N). Linux parses the CPER sections via one of two similar paths, either ELOG or GHES.

Currently, ELOG and GHES show some inconsistencies in how they report to userspace via trace events.

Therfore make the two mentioned paths act similarly by tracing the CPER CXL Protocol Error Section (UEFI v2.10, Appendix N.2.13) signaled by the I/O Machine Check Architecture and reported by BIOS in FW-First.

---
 drivers/acpi/acpi_extlog.c | 60 ++++++++++++++++++++++++++++++++++++++
 drivers/cxl/core/ras.c     |  6 ++++
 include/cxl/event.h        |  2 ++
 3 files changed, 68 insertions(+)
drivers/acpi/acpi_extlog.c +60 -0
index 7d7a813169f1..8f2ff3505d47 100644
@@ -12,6 +12,7 @@
 #include <linux/ratelimit.h>
 #include <linux/edac.h>
 #include <linux/ras.h>
+#include <cxl/event.h>
 #include <acpi/ghes.h>
 #include <asm/cpu.h>
 #include <asm/mce.h>
@@ -157,6 +158,60 @@ static void extlog_print_pcie(struct cper_sec_pcie *pcie_err,
 	}
 }
 
+static void
+extlog_cxl_cper_handle_prot_err(struct cxl_cper_sec_prot_err *prot_err,
+				int severity)
+{
+#ifdef CONFIG_ACPI_APEI_PCIEAER
+	struct cxl_cper_prot_err_work_data wd;
+	u8 *dvsec_start, *cap_start;
+
+	if (!(prot_err->valid_bits & PROT_ERR_VALID_AGENT_ADDRESS)) {
+		pr_err_ratelimited("CXL CPER invalid agent type\n");
+		return;
+	}
+
+	if (!(prot_err->valid_bits & PROT_ERR_VALID_ERROR_LOG)) {
+		pr_err_ratelimited("CXL CPER invalid protocol error log\n");
+		return;
+	}
+
+	if (prot_err->err_len != sizeof(struct cxl_ras_capability_regs)) {
+		pr_err_ratelimited("CXL CPER invalid RAS Cap size (%u)\n",
+				   prot_err->err_len);
+		return;
+	}
+
+	if (!(prot_err->valid_bits & PROT_ERR_VALID_SERIAL_NUMBER))
+		pr_warn(FW_WARN "CXL CPER no device serial number\n");
+
+	switch (prot_err->agent_type) {
+	case RCD:
+	case DEVICE:
+	case LD:
+	case FMLD:
+	case RP:
+	case DSP:
+	case USP:
+		memcpy(&wd.prot_err, prot_err, sizeof(wd.prot_err));
+
+		dvsec_start = (u8 *)(prot_err + 1);
+		cap_start = dvsec_start + prot_err->dvsec_len;
+
+		memcpy(&wd.ras_cap, cap_start, sizeof(wd.ras_cap));
+		wd.severity = cper_severity_to_aer(severity);
+		break;
+	default:
+		pr_err_ratelimited("CXL CPER invalid agent type: %d\n",
+				   prot_err->agent_type);
+		return;
+	}
+
+	cxl_cper_ras_handle_prot_err(&wd);
+
+#endif
+}
+
 static int extlog_print(struct notifier_block *nb, unsigned long val,
 			void *data)
 {
@@ -208,6 +263,10 @@ static int extlog_print(struct notifier_block *nb, unsigned long val,
 			if (gdata->error_data_length >= sizeof(*mem))
 				trace_extlog_mem_event(mem, err_seq, fru_id, fru_text,
 						       (u8)gdata->error_severity);
+		} else if (guid_equal(sec_type, &CPER_SEC_CXL_PROT_ERR)) {
+			struct cxl_cper_sec_prot_err *prot_err = acpi_hest_get_payload(gdata);
+
+			extlog_cxl_cper_handle_prot_err(prot_err, gdata->error_severity);
 		} else if (guid_equal(sec_type, &CPER_SEC_PCIE)) {
 			struct cper_sec_pcie *pcie_err = acpi_hest_get_payload(gdata);
 
@@ -375,3 +434,4 @@ module_exit(extlog_exit);
 MODULE_AUTHOR("Chen, Gong <[email protected]>");
 MODULE_DESCRIPTION("Extended MCA Error Log Driver");
 MODULE_LICENSE("GPL");
+MODULE_IMPORT_NS("CXL");
drivers/cxl/core/ras.c +6 -0
index 485a831695c7..56db290c88d3 100644
@@ -98,6 +98,12 @@ static void cxl_cper_handle_prot_err(struct cxl_cper_prot_err_work_data *data)
 		cxl_cper_trace_uncorr_prot_err(pdev, data->ras_cap);
 }
 
+void cxl_cper_ras_handle_prot_err(struct cxl_cper_prot_err_work_data *wd)
+{
+	cxl_cper_handle_prot_err(wd);
+}
+EXPORT_SYMBOL_NS_GPL(cxl_cper_ras_handle_prot_err, "CXL");
+
 static void cxl_cper_prot_err_work_fn(struct work_struct *work)
 {
 	struct cxl_cper_prot_err_work_data wd;
include/cxl/event.h +2 -0
index f9ae1796da85..aef906e26033 100644
@@ -285,4 +285,6 @@ static inline int cxl_cper_prot_err_kfifo_get(struct cxl_cper_prot_err_work_data
 }
 #endif
 
+void cxl_cper_ras_handle_prot_err(struct cxl_cper_prot_err_work_data *wd);
+
 #endif /* _LINUX_CXL_EVENT_H */
Signature · 1 line
2.48.1
#9 lore raw

Re: [PATCH 4/4 v2] ACPI: extlog: Trace CPER CXL Protocol Errors

Quoted 33 lines · Fabio M. De Francesco

When Firmware First is enabled, BIOS handles errors first and then it makes them available to the kernel via the Common Platform Error Record (CPER) sections (UEFI 2.10 Appendix N). Linux parses the CPER sections via one of two similar paths, either ELOG or GHES.

Currently, ELOG and GHES show some inconsistencies in how they report to userspace via trace events.

Therfore make the two mentioned paths act similarly by tracing the CPER CXL Protocol Error Section (UEFI v2.10, Appendix N.2.13) signaled by the I/O Machine Check Architecture and reported by BIOS in FW-First.

---
 drivers/acpi/acpi_extlog.c | 60 ++++++++++++++++++++++++++++++++++++++
 drivers/cxl/core/ras.c     |  6 ++++
 include/cxl/event.h        |  2 ++
 3 files changed, 68 insertions(+)
drivers/acpi/acpi_extlog.c +1 -0
index 7d7a813169f1..8f2ff3505d47 100644
@@ -12,6 +12,7 @@
 #include <linux/ratelimit.h>
 #include <linux/edac.h>
 #include <linux/ras.h>
+#include <cxl/event.h>
 #include <acpi/ghes.h>
 #include <asm/cpu.h>
 #include <asm/mce.h>

drivers/acpi/acpi_extlog.c static void extlog_print_pcie(struct cper_sec_pcie *pcie_err, · line 165

Quoted 8 lines
drivers/acpi/acpi_extlog.c +5 -0
@@ -157,6 +158,60 @@ static void extlog_print_pcie(struct cper_sec_pcie *pcie_err,
 	}
 }
 
+static void
+extlog_cxl_cper_handle_prot_err(struct cxl_cper_sec_prot_err *prot_err,
+				int severity)
+{
+#ifdef CONFIG_ACPI_APEI_PCIEAER

Why not apply this check on the function prototype?

Reference: Documentation/process/coding-style.rst
Section 21) Conditional Compilation

Quoted 11 lines
drivers/acpi/acpi_extlog.c +11 -0
+	struct cxl_cper_prot_err_work_data wd;
+	u8 *dvsec_start, *cap_start;
+
+	if (!(prot_err->valid_bits & PROT_ERR_VALID_AGENT_ADDRESS)) {
+		pr_err_ratelimited("CXL CPER invalid agent type\n");
+		return;
+	}
+
+	if (!(prot_err->valid_bits & PROT_ERR_VALID_ERROR_LOG)) {
+		pr_err_ratelimited("CXL CPER invalid protocol error log\n");
+		return;

drivers/acpi/acpi_extlog.c static void extlog_print_pcie(struct cper_sec_pcie *pcie_err, · line 186

Quoted 10 lines
drivers/acpi/acpi_extlog.c +10 -0
+	}
+
+	if (prot_err->err_len != sizeof(struct cxl_ras_capability_regs)) {
+		pr_err_ratelimited("CXL CPER invalid RAS Cap size (%u)\n",
+				   prot_err->err_len);
+		return;
+	}
+
+	if (!(prot_err->valid_bits & PROT_ERR_VALID_SERIAL_NUMBER))
+		pr_warn(FW_WARN "CXL CPER no device serial number\n");

Is this a requirement (in the spec) that we should warn users about?

The UEFI spec says that serial number is only used if "CXL agent" is a "CXL device".

"CXL ports" won't have serial numbers. So this will be a false warning for port errors.

Quoted 9 lines
drivers/acpi/acpi_extlog.c +9 -0
+
+	switch (prot_err->agent_type) {
+	case RCD:
+	case DEVICE:
+	case LD:
+	case FMLD:
+	case RP:
+	case DSP:
+	case USP:

drivers/acpi/acpi_extlog.c static void extlog_print_pcie(struct cper_sec_pcie *pcie_err, · line 205

Quoted 10 lines
drivers/acpi/acpi_extlog.c +10 -0
+		memcpy(&wd.prot_err, prot_err, sizeof(wd.prot_err));
+
+		dvsec_start = (u8 *)(prot_err + 1);
+		cap_start = dvsec_start + prot_err->dvsec_len;
+
+		memcpy(&wd.ras_cap, cap_start, sizeof(wd.ras_cap));
+		wd.severity = cper_severity_to_aer(severity);
+		break;
+	default:
+		pr_err_ratelimited("CXL CPER invalid agent type: %d\n",

"invalid" is too harsh given that the specs may be updated. Maybe say "reserved" or "unknown" or "unrecognized" instead.

Hopefully things will settle down to where a user will be able to have a system with newer CXL "agents" without *requiring* a kernel update. :)

Thanks,
Yazen

#10 lore raw

Re: [PATCH 4/4 v2] ACPI: extlog: Trace CPER CXL Protocol Errors

Quoted 37 lines · Central European Summer Time Yazen Ghannam
Quoted 36 lines · Fabio M. De Francesco

When Firmware First is enabled, BIOS handles errors first and then it makes them available to the kernel via the Common Platform Error Record (CPER) sections (UEFI 2.10 Appendix N). Linux parses the CPER sections via one of two similar paths, either ELOG or GHES.

Currently, ELOG and GHES show some inconsistencies in how they report to userspace via trace events.

Therfore make the two mentioned paths act similarly by tracing the CPER CXL Protocol Error Section (UEFI v2.10, Appendix N.2.13) signaled by the I/O Machine Check Architecture and reported by BIOS in FW-First.

---
 drivers/acpi/acpi_extlog.c | 60 ++++++++++++++++++++++++++++++++++++++
 drivers/cxl/core/ras.c     |  6 ++++
 include/cxl/event.h        |  2 ++
 3 files changed, 68 insertions(+)
drivers/acpi/acpi_extlog.c +1 -0
index 7d7a813169f1..8f2ff3505d47 100644
@@ -12,6 +12,7 @@
 #include <linux/ratelimit.h>
 #include <linux/edac.h>
 #include <linux/ras.h>
+#include <cxl/event.h>
 #include <acpi/ghes.h>
 #include <asm/cpu.h>
 #include <asm/mce.h>
@@ -157,6 +158,60 @@ static void extlog_print_pcie(struct cper_sec_pcie *pcie_err,
 	}
 }
 
Quoted 8 lines

drivers/acpi/acpi_extlog.c static void extlog_print_pcie(struct cper_sec_pcie *pcie_err, · line 165

Quoted 5 lines
drivers/acpi/acpi_extlog.c +5 -0
+static void
+extlog_cxl_cper_handle_prot_err(struct cxl_cper_sec_prot_err *prot_err,
+				int severity)
+{
+#ifdef CONFIG_ACPI_APEI_PCIEAER

Why not apply this check on the function prototype?

This function is static.

Quoted 28 lines

Reference: Documentation/process/coding-style.rst
Section 21) Conditional Compilation

Quoted 11 lines
drivers/acpi/acpi_extlog.c +11 -0
+	struct cxl_cper_prot_err_work_data wd;
+	u8 *dvsec_start, *cap_start;
+
+	if (!(prot_err->valid_bits & PROT_ERR_VALID_AGENT_ADDRESS)) {
+		pr_err_ratelimited("CXL CPER invalid agent type\n");
+		return;
+	}
+
+	if (!(prot_err->valid_bits & PROT_ERR_VALID_ERROR_LOG)) {
+		pr_err_ratelimited("CXL CPER invalid protocol error log\n");
+		return;

drivers/acpi/acpi_extlog.c static void extlog_print_pcie(struct cper_sec_pcie *pcie_err, · line 186

Quoted 10 lines
drivers/acpi/acpi_extlog.c +10 -0
+	}
+
+	if (prot_err->err_len != sizeof(struct cxl_ras_capability_regs)) {
+		pr_err_ratelimited("CXL CPER invalid RAS Cap size (%u)\n",
+				   prot_err->err_len);
+		return;
+	}
+
+	if (!(prot_err->valid_bits & PROT_ERR_VALID_SERIAL_NUMBER))
+		pr_warn(FW_WARN "CXL CPER no device serial number\n");

Is this a requirement (in the spec) that we should warn users about?

Quoted 6 lines

The UEFI spec says that serial number is only used if "CXL agent" is a "CXL device".

"CXL ports" won't have serial numbers. So this will be a false warning for port errors.

I'll add a test and print that warning only if agent is a device (RCD, DEVICE, LD, FMLD).

Quoted 24 lines
Quoted 10 lines
drivers/acpi/acpi_extlog.c +9 -0
+
+	switch (prot_err->agent_type) {
+	case RCD:
+	case DEVICE:
+	case LD:
+	case FMLD:
+	case RP:
+	case DSP:
+	case USP:

drivers/acpi/acpi_extlog.c static void extlog_print_pcie(struct cper_sec_pcie *pcie_err, · line 205

Quoted 10 lines
drivers/acpi/acpi_extlog.c +10 -0
+		memcpy(&wd.prot_err, prot_err, sizeof(wd.prot_err));
+
+		dvsec_start = (u8 *)(prot_err + 1);
+		cap_start = dvsec_start + prot_err->dvsec_len;
+
+		memcpy(&wd.ras_cap, cap_start, sizeof(wd.ras_cap));
+		wd.severity = cper_severity_to_aer(severity);
+		break;
+	default:
+		pr_err_ratelimited("CXL CPER invalid agent type: %d\n",

"invalid" is too harsh given that the specs may be updated. Maybe say "reserved" or "unknown" or "unrecognized" instead.

Quoted 3 lines

Hopefully things will settle down to where a user will be able to have a system with newer CXL "agents" without *requiring* a kernel update. :)

I'll replace "invalid" with "unknown".

Quoted 4 lines

Thanks,
Yazen

Thanks,

Fabio