ACPI, APEI, Add PCIe AER error information printing support
authorHuang Ying <ying.huang@intel.com>
Mon, 21 Feb 2011 05:54:43 +0000 (13:54 +0800)
committerLen Brown <len.brown@intel.com>
Tue, 22 Mar 2011 02:59:08 +0000 (22:59 -0400)
The AER error information printing support is implemented in
drivers/pci/pcie/aer/aer_print.c.  So some string constants, functions
and macros definitions can be re-used without being exported.

The original PCIe AER error information printing function is not
re-used directly because the overall format is quite different.  And
changing the original printing format may make some original users'
scripts broken.

Signed-off-by: Huang Ying <ying.huang@intel.com>
CC: Jesse Barnes <jbarnes@virtuousgeek.org>
CC: Zhang Yanmin <yanmin.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Documentation/acpi/apei/output_format.txt
drivers/acpi/apei/Kconfig
drivers/acpi/apei/cper.c
drivers/pci/pcie/aer/aerdrv.h
drivers/pci/pcie/aer/aerdrv_errprint.c
include/linux/aer.h
include/linux/cper.h

index 9146952..0c49c19 100644 (file)
@@ -92,6 +92,11 @@ vendor_id: <integer>, device_id: <integer>
 class_code: <integer>]
 [serial number: <integer>, <integer>]
 [bridge: secondary_status: <integer>, control: <integer>]
+[aer_status: <integer>, aer_mask: <integer>
+<aer status string>
+[aer_uncor_severity: <integer>]
+aer_layer=<aer layer string>, aer_agent=<aer agent string>
+aer_tlp_header: <integer> <integer> <integer> <integer>]
 
 <pcie port type string>* := PCIe end point | legacy PCI end point | \
 unknown | unknown | root port | upstream switch port | \
@@ -99,6 +104,26 @@ downstream switch port | PCIe to PCI/PCI-X bridge | \
 PCI/PCI-X to PCIe bridge | root complex integrated endpoint device | \
 root complex event collector
 
+if section severity is fatal or recoverable
+<aer status string># :=
+unknown | unknown | unknown | unknown | Data Link Protocol | \
+unknown | unknown | unknown | unknown | unknown | unknown | unknown | \
+Poisoned TLP | Flow Control Protocol | Completion Timeout | \
+Completer Abort | Unexpected Completion | Receiver Overflow | \
+Malformed TLP | ECRC | Unsupported Request
+else
+<aer status string># :=
+Receiver Error | unknown | unknown | unknown | unknown | unknown | \
+Bad TLP | Bad DLLP | RELAY_NUM Rollover | unknown | unknown | unknown | \
+Replay Timer Timeout | Advisory Non-Fatal
+fi
+
+<aer layer string> :=
+Physical Layer | Data Link Layer | Transaction Layer
+
+<aer agent string> :=
+Receiver ID | Requester ID | Completer ID | Transmitter ID
+
 Where, [] designate corresponding content is optional
 
 All <field string> description with * has the following format:
index fca34cc..9ecf6fe 100644 (file)
@@ -21,6 +21,13 @@ config ACPI_APEI_GHES
          by firmware to produce more valuable hardware error
          information for Linux.
 
+config ACPI_APEI_PCIEAER
+       bool "APEI PCIe AER logging/recovering support"
+       depends on ACPI_APEI && PCIEAER
+       help
+         PCIe AER errors may be reported via APEI firmware first mode.
+         Turn on this option to enable the corresponding support.
+
 config ACPI_APEI_EINJ
        tristate "APEI Error INJection (EINJ)"
        depends on ACPI_APEI && DEBUG_FS
index 31464a0..5d41894 100644 (file)
@@ -29,6 +29,7 @@
 #include <linux/time.h>
 #include <linux/cper.h>
 #include <linux/acpi.h>
+#include <linux/aer.h>
 
 /*
  * CPER record ID need to be unique even after reboot, because record
@@ -70,8 +71,8 @@ static const char *cper_severity_str(unsigned int severity)
  * If the output length is longer than 80, multiple line will be
  * printed, with @pfx is printed at the beginning of each line.
  */
-static void cper_print_bits(const char *pfx, unsigned int bits,
-                           const char *strs[], unsigned int strs_size)
+void cper_print_bits(const char *pfx, unsigned int bits,
+                    const char *strs[], unsigned int strs_size)
 {
        int i, len = 0;
        const char *str;
@@ -81,6 +82,8 @@ static void cper_print_bits(const char *pfx, unsigned int bits,
                if (!(bits & (1U << i)))
                        continue;
                str = strs[i];
+               if (!str)
+                       continue;
                if (len && len + strlen(str) + 2 > 80) {
                        printk("%s\n", buf);
                        len = 0;
@@ -243,7 +246,8 @@ static const char *cper_pcie_port_type_strs[] = {
        "root complex event collector",
 };
 
-static void cper_print_pcie(const char *pfx, const struct cper_sec_pcie *pcie)
+static void cper_print_pcie(const char *pfx, const struct cper_sec_pcie *pcie,
+                           const struct acpi_hest_generic_data *gdata)
 {
        if (pcie->validation_bits & CPER_PCIE_VALID_PORT_TYPE)
                printk("%s""port_type: %d, %s\n", pfx, pcie->port_type,
@@ -276,6 +280,12 @@ static void cper_print_pcie(const char *pfx, const struct cper_sec_pcie *pcie)
                printk(
        "%s""bridge: secondary_status: 0x%04x, control: 0x%04x\n",
        pfx, pcie->bridge.secondary_status, pcie->bridge.control);
+#ifdef CONFIG_ACPI_APEI_PCIEAER
+       if (pcie->validation_bits & CPER_PCIE_VALID_AER_INFO) {
+               struct aer_capability_regs *aer_regs = (void *)pcie->aer_info;
+               cper_print_aer(pfx, gdata->error_severity, aer_regs);
+       }
+#endif
 }
 
 static const char *apei_estatus_section_flag_strs[] = {
@@ -322,7 +332,7 @@ static void apei_estatus_print_section(
                struct cper_sec_pcie *pcie = (void *)(gdata + 1);
                printk("%s""section_type: PCIe error\n", pfx);
                if (gdata->error_data_length >= sizeof(*pcie))
-                       cper_print_pcie(pfx, pcie);
+                       cper_print_pcie(pfx, pcie, gdata);
                else
                        goto err_section_too_small;
        } else
index 80c11d1..3eb7708 100644 (file)
                                        PCI_ERR_UNC_UNX_COMP|           \
                                        PCI_ERR_UNC_MALF_TLP)
 
-struct header_log_regs {
-       unsigned int dw0;
-       unsigned int dw1;
-       unsigned int dw2;
-       unsigned int dw3;
-};
-
 #define AER_MAX_MULTI_ERR_DEVICES      5       /* Not likely to have more */
 struct aer_err_info {
        struct pci_dev *dev[AER_MAX_MULTI_ERR_DEVICES];
@@ -59,7 +52,7 @@ struct aer_err_info {
 
        unsigned int status;            /* COR/UNCOR Error Status */
        unsigned int mask;              /* COR/UNCOR Error Mask */
-       struct header_log_regs tlp;     /* TLP Header */
+       struct aer_header_log_regs tlp; /* TLP Header */
 };
 
 struct aer_err_source {
index 7a237f6..b07a42e 100644 (file)
@@ -19,6 +19,7 @@
 #include <linux/errno.h>
 #include <linux/pm.h>
 #include <linux/suspend.h>
+#include <linux/cper.h>
 
 #include "aerdrv.h"
 
@@ -201,3 +202,61 @@ void aer_print_port_info(struct pci_dev *dev, struct aer_err_info *info)
                info->multi_error_valid ? "Multiple " : "",
                aer_error_severity_string[info->severity], info->id);
 }
+
+#ifdef CONFIG_ACPI_APEI_PCIEAER
+static int cper_severity_to_aer(int cper_severity)
+{
+       switch (cper_severity) {
+       case CPER_SEV_RECOVERABLE:
+               return AER_NONFATAL;
+       case CPER_SEV_FATAL:
+               return AER_FATAL;
+       default:
+               return AER_CORRECTABLE;
+       }
+}
+
+void cper_print_aer(const char *prefix, int cper_severity,
+                   struct aer_capability_regs *aer)
+{
+       int aer_severity, layer, agent, status_strs_size, tlp_header_valid = 0;
+       u32 status, mask;
+       const char **status_strs;
+
+       aer_severity = cper_severity_to_aer(cper_severity);
+       if (aer_severity == AER_CORRECTABLE) {
+               status = aer->cor_status;
+               mask = aer->cor_mask;
+               status_strs = aer_correctable_error_string;
+               status_strs_size = ARRAY_SIZE(aer_correctable_error_string);
+       } else {
+               status = aer->uncor_status;
+               mask = aer->uncor_mask;
+               status_strs = aer_uncorrectable_error_string;
+               status_strs_size = ARRAY_SIZE(aer_uncorrectable_error_string);
+               tlp_header_valid = status & AER_LOG_TLP_MASKS;
+       }
+       layer = AER_GET_LAYER_ERROR(aer_severity, status);
+       agent = AER_GET_AGENT(aer_severity, status);
+       printk("%s""aer_status: 0x%08x, aer_mask: 0x%08x\n",
+              prefix, status, mask);
+       cper_print_bits(prefix, status, status_strs, status_strs_size);
+       printk("%s""aer_layer=%s, aer_agent=%s\n", prefix,
+              aer_error_layer[layer], aer_agent_string[agent]);
+       if (aer_severity != AER_CORRECTABLE)
+               printk("%s""aer_uncor_severity: 0x%08x\n",
+                      prefix, aer->uncor_severity);
+       if (tlp_header_valid) {
+               const unsigned char *tlp;
+               tlp = (const unsigned char *)&aer->header_log;
+               printk("%s""aer_tlp_header:"
+                       " %02x%02x%02x%02x %02x%02x%02x%02x"
+                       " %02x%02x%02x%02x %02x%02x%02x%02x\n",
+                       prefix, *(tlp + 3), *(tlp + 2), *(tlp + 1), *tlp,
+                       *(tlp + 7), *(tlp + 6), *(tlp + 5), *(tlp + 4),
+                       *(tlp + 11), *(tlp + 10), *(tlp + 9),
+                       *(tlp + 8), *(tlp + 15), *(tlp + 14),
+                       *(tlp + 13), *(tlp + 12));
+       }
+}
+#endif
index f7df1ee..8414de2 100644 (file)
@@ -7,6 +7,28 @@
 #ifndef _AER_H_
 #define _AER_H_
 
+struct aer_header_log_regs {
+       unsigned int dw0;
+       unsigned int dw1;
+       unsigned int dw2;
+       unsigned int dw3;
+};
+
+struct aer_capability_regs {
+       u32 header;
+       u32 uncor_status;
+       u32 uncor_mask;
+       u32 uncor_severity;
+       u32 cor_status;
+       u32 cor_mask;
+       u32 cap_control;
+       struct aer_header_log_regs header_log;
+       u32 root_command;
+       u32 root_status;
+       u16 cor_err_source;
+       u16 uncor_err_source;
+};
+
 #if defined(CONFIG_PCIEAER)
 /* pci-e port driver needs this function to enable aer */
 extern int pci_enable_pcie_error_reporting(struct pci_dev *dev);
@@ -27,5 +49,7 @@ static inline int pci_cleanup_aer_uncorrect_error_status(struct pci_dev *dev)
 }
 #endif
 
+extern void cper_print_aer(const char *prefix, int cper_severity,
+                          struct aer_capability_regs *aer);
 #endif //_AER_H_
 
index 3104aaf..372a258 100644 (file)
@@ -388,5 +388,7 @@ struct cper_sec_pcie {
 #pragma pack()
 
 u64 cper_next_record_id(void);
+void cper_print_bits(const char *prefix, unsigned int bits,
+                    const char *strs[], unsigned int strs_size);
 
 #endif