[sonic_pcie] Add get_pcie_aer_stats and its common implementation (#144)

ArunSaravananBalachandran · web-flow · commit 7fc76b93b2d3 · 2021-01-14T09:18:50.000-08:00
Introduce a new platform API get_pcie_aer_stats in class PcieBase for retrieving AER stats of a PCIe device. Add common implementation of get_pcie_aer_stats in class PcieUtil Ref: sonic-net/SONiC#702
diff --git a/sonic_platform_base/sonic_pcie/pcie_base.py b/sonic_platform_base/sonic_pcie/pcie_base.py
@@ -1,41 +1,65 @@
-#
-# pcie_base.py
-#
-# Abstract base class for implementing platform-specific
-#  PCIE functionality for SONiC
-#
-
-try:
-    import abc
-except ImportError as e:
-    raise ImportError (str(e) + " - required module not found")
-
-class PcieBase(object): 
-    def __init__(self, path):
-       """
-         Constructor
-
-         Args:
-           pcieutil file and config file path
-        """
-
-    @abc.abstractmethod
-    def get_pcie_device(self):
-        """
-         get current device pcie info
-        
-          Returns:
-            A list including pcie device info
-         """
-        return [] 
-
-
-    @abc.abstractmethod
-    def get_pcie_check(self):
-        """
-         Check Pcie device with config file
-
-         Returns:
-            A list including pcie device and test result info
-        """
-        return []    
+#
+# pcie_base.py
+#
+# Abstract base class for implementing platform-specific
+#  PCIE functionality for SONiC
+#
+
+try:
+    import abc
+except ImportError as e:
+    raise ImportError(str(e) + " - required module not found")
+
+
+class PcieBase(object):
+    def __init__(self, path):
+        """
+         Constructor
+
+         Args:
+           pcieutil file and config file path
+        """
+
+    @abc.abstractmethod
+    def get_pcie_device(self):
+        """
+         get current device pcie info
+
+          Returns:
+            A list including pcie device info
+        """
+        return []
+
+    @abc.abstractmethod
+    def get_pcie_check(self):
+        """
+         Check Pcie device with config file
+
+         Returns:
+            A list including pcie device and test result info
+        """
+        return []
+
+    @abc.abstractmethod
+    def get_pcie_aer_stats(self, domain, bus, dev, fn):
+        """
+        Returns a nested dictionary containing the AER stats belonging to a
+        PCIe device
+
+        Args:
+            domain, bus, dev, fn: Domain, bus, device, function of the PCIe
+            device respectively
+
+        Returns:
+            A nested dictionary where key is severity 'correctable', 'fatal' or
+            'non_fatal', value is a dictionary of key, value pairs in the format:
+                {'AER Error type': Error count}
+
+            Ex. {'correctable': {'BadDLLP': 0, 'BadTLP': 0},
+                 'fatal': {'RxOF': 0, 'MalfTLP': 0},
+                 'non_fatal': {'RxOF': 0, 'MalfTLP': 0}}
+
+            For PCIe devices that do not support AER, the value for each
+            severity key is an empty dictionary.
+        """
+        return {}
diff --git a/sonic_platform_base/sonic_pcie/pcie_common.py b/sonic_platform_base/sonic_pcie/pcie_common.py
@@ -98,6 +98,42 @@ def get_pcie_check(self):
                 item_conf["result"] = "Failed"
         return self.confInfo
 
+    # return AER stats of PCIe device
+    def get_pcie_aer_stats(self, domain=0, bus=0, device=0, func=0):
+        aer_stats = {'correctable': {}, 'fatal': {}, 'non_fatal': {}}
+        dev_path = os.path.join('/sys/bus/pci/devices', '%04x:%02x:%02x.%d' % (domain, bus, device, func))
+
+        # construct AER sysfs filepath
+        correctable_path = os.path.join(dev_path, "aer_dev_correctable")
+        fatal_path = os.path.join(dev_path, "aer_dev_fatal")
+        non_fatal_path = os.path.join(dev_path, "aer_dev_nonfatal")
+
+        # update AER-correctable fields
+        if os.path.isfile(correctable_path):
+            with open(correctable_path, 'r') as fh:
+                lines = fh.readlines()
+            for line in lines:
+                correctable_field, value = line.split()
+                aer_stats['correctable'][correctable_field] = value
+
+        # update AER-Fatal fields
+        if os.path.isfile(fatal_path):
+            with open(fatal_path, 'r') as fh:
+                lines = fh.readlines()
+            for line in lines:
+                fatal_field, value = line.split()
+                aer_stats['fatal'][fatal_field] = value
+
+        # update AER-Non Fatal fields
+        if os.path.isfile(non_fatal_path):
+            with open(non_fatal_path, 'r') as fh:
+                lines = fh.readlines()
+            for line in lines:
+                non_fatal_field, value = line.split()
+                aer_stats['non_fatal'][non_fatal_field] = value
+
+        return aer_stats
+
     # generate the config file with current pci device
     def dump_conf_yaml(self):
         curInfo = self.get_pcie_device()
