pcie_errors: plugin to read PCIe errors

The pcie plugin collects PCI Express errors from Device Status in Capability
structure and from Advanced Error Reporting Extended Capability where available.
At every read it polls config space of PCI Express devices and dispatches
notification for every error that is found. Notif OK is sent after error is cleared.

Change-Id: I559f4035df76ab2934969a3c46cd4e98b93aba9a
Signed-off-by: Kamil Wiatrowski <kamilx.wiatrowski@intel.com>

Makefile.am

@@ -1381,6 +1381,24 @@ ovs_stats_la_LDFLAGS = $(PLUGIN_LDFLAGS) $(BUILD_WITH_LIBYAJL_LDFLAGS)
 ovs_stats_la_LIBADD = $(BUILD_WITH_LIBYAJL_LIBS)
 endif
 
+if BUILD_PLUGIN_PCIE_ERRORS
+pkglib_LTLIBRARIES += pcie_errors.la
+pcie_errors_la_SOURCES = src/pcie_errors.c
+pcie_errors_la_CPPFLAGS = $(AM_CPPFLAGS)
+pcie_errors_la_LDFLAGS = $(PLUGIN_LDFLAGS)
+
+test_plugin_pcie_errors_SOURCES = \
+	src/pcie_errors_test.c \
+	src/daemon/utils_llist.c \
+	src/daemon/configfile.c \
+	src/daemon/types_list.c
+test_plugin_pcie_errors_CPPFLAGS = $(AM_CPPFLAGS)
+test_plugin_pcie_errors_LDFLAGS = $(PLUGIN_LDFLAGS)
+test_plugin_pcie_errors_LDADD = liboconfig.la libplugin_mock.la
+check_PROGRAMS += test_plugin_pcie_errors
+TESTS += test_plugin_pcie_errors
+endif
+
 if BUILD_PLUGIN_PERL
 pkglib_LTLIBRARIES += perl.la
 perl_la_SOURCES = src/perl.c

README

@@ -314,6 +314,10 @@ Features
       OVS documentation.
       <http://openvswitch.org/support/dist-docs/INSTALL.rst.html>
 
+    - pcie_errors
+      Read errors from PCI Express Device Status and AER extended capabilities.
+      <https://www.design-reuse.com/articles/38374/pcie-error-logging-and-handling-on-a-typical-soc.html>
+
     - perl
       The perl plugin implements a Perl-interpreter into collectd. You can
       write your own plugins in Perl and return arbitrary values using this

configure.ac

@@ -550,6 +550,12 @@ if test "x$ac_system" = "xLinux"; then
     AC_DEFINE([HAVE_CAPABILITY], [1], [Define to 1 if you have cap_get_proc() (-lcap).])
   fi
 
+  # For pcie_errors plugin
+  AC_CHECK_HEADERS([linux/pci_regs.h],
+    [have_pci_regs_h="yes"],
+    [have_pci_regs_h="no (linux/pci_regs.h not found)"]
+  )
+
 else
   have_linux_raid_md_u_h="no"
   have_linux_wireless_h="no"
@@ -6229,6 +6235,7 @@ plugin_nfs="no"
 plugin_numa="no"
 plugin_ovs_events="no"
 plugin_ovs_stats="no"
+plugin_pcie_errors="no"
 plugin_perl="no"
 plugin_pinba="no"
 plugin_processes="no"
@@ -6307,6 +6314,10 @@ if test "x$ac_system" = "xLinux"; then
     plugin_ovs_events="yes"
     plugin_ovs_stats="yes"
   fi
+
+  if test "x$have_pci_regs_h" = "xyes"; then
+    plugin_pcie_errors="yes"
+  fi
 fi
 
 if test "x$ac_system" = "xOpenBSD"; then
@@ -6684,6 +6695,7 @@ AC_PLUGIN([openvpn],             [yes],                     [OpenVPN client stat
 AC_PLUGIN([oracle],              [$with_oracle],            [Oracle plugin])
 AC_PLUGIN([ovs_events],          [$plugin_ovs_events],      [OVS events plugin])
 AC_PLUGIN([ovs_stats],           [$plugin_ovs_stats],       [OVS statistics plugin])
+AC_PLUGIN([pcie_errors],         [$plugin_pcie_errors],     [PCIe errors plugin])
 AC_PLUGIN([perl],                [$plugin_perl],            [Embed a Perl interpreter])
 AC_PLUGIN([pf],                  [$have_net_pfvar_h],       [BSD packet filter (PF) statistics])
 # FIXME: Check for libevent, too.
@@ -7105,6 +7117,7 @@ AC_MSG_RESULT([    openvpn . . . . . . . $enable_openvpn])
 AC_MSG_RESULT([    oracle  . . . . . . . $enable_oracle])
 AC_MSG_RESULT([    ovs_events  . . . . . $enable_ovs_events])
 AC_MSG_RESULT([    ovs_stats . . . . . . $enable_ovs_stats])
+AC_MSG_RESULT([    pcie_errors . . . . . $enable_pcie_errors])
 AC_MSG_RESULT([    perl  . . . . . . . . $enable_perl])
 AC_MSG_RESULT([    pf  . . . . . . . . . $enable_pf])
 AC_MSG_RESULT([    pinba . . . . . . . . $enable_pinba])

src/collectd.conf.in

@@ -173,6 +173,7 @@
 #@BUILD_PLUGIN_ORACLE_TRUE@LoadPlugin oracle
 #@BUILD_PLUGIN_OVS_EVENTS_TRUE@LoadPlugin ovs_events
 #@BUILD_PLUGIN_OVS_STATS_TRUE@LoadPlugin ovs_stats
+#@BUILD_PLUGIN_PCIE_ERRORS_TRUE@LoadPlugin pcie_errors
 #@BUILD_PLUGIN_PERL_TRUE@LoadPlugin perl
 #@BUILD_PLUGIN_PINBA_TRUE@LoadPlugin pinba
 #@BUILD_PLUGIN_PING_TRUE@LoadPlugin ping
@@ -1130,6 +1131,12 @@
 #  Bridges "br0" "br_ext"
 #</Plugin>
 
+#<Plugin pcie_errors>
+#  Source "sysfs"
+#  ReportMasked false
+#  PersistentNotifications false
+#</Plugin>
+
 #<Plugin perl>
 #	IncludeDir "/my/include/path"
 #	BaseName "Collectd::Plugins"

src/collectd.conf.pod

@@ -6265,6 +6265,53 @@ Default: empty (monitor all bridges)
 
 =back
 
+=head2 Plugin C<pcie_errors>
+
+The I<pcie_errors> plugin collects PCI Express errors from Device Status in Capability
+structure and from Advanced Error Reporting Extended Capability where available.
+At every read it polls config space of PCI Express devices and dispatches
+notification for every error that is set. It checks for new errors at every read.
+The device is indicated in plugin_instance according to format "domain:bus:dev.fn".
+Errors are divided into categories indicated by type_instance: "correctable", and
+for uncorrectable errors "non_fatal" or "fatal".
+Fatal errros are reported as I<NOTIF_FAILURE> and all others as I<NOTIF_WARNING>.
+
+B<Synopsis:>
+
+  <Plugin "pcie_errors">
+    Source "sysfs"
+    AccessDir "/sys/bus/pci"
+    ReportMasked false
+    PersistentNotifications false
+  </Plugin>
+
+B<Options:>
+
+=over 4
+
+=item B<Source> B<sysfs>|B<proc>
+
+Use B<sysfs> or B<proc> to read data from /sysfs or /proc.
+The default value is B<sysfs>.
+
+=item B<AccessDir> I<dir>
+
+Directory used to access device config space. It is optional and defaults to
+/sys/bus/pci for B<sysfs> and to /proc/bus/pci for B<proc>.
+
+=item B<ReportMasked> B<false>|B<true>
+
+If true plugin will notify errors that are set to masked in Error Mask register.
+Such errors are not reported to the PCI Express Root Complex. Defaults to
+B<false>.
+
+=item B<PersistentNotifications> B<false>|B<true>
+
+If false plugin will dispatch notfication only on set/clear of error.
+The ones already reported will be ignored. Defaults to B<false>.
+
+=back
+
 =head2 Plugin C<perl>
 
 This plugin embeds a Perl-interpreter into collectd and provides an interface