diff --git a/.rasdaemon.metadata b/.rasdaemon.metadata index fea2747..a3cbaea 100644 --- a/.rasdaemon.metadata +++ b/.rasdaemon.metadata @@ -1 +1 @@ -bf07a1cbbbd3cc0b2ff60eff9ee3cba3856e9e6c SOURCES/rasdaemon-0.6.6.tar.bz2 +667c81ed421c236c910e8e6a51ca1fff8775f5a8 SOURCES/rasdaemon-0.6.6.tar.bz2 diff --git a/SOURCES/129f20ba5e0b8b5881539e0038e97ef353d5d294.patch b/SOURCES/129f20ba5e0b8b5881539e0038e97ef353d5d294.patch new file mode 100644 index 0000000..886dedd --- /dev/null +++ b/SOURCES/129f20ba5e0b8b5881539e0038e97ef353d5d294.patch @@ -0,0 +1,26 @@ +commit 129f20ba5e0b8b5881539e0038e97ef353d5d294 +Author: Cong Wang +Date: Fri Feb 28 12:37:15 2020 -0800 + + Match rankX in ras-mc-ctl + + According to kernel doc: + https://www.kernel.org/doc/html/v4.10/admin-guide/ras.html + mcX directory contains either dimmX or rankX directories. + + Signed-off-by: Cong Wang + Signed-off-by: Mauro Carvalho Chehab + +diff --git a/util/ras-mc-ctl.in b/util/ras-mc-ctl.in +index b4e5856..5a4aaf6 100755 +--- a/util/ras-mc-ctl.in ++++ b/util/ras-mc-ctl.in +@@ -247,7 +247,7 @@ sub parse_dimm_nodes + $mc =~ s,.*mc(\d+).*,$1,; + + my $dimm = $file; +- $dimm =~ s,.*dimm(\d+).*,$1,; ++ $dimm =~ s,.*(rank|dimm)(\d+).*,$2,; + + open IN, $file; + my $location = ; diff --git a/SOURCES/3f1fa60f2fef32c05bed1f21617311216ee80026.patch b/SOURCES/3f1fa60f2fef32c05bed1f21617311216ee80026.patch new file mode 100644 index 0000000..1d804db --- /dev/null +++ b/SOURCES/3f1fa60f2fef32c05bed1f21617311216ee80026.patch @@ -0,0 +1,525 @@ +commit 3f1fa60f2fef32c05bed1f21617311216ee80026 +Author: Subhendu Saha +Date: Tue Jan 12 03:29:55 2021 -0500 + + Fix ras-mc-ctl script. + + When rasdaemon is compiled without enabling aer, mce, devlink, + etc., those tables are not created in the database file. Then + ras-mc-ctl script breaks trying to query data from non-existent + tables. + + Signed-off-by: Subhendu Saha subhends@akamai.com + Signed-off-by: Mauro Carvalho Chehab + +diff --git a/util/ras-mc-ctl.in b/util/ras-mc-ctl.in +index dd7d56f..1fbeb63 100755 +--- a/util/ras-mc-ctl.in ++++ b/util/ras-mc-ctl.in +@@ -41,6 +41,20 @@ my $sysconfdir = "@sysconfdir@"; + my $dmidecode = find_prog ("dmidecode"); + my $modprobe = find_prog ("modprobe") or exit (1); + ++my $has_aer = 0; ++my $has_arm = 0; ++my $has_devlink = 0; ++my $has_disk_errors = 0; ++my $has_extlog = 0; ++my $has_mce = 0; ++ ++@WITH_AER_TRUE@$has_aer = 1; ++@WITH_ARM_TRUE@$has_arm = 1; ++@WITH_DEVLINK_TRUE@$has_devlink = 1; ++@WITH_DISKERROR_TRUE@$has_disk_errors = 1; ++@WITH_EXTLOG_TRUE@$has_extlog = 1; ++@WITH_MCE_TRUE@$has_mce = 1; ++ + my %conf = (); + my %bus = (); + my %dimm_size = (); +@@ -1144,102 +1158,114 @@ sub summary + $query_handle->finish; + + # PCIe AER aer_event errors +- $query = "select err_type, err_msg, count(*) from aer_event group by err_type, err_msg"; +- $query_handle = $dbh->prepare($query); +- $query_handle->execute(); +- $query_handle->bind_columns(\($err_type, $msg, $count)); +- $out = ""; +- while($query_handle->fetch()) { +- $out .= "\t$count $err_type errors: $msg\n"; +- } +- if ($out ne "") { +- print "PCIe AER events summary:\n$out\n"; +- } else { +- print "No PCIe AER errors.\n\n"; ++ if ($has_aer == 1) { ++ $query = "select err_type, err_msg, count(*) from aer_event group by err_type, err_msg"; ++ $query_handle = $dbh->prepare($query); ++ $query_handle->execute(); ++ $query_handle->bind_columns(\($err_type, $msg, $count)); ++ $out = ""; ++ while($query_handle->fetch()) { ++ $out .= "\t$count $err_type errors: $msg\n"; ++ } ++ if ($out ne "") { ++ print "PCIe AER events summary:\n$out\n"; ++ } else { ++ print "No PCIe AER errors.\n\n"; ++ } ++ $query_handle->finish; + } +- $query_handle->finish; + + # ARM processor arm_event errors +- $query = "select affinity, mpidr, count(*) from arm_event group by affinity, mpidr"; +- $query_handle = $dbh->prepare($query); +- $query_handle->execute(); +- $query_handle->bind_columns(\($affinity, $mpidr, $count)); +- $out = ""; +- while($query_handle->fetch()) { +- $out .= "\t$count errors\n"; +- } +- if ($out ne "") { +- print "ARM processor events summary:\n$out\n"; +- } else { +- print "No ARM processor errors.\n\n"; ++ if ($has_arm == 1) { ++ $query = "select affinity, mpidr, count(*) from arm_event group by affinity, mpidr"; ++ $query_handle = $dbh->prepare($query); ++ $query_handle->execute(); ++ $query_handle->bind_columns(\($affinity, $mpidr, $count)); ++ $out = ""; ++ while($query_handle->fetch()) { ++ $out .= "\t$count errors\n"; ++ } ++ if ($out ne "") { ++ print "ARM processor events summary:\n$out\n"; ++ } else { ++ print "No ARM processor errors.\n\n"; ++ } ++ $query_handle->finish; + } +- $query_handle->finish; + + # extlog errors +- $query = "select etype, severity, count(*) from extlog_event group by etype, severity"; +- $query_handle = $dbh->prepare($query); +- $query_handle->execute(); +- $query_handle->bind_columns(\($etype, $severity, $count)); +- $out = ""; +- while($query_handle->fetch()) { +- $etype_string = get_extlog_type($etype); +- $severity_string = get_extlog_severity($severity); +- $out .= "\t$count $etype_string $severity_string errors\n"; +- } +- if ($out ne "") { +- print "Extlog records summary:\n$out"; +- } else { +- print "No Extlog errors.\n\n"; ++ if ($has_extlog == 1) { ++ $query = "select etype, severity, count(*) from extlog_event group by etype, severity"; ++ $query_handle = $dbh->prepare($query); ++ $query_handle->execute(); ++ $query_handle->bind_columns(\($etype, $severity, $count)); ++ $out = ""; ++ while($query_handle->fetch()) { ++ $etype_string = get_extlog_type($etype); ++ $severity_string = get_extlog_severity($severity); ++ $out .= "\t$count $etype_string $severity_string errors\n"; ++ } ++ if ($out ne "") { ++ print "Extlog records summary:\n$out"; ++ } else { ++ print "No Extlog errors.\n\n"; ++ } ++ $query_handle->finish; + } +- $query_handle->finish; + + # devlink errors +- $query = "select dev_name, count(*) from devlink_event group by dev_name"; +- $query_handle = $dbh->prepare($query); +- $query_handle->execute(); +- $query_handle->bind_columns(\($dev_name, $count)); +- $out = ""; +- while($query_handle->fetch()) { +- $out .= "\t$dev_name has $count errors\n"; +- } +- if ($out ne "") { +- print "Devlink records summary:\n$out"; +- } else { +- print "No devlink errors.\n"; ++ if ($has_devlink == 1) { ++ $query = "select dev_name, count(*) from devlink_event group by dev_name"; ++ $query_handle = $dbh->prepare($query); ++ $query_handle->execute(); ++ $query_handle->bind_columns(\($dev_name, $count)); ++ $out = ""; ++ while($query_handle->fetch()) { ++ $out .= "\t$dev_name has $count errors\n"; ++ } ++ if ($out ne "") { ++ print "Devlink records summary:\n$out"; ++ } else { ++ print "No devlink errors.\n"; ++ } ++ $query_handle->finish; + } +- $query_handle->finish; + + # Disk errors +- $query = "select dev, count(*) from disk_errors group by dev"; +- $query_handle = $dbh->prepare($query); +- $query_handle->execute(); +- $query_handle->bind_columns(\($dev, $count)); +- $out = ""; +- while($query_handle->fetch()) { +- $out .= "\t$dev has $count errors\n"; +- } +- if ($out ne "") { +- print "Disk errors summary:\n$out"; +- } else { +- print "No disk errors.\n"; ++ if ($has_disk_errors == 1) { ++ $query = "select dev, count(*) from disk_errors group by dev"; ++ $query_handle = $dbh->prepare($query); ++ $query_handle->execute(); ++ $query_handle->bind_columns(\($dev, $count)); ++ $out = ""; ++ while($query_handle->fetch()) { ++ $out .= "\t$dev has $count errors\n"; ++ } ++ if ($out ne "") { ++ print "Disk errors summary:\n$out"; ++ } else { ++ print "No disk errors.\n"; ++ } ++ $query_handle->finish; + } +- $query_handle->finish; + + # MCE mce_record errors +- $query = "select error_msg, count(*) from mce_record group by error_msg"; +- $query_handle = $dbh->prepare($query); +- $query_handle->execute(); +- $query_handle->bind_columns(\($msg, $count)); +- $out = ""; +- while($query_handle->fetch()) { +- $out .= "\t$count $msg errors\n"; +- } +- if ($out ne "") { +- print "MCE records summary:\n$out"; +- } else { +- print "No MCE errors.\n"; ++ if ($has_mce == 1) { ++ $query = "select error_msg, count(*) from mce_record group by error_msg"; ++ $query_handle = $dbh->prepare($query); ++ $query_handle->execute(); ++ $query_handle->bind_columns(\($msg, $count)); ++ $out = ""; ++ while($query_handle->fetch()) { ++ $out .= "\t$count $msg errors\n"; ++ } ++ if ($out ne "") { ++ print "MCE records summary:\n$out"; ++ } else { ++ print "No MCE errors.\n"; ++ } ++ $query_handle->finish; + } +- $query_handle->finish; + + undef($dbh); + } +@@ -1277,150 +1303,162 @@ sub errors + $query_handle->finish; + + # PCIe AER aer_event errors +- $query = "select id, timestamp, dev_name, err_type, err_msg from aer_event order by id"; +- $query_handle = $dbh->prepare($query); +- $query_handle->execute(); +- $query_handle->bind_columns(\($id, $time, $devname, $type, $msg)); +- $out = ""; +- while($query_handle->fetch()) { +- $out .= "$id $time $devname $type error: $msg\n"; +- } +- if ($out ne "") { +- print "PCIe AER events:\n$out\n"; +- } else { +- print "No PCIe AER errors.\n\n"; ++ if ($has_aer == 1) { ++ $query = "select id, timestamp, dev_name, err_type, err_msg from aer_event order by id"; ++ $query_handle = $dbh->prepare($query); ++ $query_handle->execute(); ++ $query_handle->bind_columns(\($id, $time, $devname, $type, $msg)); ++ $out = ""; ++ while($query_handle->fetch()) { ++ $out .= "$id $time $devname $type error: $msg\n"; ++ } ++ if ($out ne "") { ++ print "PCIe AER events:\n$out\n"; ++ } else { ++ print "No PCIe AER errors.\n\n"; ++ } ++ $query_handle->finish; + } +- $query_handle->finish; + + # ARM processor arm_event errors +- $query = "select id, timestamp, error_count, affinity, mpidr, running_state, psci_state from arm_event order by id"; +- $query_handle = $dbh->prepare($query); +- $query_handle->execute(); +- $query_handle->bind_columns(\($id, $timestamp, $error_count, $affinity, $mpidr, $r_state, $psci_state)); +- $out = ""; +- while($query_handle->fetch()) { +- $out .= "$id $timestamp error: "; +- $out .= "error_count=$error_count, " if ($error_count); +- $out .= "affinity_level=$affinity, "; +- $out .= sprintf "mpidr=0x%x, ", $mpidr; +- $out .= sprintf "running_state=0x%x, ", $r_state; +- $out .= sprintf "psci_state=0x%x", $psci_state; +- $out .= "\n"; +- } +- if ($out ne "") { +- print "ARM processor events:\n$out\n"; +- } else { +- print "No ARM processor errors.\n\n"; ++ if ($has_arm == 1) { ++ $query = "select id, timestamp, error_count, affinity, mpidr, running_state, psci_state from arm_event order by id"; ++ $query_handle = $dbh->prepare($query); ++ $query_handle->execute(); ++ $query_handle->bind_columns(\($id, $timestamp, $error_count, $affinity, $mpidr, $r_state, $psci_state)); ++ $out = ""; ++ while($query_handle->fetch()) { ++ $out .= "$id $timestamp error: "; ++ $out .= "error_count=$error_count, " if ($error_count); ++ $out .= "affinity_level=$affinity, "; ++ $out .= sprintf "mpidr=0x%x, ", $mpidr; ++ $out .= sprintf "running_state=0x%x, ", $r_state; ++ $out .= sprintf "psci_state=0x%x", $psci_state; ++ $out .= "\n"; ++ } ++ if ($out ne "") { ++ print "ARM processor events:\n$out\n"; ++ } else { ++ print "No ARM processor errors.\n\n"; ++ } ++ $query_handle->finish; + } +- $query_handle->finish; + + # Extlog errors +- $query = "select id, timestamp, etype, severity, address, fru_id, fru_text, cper_data from extlog_event order by id"; +- $query_handle = $dbh->prepare($query); +- $query_handle->execute(); +- $query_handle->bind_columns(\($id, $timestamp, $etype, $severity, $addr, $fru_id, $fru_text, $cper_data)); +- $out = ""; +- while($query_handle->fetch()) { +- $etype_string = get_extlog_type($etype); +- $severity_string = get_extlog_severity($severity); +- $out .= "$id $timestamp error: "; +- $out .= "type=$etype_string, "; +- $out .= "severity=$severity_string, "; +- $out .= sprintf "address=0x%08x, ", $addr; +- $out .= sprintf "fru_id=%s, ", get_uuid_le($fru_id); +- $out .= "fru_text='$fru_text', "; +- $out .= get_cper_data_text($cper_data) if ($cper_data); +- $out .= "\n"; +- } +- if ($out ne "") { +- print "Extlog events:\n$out\n"; +- } else { +- print "No Extlog errors.\n\n"; ++ if ($has_extlog == 1) { ++ $query = "select id, timestamp, etype, severity, address, fru_id, fru_text, cper_data from extlog_event order by id"; ++ $query_handle = $dbh->prepare($query); ++ $query_handle->execute(); ++ $query_handle->bind_columns(\($id, $timestamp, $etype, $severity, $addr, $fru_id, $fru_text, $cper_data)); ++ $out = ""; ++ while($query_handle->fetch()) { ++ $etype_string = get_extlog_type($etype); ++ $severity_string = get_extlog_severity($severity); ++ $out .= "$id $timestamp error: "; ++ $out .= "type=$etype_string, "; ++ $out .= "severity=$severity_string, "; ++ $out .= sprintf "address=0x%08x, ", $addr; ++ $out .= sprintf "fru_id=%s, ", get_uuid_le($fru_id); ++ $out .= "fru_text='$fru_text', "; ++ $out .= get_cper_data_text($cper_data) if ($cper_data); ++ $out .= "\n"; ++ } ++ if ($out ne "") { ++ print "Extlog events:\n$out\n"; ++ } else { ++ print "No Extlog errors.\n\n"; ++ } ++ $query_handle->finish; + } +- $query_handle->finish; + + # devlink errors +- $query = "select id, timestamp, bus_name, dev_name, driver_name, reporter_name, msg from devlink_event order by id"; +- $query_handle = $dbh->prepare($query); +- $query_handle->execute(); +- $query_handle->bind_columns(\($id, $timestamp, $bus_name, $dev_name, $driver_name, $reporter_name, $msg)); +- $out = ""; +- while($query_handle->fetch()) { +- $out .= "$id $timestamp error: "; +- $out .= "bus_name=$bus_name, "; +- $out .= "dev_name=$dev_name, "; +- $out .= "driver_name=$driver_name, "; +- $out .= "reporter_name=$reporter_name, "; +- $out .= "message='$msg', "; +- $out .= "\n"; +- } +- if ($out ne "") { +- print "Devlink events:\n$out\n"; +- } else { +- print "No devlink errors.\n\n"; ++ if ($has_devlink == 1) { ++ $query = "select id, timestamp, bus_name, dev_name, driver_name, reporter_name, msg from devlink_event order by id"; ++ $query_handle = $dbh->prepare($query); ++ $query_handle->execute(); ++ $query_handle->bind_columns(\($id, $timestamp, $bus_name, $dev_name, $driver_name, $reporter_name, $msg)); ++ $out = ""; ++ while($query_handle->fetch()) { ++ $out .= "$id $timestamp error: "; ++ $out .= "bus_name=$bus_name, "; ++ $out .= "dev_name=$dev_name, "; ++ $out .= "driver_name=$driver_name, "; ++ $out .= "reporter_name=$reporter_name, "; ++ $out .= "message='$msg', "; ++ $out .= "\n"; ++ } ++ if ($out ne "") { ++ print "Devlink events:\n$out\n"; ++ } else { ++ print "No devlink errors.\n\n"; ++ } ++ $query_handle->finish; + } +- $query_handle->finish; + + # Disk errors +- $query = "select id, timestamp, dev, sector, nr_sector, error, rwbs, cmd from disk_errors order by id"; +- $query_handle = $dbh->prepare($query); +- $query_handle->execute(); +- $query_handle->bind_columns(\($id, $timestamp, $dev, $sector, $nr_sector, $error, $rwbs, $cmd)); +- $out = ""; +- while($query_handle->fetch()) { +- $out .= "$id $timestamp error: "; +- $out .= "dev=$dev, "; +- $out .= "sector=$sector, "; +- $out .= "nr_sector=$nr_sector, "; +- $out .= "error='$error', "; +- $out .= "rwbs='$rwbs', "; +- $out .= "cmd='$cmd', "; +- $out .= "\n"; +- } +- if ($out ne "") { +- print "Disk errors\n$out\n"; +- } else { +- print "No disk errors.\n\n"; ++ if ($has_disk_errors == 1) { ++ $query = "select id, timestamp, dev, sector, nr_sector, error, rwbs, cmd from disk_errors order by id"; ++ $query_handle = $dbh->prepare($query); ++ $query_handle->execute(); ++ $query_handle->bind_columns(\($id, $timestamp, $dev, $sector, $nr_sector, $error, $rwbs, $cmd)); ++ $out = ""; ++ while($query_handle->fetch()) { ++ $out .= "$id $timestamp error: "; ++ $out .= "dev=$dev, "; ++ $out .= "sector=$sector, "; ++ $out .= "nr_sector=$nr_sector, "; ++ $out .= "error='$error', "; ++ $out .= "rwbs='$rwbs', "; ++ $out .= "cmd='$cmd', "; ++ $out .= "\n"; ++ } ++ if ($out ne "") { ++ print "Disk errors\n$out\n"; ++ } else { ++ print "No disk errors.\n\n"; ++ } ++ $query_handle->finish; + } +- $query_handle->finish; + + # MCE mce_record errors +- $query = "select id, timestamp, mcgcap, mcgstatus, status, addr, misc, ip, tsc, walltime, cpu, cpuid, apicid, socketid, cs, bank, cpuvendor, bank_name, error_msg, mcgstatus_msg, mcistatus_msg, user_action, mc_location from mce_record order by id"; +- $query_handle = $dbh->prepare($query); +- $query_handle->execute(); +- $query_handle->bind_columns(\($id, $time, $mcgcap,$mcgstatus, $status, $addr, $misc, $ip, $tsc, $walltime, $cpu, $cpuid, $apicid, $socketid, $cs, $bank, $cpuvendor, $bank_name, $msg, $mcgstatus_msg, $mcistatus_msg, $user_action, $mc_location)); +- $out = ""; +- while($query_handle->fetch()) { +- $out .= "$id $time error: $msg"; +- $out .= ", CPU $cpuvendor" if ($cpuvendor); +- $out .= ", bank $bank_name" if ($bank_name); +- $out .= ", mcg $mcgstatus_msg" if ($mcgstatus_msg); +- $out .= ", mci $mcistatus_msg" if ($mcistatus_msg); +- $out .= ", $mc_location" if ($mc_location); +- $out .= ", $user_action" if ($user_action); +- $out .= sprintf ", mcgcap=0x%08x", $mcgcap if ($mcgcap); +- $out .= sprintf ", mcgstatus=0x%08x", $mcgstatus if ($mcgstatus); +- $out .= sprintf ", status=0x%08x", $status if ($status); +- $out .= sprintf ", addr=0x%08x", $addr if ($addr); +- $out .= sprintf ", misc=0x%08x", $misc if ($misc); +- $out .= sprintf ", ip=0x%08x", $ip if ($ip); +- $out .= sprintf ", tsc=0x%08x", $tsc if ($tsc); +- $out .= sprintf ", walltime=0x%08x", $walltime if ($walltime); +- $out .= sprintf ", cpu=0x%08x", $cpu if ($cpu); +- $out .= sprintf ", cpuid=0x%08x", $cpuid if ($cpuid); +- $out .= sprintf ", apicid=0x%08x", $apicid if ($apicid); +- $out .= sprintf ", socketid=0x%08x", $socketid if ($socketid); +- $out .= sprintf ", cs=0x%08x", $cs if ($cs); +- $out .= sprintf ", bank=0x%08x", $bank if ($bank); +- +- $out .= "\n"; +- } +- if ($out ne "") { +- print "MCE events:\n$out\n"; +- } else { +- print "No MCE errors.\n\n"; ++ if ($has_mce == 1) { ++ $query = "select id, timestamp, mcgcap, mcgstatus, status, addr, misc, ip, tsc, walltime, cpu, cpuid, apicid, socketid, cs, bank, cpuvendor, bank_name, error_msg, mcgstatus_msg, mcistatus_msg, user_action, mc_location from mce_record order by id"; ++ $query_handle = $dbh->prepare($query); ++ $query_handle->execute(); ++ $query_handle->bind_columns(\($id, $time, $mcgcap,$mcgstatus, $status, $addr, $misc, $ip, $tsc, $walltime, $cpu, $cpuid, $apicid, $socketid, $cs, $bank, $cpuvendor, $bank_name, $msg, $mcgstatus_msg, $mcistatus_msg, $user_action, $mc_location)); ++ $out = ""; ++ while($query_handle->fetch()) { ++ $out .= "$id $time error: $msg"; ++ $out .= ", CPU $cpuvendor" if ($cpuvendor); ++ $out .= ", bank $bank_name" if ($bank_name); ++ $out .= ", mcg $mcgstatus_msg" if ($mcgstatus_msg); ++ $out .= ", mci $mcistatus_msg" if ($mcistatus_msg); ++ $out .= ", $mc_location" if ($mc_location); ++ $out .= ", $user_action" if ($user_action); ++ $out .= sprintf ", mcgcap=0x%08x", $mcgcap if ($mcgcap); ++ $out .= sprintf ", mcgstatus=0x%08x", $mcgstatus if ($mcgstatus); ++ $out .= sprintf ", status=0x%08x", $status if ($status); ++ $out .= sprintf ", addr=0x%08x", $addr if ($addr); ++ $out .= sprintf ", misc=0x%08x", $misc if ($misc); ++ $out .= sprintf ", ip=0x%08x", $ip if ($ip); ++ $out .= sprintf ", tsc=0x%08x", $tsc if ($tsc); ++ $out .= sprintf ", walltime=0x%08x", $walltime if ($walltime); ++ $out .= sprintf ", cpu=0x%08x", $cpu if ($cpu); ++ $out .= sprintf ", cpuid=0x%08x", $cpuid if ($cpuid); ++ $out .= sprintf ", apicid=0x%08x", $apicid if ($apicid); ++ $out .= sprintf ", socketid=0x%08x", $socketid if ($socketid); ++ $out .= sprintf ", cs=0x%08x", $cs if ($cs); ++ $out .= sprintf ", bank=0x%08x", $bank if ($bank); ++ ++ $out .= "\n"; ++ } ++ if ($out ne "") { ++ print "MCE events:\n$out\n"; ++ } else { ++ print "No MCE errors.\n\n"; ++ } ++ $query_handle->finish; + } +- $query_handle->finish; + + undef($dbh); + } diff --git a/SOURCES/3f9af0adc4c11b60c3969c0fc0bb209c0fbfdc8e.patch b/SOURCES/3f9af0adc4c11b60c3969c0fc0bb209c0fbfdc8e.patch new file mode 100644 index 0000000..52a793c --- /dev/null +++ b/SOURCES/3f9af0adc4c11b60c3969c0fc0bb209c0fbfdc8e.patch @@ -0,0 +1,50 @@ +commit 3f9af0adc4c11b60c3969c0fc0bb209c0fbfdc8e +Author: dann frazier +Date: Tue Apr 21 15:56:04 2020 -0600 + + ras-mc-ctl: PCIe AER: display PCIe dev name + + Storage of PCIe dev name was added in commit 8e96ca2c1c59 ("rasdaemon: + store PCIe dev name and TLP header for the aer event"). This makes + ras-mc-ctl extract and emit it like so: + + PCIe AER events: + 1 2020-04-16 22:09:48 +0000 0000:0b:00.0 Corrected error: Receiver Error + 2 2020-04-16 22:23:24 +0000 0000:0b:00.0 Corrected error: Receiver Error + 3 2020-04-17 23:00:37 +0000 0000:d9:01.0 Corrected error: Advisory Non-Fatal, BIT15 + 4 2020-04-17 23:21:52 +0000 0000:d9:01.0 Corrected error: Advisory Non-Fatal + 5 2020-04-18 02:04:24 +0000 0000:5e:00.0 Corrected error: Receiver Error + + Signed-off-by: Dann Frazier + Tested-by: Shiju Jose + +diff --git a/util/ras-mc-ctl.in b/util/ras-mc-ctl.in +index f361fd0..dd7d56f 100755 +--- a/util/ras-mc-ctl.in ++++ b/util/ras-mc-ctl.in +@@ -1247,7 +1247,7 @@ sub summary + sub errors + { + require DBI; +- my ($query, $query_handle, $id, $time, $count, $type, $msg, $label, $mc, $top, $mid, $low, $addr, $grain, $syndrome, $detail, $out); ++ my ($query, $query_handle, $id, $time, $devname, $count, $type, $msg, $label, $mc, $top, $mid, $low, $addr, $grain, $syndrome, $detail, $out); + my ($mcgcap,$mcgstatus, $status, $misc, $ip, $tsc, $walltime, $cpu, $cpuid, $apicid, $socketid, $cs, $bank, $cpuvendor, $bank_name, $mcgstatus_msg, $mcistatus_msg, $user_action, $mc_location); + my ($timestamp, $etype, $severity, $etype_string, $severity_string, $fru_id, $fru_text, $cper_data); + my ($bus_name, $dev_name, $driver_name, $reporter_name); +@@ -1277,13 +1277,13 @@ sub errors + $query_handle->finish; + + # PCIe AER aer_event errors +- $query = "select id, timestamp, err_type, err_msg from aer_event order by id"; ++ $query = "select id, timestamp, dev_name, err_type, err_msg from aer_event order by id"; + $query_handle = $dbh->prepare($query); + $query_handle->execute(); +- $query_handle->bind_columns(\($id, $time, $type, $msg)); ++ $query_handle->bind_columns(\($id, $time, $devname, $type, $msg)); + $out = ""; + while($query_handle->fetch()) { +- $out .= "$id $time $type error: $msg\n"; ++ $out .= "$id $time $devname $type error: $msg\n"; + } + if ($out ne "") { + print "PCIe AER events:\n$out\n"; diff --git a/SOURCES/51f989b7c599df88a2aad8dd27b129f448009fc4.patch b/SOURCES/51f989b7c599df88a2aad8dd27b129f448009fc4.patch new file mode 100644 index 0000000..420f85d --- /dev/null +++ b/SOURCES/51f989b7c599df88a2aad8dd27b129f448009fc4.patch @@ -0,0 +1,83 @@ +commit 51f989b7c599df88a2aad8dd27b129f448009fc4 +Author: Shiju Jose +Date: Tue Aug 11 13:31:46 2020 +0100 + + rasdaemon: ras-mc-ctl: Add ARM processor error information + + Add supporting ARM processor error in the ras-mc-ctl tool. + + Signed-off-by: Shiju Jose + Signed-off-by: Mauro Carvalho Chehab + +diff --git a/util/ras-mc-ctl.in b/util/ras-mc-ctl.in +index 5a4aaf6..f361fd0 100755 +--- a/util/ras-mc-ctl.in ++++ b/util/ras-mc-ctl.in +@@ -1123,6 +1123,7 @@ sub summary + my ($err_type, $label, $mc, $top, $mid, $low, $count, $msg); + my ($etype, $severity, $etype_string, $severity_string); + my ($dev_name, $dev); ++ my ($affinity, $mpidr); + + my $dbh = DBI->connect("dbi:SQLite:dbname=$dbname", "", "", {}); + +@@ -1158,6 +1159,22 @@ sub summary + } + $query_handle->finish; + ++ # ARM processor arm_event errors ++ $query = "select affinity, mpidr, count(*) from arm_event group by affinity, mpidr"; ++ $query_handle = $dbh->prepare($query); ++ $query_handle->execute(); ++ $query_handle->bind_columns(\($affinity, $mpidr, $count)); ++ $out = ""; ++ while($query_handle->fetch()) { ++ $out .= "\t$count errors\n"; ++ } ++ if ($out ne "") { ++ print "ARM processor events summary:\n$out\n"; ++ } else { ++ print "No ARM processor errors.\n\n"; ++ } ++ $query_handle->finish; ++ + # extlog errors + $query = "select etype, severity, count(*) from extlog_event group by etype, severity"; + $query_handle = $dbh->prepare($query); +@@ -1235,6 +1252,7 @@ sub errors + my ($timestamp, $etype, $severity, $etype_string, $severity_string, $fru_id, $fru_text, $cper_data); + my ($bus_name, $dev_name, $driver_name, $reporter_name); + my ($dev, $sector, $nr_sector, $error, $rwbs, $cmd); ++ my ($error_count, $affinity, $mpidr, $r_state, $psci_state); + + my $dbh = DBI->connect("dbi:SQLite:dbname=$dbname", "", "", {}); + +@@ -1274,6 +1292,28 @@ sub errors + } + $query_handle->finish; + ++ # ARM processor arm_event errors ++ $query = "select id, timestamp, error_count, affinity, mpidr, running_state, psci_state from arm_event order by id"; ++ $query_handle = $dbh->prepare($query); ++ $query_handle->execute(); ++ $query_handle->bind_columns(\($id, $timestamp, $error_count, $affinity, $mpidr, $r_state, $psci_state)); ++ $out = ""; ++ while($query_handle->fetch()) { ++ $out .= "$id $timestamp error: "; ++ $out .= "error_count=$error_count, " if ($error_count); ++ $out .= "affinity_level=$affinity, "; ++ $out .= sprintf "mpidr=0x%x, ", $mpidr; ++ $out .= sprintf "running_state=0x%x, ", $r_state; ++ $out .= sprintf "psci_state=0x%x", $psci_state; ++ $out .= "\n"; ++ } ++ if ($out ne "") { ++ print "ARM processor events:\n$out\n"; ++ } else { ++ print "No ARM processor errors.\n\n"; ++ } ++ $query_handle->finish; ++ + # Extlog errors + $query = "select id, timestamp, etype, severity, address, fru_id, fru_text, cper_data from extlog_event order by id"; + $query_handle = $dbh->prepare($query); diff --git a/SOURCES/659abc11853a4bfc7e9dd51f7c7a90449b1747a9.patch b/SOURCES/659abc11853a4bfc7e9dd51f7c7a90449b1747a9.patch new file mode 100644 index 0000000..1f27670 --- /dev/null +++ b/SOURCES/659abc11853a4bfc7e9dd51f7c7a90449b1747a9.patch @@ -0,0 +1,24 @@ +commit 659abc11853a4bfc7e9dd51f7c7a90449b1747a9 +Author: Cong Wang +Date: Thu Feb 27 16:24:06 2020 -0800 + + Fix a typo in ras-mc-ctl + + Signed-off-by: Cong Wang + Signed-off-by: Mauro Carvalho Chehab + +diff --git a/util/ras-mc-ctl.in b/util/ras-mc-ctl.in +index 8d6d866..b4e5856 100755 +--- a/util/ras-mc-ctl.in ++++ b/util/ras-mc-ctl.in +@@ -270,8 +270,8 @@ sub parse_dimm_nodes + for (my $i = 1; $i < scalar(@temp); $i += 2) { + $pos[$i / 2] = $temp[$i]; + +- if ($pos[$i / 2] > $max_pos[$i / 2]) { +- $max_pos[$i / 2 + 1] = $pos[$i / 2]; ++ if ($pos[$i / 2] > $max_pos[$i / 2 + 1]) { ++ $max_pos[$i / 2 + 1] = $pos[$i / 2]; + } + } + if ($mc > $max_pos[0]) { diff --git a/SPECS/rasdaemon.spec b/SPECS/rasdaemon.spec index 260f13e..96f0e5c 100644 --- a/SPECS/rasdaemon.spec +++ b/SPECS/rasdaemon.spec @@ -24,6 +24,12 @@ Requires(post): systemd Requires(preun): systemd Requires(postun): systemd +Patch1: 659abc11853a4bfc7e9dd51f7c7a90449b1747a9.patch +Patch2: 129f20ba5e0b8b5881539e0038e97ef353d5d294.patch +Patch3: 51f989b7c599df88a2aad8dd27b129f448009fc4.patch +Patch4: 3f9af0adc4c11b60c3969c0fc0bb209c0fbfdc8e.patch +Patch5: 3f1fa60f2fef32c05bed1f21617311216ee80026.patch + %description %{name} is a RAS (Reliability, Availability and Serviceability) logging tool. It currently records memory errors, using the EDAC tracing events. @@ -36,6 +42,11 @@ an utility for reporting current error counts from the EDAC sysfs files. %prep %setup -q +%patch1 -p1 +%patch2 -p1 +%patch3 -p1 +%patch4 -p1 +%patch5 -p1 %build %ifarch %{arm} aarch64 @@ -62,8 +73,7 @@ rm INSTALL %{buildroot}/usr/include/*.h %changelog * Tue Feb 23 2021 David Johansen - 0.6.6-1 -- Bump to version 0.6.6 -- cherry-picked ras-mc-ctl fixes: 059a901, d98326e, 5d00690, 0862a09, 546cf71 +- Bump to version 0.6.6 and include patches for ras-mc-ctl fixes * Wed Jan 27 2021 Fedora Release Engineering - 0.6.4-4 - Rebuilt for https://fedoraproject.org/wiki/Fedora_34_Mass_Rebuild