rpms/rasdaemon
6
0
Fork 0
Code Issues Pull Requests Releases Activity
c9-beta
rasdaemon/SOURCES/045ab08eaa00172d50621df9502f6910f3fe3af4.patch

155 lines
8.2 KiB
Diff
Raw Permalink Blame History

commit 045ab08eaa00172d50621df9502f6910f3fe3af4
Author: Avadhut Naik <avadhut.naik@amd.com>
Date: Mon Apr 1 23:33:07 2024 -0500
rasdaemon: Add support to parse the PPIN field of mce tracepoint
Support for exporting the PPIN (Protected Processor Inventory Number)
is being added to the mce_record tracepoint.
Add the required, corresponding support in the rasdaemon for the field
to be parsed and logged or added to the database and viewed later through
ras-mc-ctl utility.
Signed-off-by: Avadhut Naik <avadhut.naik@amd.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
---
ras-mce-handler.c | 7 +++++++
ras-mce-handler.h | 1 +
ras-record.c | 42 ++++++++++++++++++++++--------------------
util/ras-mc-ctl.in | 7 ++++---
4 files changed, 34 insertions(+), 23 deletions(-)
--- rasdaemon-0.6.7.orig/ras-mce-handler.c 2024-08-22 14:30:41.581276901 -0400
+++ rasdaemon-0.6.7/ras-mce-handler.c 2024-08-22 14:33:19.940957855 -0400
@@ -369,6 +369,9 @@ #if 0
trace_seq_printf(s, ", apicid= %x", e->apicid);
+ if (e->ppin)
+ trace_seq_printf(s, ", ppin= %llx", (long long)e->ppin);
+
/*
* FIXME: The original mcelog userspace tool uses DMI to map from
* address to DIMM. From the comments there, the code there doesn't
@@ -545,6 +548,10 @@ if (pevent_get_field_val(s, event, "ipid
return -1;
e.ipid = val;
+ /* Get PPIN */
+ if (!pevent_get_field_val(s, event, "ppin", record, &val, 1))
+ e.ppin = val;
+
switch (mce->cputype) {
case CPU_GENERIC:
break;
--- rasdaemon-0.6.7.orig/ras-mce-handler.h 2024-08-22 14:30:41.581276901 -0400
+++ rasdaemon-0.6.7/ras-mce-handler.h 2024-08-22 14:33:47.312729865 -0400
@@ -74,6 +74,7 @@ struct mce_event {
uint8_t cpuvendor;
uint64_t synd; /* MCA_SYND MSR: only valid on SMCA systems */
uint64_t ipid; /* MCA_IPID MSR: only valid on SMCA systems */
+ uint64_t ppin;
/* Parsed data */
char timestamp[64];
--- rasdaemon-0.6.7.orig/ras-record.c 2024-08-22 14:30:41.581276901 -0400
+++ rasdaemon-0.6.7/ras-record.c 2024-08-22 14:37:20.884941279 -0400
@@ -330,19 +330,20 @@ static const struct db_fields mce_record
{ .name="ip", .type="INTEGER" },
{ .name="tsc", .type="INTEGER" },
{ .name="walltime", .type="INTEGER" },
- { .name="cpu", .type="INTEGER" }, // 10
+ { .name = "ppin", .type = "INTEGER" }, // 10
+ { .name="cpu", .type="INTEGER" },
{ .name="cpuid", .type="INTEGER" },
{ .name="apicid", .type="INTEGER" },
{ .name="socketid", .type="INTEGER" },
- { .name="cs", .type="INTEGER" },
- { .name="bank", .type="INTEGER" }, //15
+ { .name="cs", .type="INTEGER" }, //15
+ { .name="bank", .type="INTEGER" },
{ .name="cpuvendor", .type="INTEGER" },
/* Parsed data - will likely change */
{ .name="bank_name", .type="TEXT" },
{ .name="error_msg", .type="TEXT" },
- { .name="mcgstatus_msg", .type="TEXT" },
- { .name="mcistatus_msg", .type="TEXT" }, // 20
+ { .name="mcgstatus_msg", .type="TEXT" }, // 20
+ { .name="mcistatus_msg", .type="TEXT" },
{ .name="mcastatus_msg", .type="TEXT" },
{ .name="user_action", .type="TEXT" },
{ .name="mc_location", .type="TEXT" },
@@ -372,21 +373,22 @@ return 0;
sqlite3_bind_int64 (priv->stmt_mce_record, 7, ev->ip);
sqlite3_bind_int64 (priv->stmt_mce_record, 8, ev->tsc);
sqlite3_bind_int64 (priv->stmt_mce_record, 9, ev->walltime);
- sqlite3_bind_int (priv->stmt_mce_record, 10, ev->cpu);
- sqlite3_bind_int (priv->stmt_mce_record, 11, ev->cpuid);
- sqlite3_bind_int (priv->stmt_mce_record, 12, ev->apicid);
- sqlite3_bind_int (priv->stmt_mce_record, 13, ev->socketid);
- sqlite3_bind_int (priv->stmt_mce_record, 14, ev->cs);
- sqlite3_bind_int (priv->stmt_mce_record, 15, ev->bank);
- sqlite3_bind_int (priv->stmt_mce_record, 16, ev->cpuvendor);
-
- sqlite3_bind_text(priv->stmt_mce_record, 17, ev->bank_name, -1, NULL);
- sqlite3_bind_text(priv->stmt_mce_record, 18, ev->error_msg, -1, NULL);
- sqlite3_bind_text(priv->stmt_mce_record, 19, ev->mcgstatus_msg, -1, NULL);
- sqlite3_bind_text(priv->stmt_mce_record, 20, ev->mcistatus_msg, -1, NULL);
- sqlite3_bind_text(priv->stmt_mce_record, 21, ev->mcastatus_msg, -1, NULL);
- sqlite3_bind_text(priv->stmt_mce_record, 22, ev->user_action, -1, NULL);
- sqlite3_bind_text(priv->stmt_mce_record, 23, ev->mc_location, -1, NULL);
+ sqlite3_bind_int64(priv->stmt_mce_record, 10, ev->ppin);
+ sqlite3_bind_int (priv->stmt_mce_record, 11, ev->cpu);
+ sqlite3_bind_int (priv->stmt_mce_record, 12, ev->cpuid);
+ sqlite3_bind_int (priv->stmt_mce_record, 13, ev->apicid);
+ sqlite3_bind_int (priv->stmt_mce_record, 14, ev->socketid);
+ sqlite3_bind_int (priv->stmt_mce_record, 15, ev->cs);
+ sqlite3_bind_int (priv->stmt_mce_record, 16, ev->bank);
+ sqlite3_bind_int (priv->stmt_mce_record, 17, ev->cpuvendor);
+
+ sqlite3_bind_text(priv->stmt_mce_record, 18, ev->bank_name, -1, NULL);
+ sqlite3_bind_text(priv->stmt_mce_record, 19, ev->error_msg, -1, NULL);
+ sqlite3_bind_text(priv->stmt_mce_record, 20, ev->mcgstatus_msg, -1, NULL);
+ sqlite3_bind_text(priv->stmt_mce_record, 21, ev->mcistatus_msg, -1, NULL);
+ sqlite3_bind_text(priv->stmt_mce_record, 22, ev->mcastatus_msg, -1, NULL);
+ sqlite3_bind_text(priv->stmt_mce_record, 23, ev->user_action, -1, NULL);
+ sqlite3_bind_text(priv->stmt_mce_record, 24, ev->mc_location, -1, NULL);
rc = sqlite3_step(priv->stmt_mce_record);
if (rc != SQLITE_OK && rc != SQLITE_DONE)
--- rasdaemon-0.6.7.orig/util/ras-mc-ctl.in 2024-08-22 14:29:51.058697724 -0400
+++ rasdaemon-0.6.7/util/ras-mc-ctl.in 2024-08-22 14:34:53.803175632 -0400
@@ -1317,7 +1317,7 @@ sub errors
{
require DBI;
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, $mcastatus_msg, $user_action, $mc_location);
+ my ($mcgcap,$mcgstatus, $status, $misc, $ip, $tsc, $walltime, $ppin, $cpu, $cpuid, $apicid, $socketid, $cs, $bank, $cpuvendor, $bank_name, $mcgstatus_msg, $mcistatus_msg, $mcastatus_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, $reporter_name);
my ($dev, $sector, $nr_sector, $error, $rwbs, $cmd);
@@ -1485,10 +1485,10 @@ $out .= sprintf "address=0x%08x, ", $add
# MCE mce_record errors
if ($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, mcastatus_msg, user_action, mc_location from mce_record order by id";
+ $query = "select id, timestamp, mcgcap, mcgstatus, status, addr, misc, ip, tsc, walltime, ppin, cpu, cpuid, apicid, socketid, cs, bank, cpuvendor, bank_name, error_msg, mcgstatus_msg, mcistatus_msg, mcastatus_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, $mcastatus_msg, $user_action, $mc_location));
+ $query_handle->bind_columns(\($id, $time, $mcgcap,$mcgstatus, $status, $addr, $misc, $ip, $tsc, $walltime, $ppin, $cpu, $cpuid, $apicid, $socketid, $cs, $bank, $cpuvendor, $bank_name, $msg, $mcgstatus_msg, $mcistatus_msg, $mcastatus_msg, $user_action, $mc_location));
$out = "";
while($query_handle->fetch()) {
$out .= "$id $time error: $msg";
@@ -1507,6 +1507,7 @@ $out .= sprintf ", misc=0x%08x", $misc i
$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 ", ppin=0x%08x", $ppin if ($ppin);
$out .= sprintf ", cpu=0x%08x", $cpu if ($cpu);
$out .= sprintf ", cpuid=0x%08x", $cpuid if ($cpuid);
$out .= sprintf ", apicid=0x%08x", $apicid if ($apicid);
Reference in New Issue View Git Blame Copy Permalink