Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix ras-mc-ctl bug and add support for parsing the new fields added to MCE tracepoint #152

Closed
wants to merge 3 commits into from
Closed

Fix ras-mc-ctl bug and add support for parsing the new fields added to MCE tracepoint #152

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
7111fd2
rasdaemon: ras-mc-ctl: Add support to display mcastatus_msg string
Mar 26, 2024
a83def7
rasdaemon: Add support to parse the PPIN field of mce tracepoint
Apr 2, 2024
6d3bf61
rasdaemon: Add support to parse microcode field of mce tracepoint
Apr 2, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 14 additions & 0 deletions ras-mce-handler.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -372,6 +372,12 @@ static void report_mce_event(struct ras_events *ras,

trace_seq_printf(s, ", apicid= %x", e->apicid);

if (e->ppin)
trace_seq_printf(s, ", ppin= %llx", (long long)e->ppin);

if (e->microcode)
trace_seq_printf(s, ", microcode= %x", e->microcode);

if (!e->vdata_len)
return;

Expand Down Expand Up @@ -566,6 +572,14 @@ int ras_mce_event_handler(struct trace_seq *s,
return -1;
e.ipid = val;

/* Get PPIN */
if (!tep_get_field_val(s, event, "ppin", record, &val, 1))
e.ppin = val;

/* Get Microcode Revision */
if (!tep_get_field_val(s, event, "microcode", record, &val, 1))
e.microcode = val;

/* Get Vendor-specfic Data, if any */
e.vdata = tep_get_field_raw(s, event, "v_data", record, &e.vdata_len, 1);

Expand Down
2 changes: 2 additions & 0 deletions ras-mce-handler.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -75,6 +75,8 @@ 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;
uint32_t microcode;
int32_t vdata_len;
const uint64_t *vdata;

Expand Down
44 changes: 24 additions & 20 deletions ras-record.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -330,19 +330,21 @@ static const struct db_fields mce_record_fields[] = {
{ .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" },
{ .name = "microcode", .type = "INTEGER" },

/* Parsed data - will likely change */
{ .name = "bank_name", .type = "TEXT" },
{ .name = "error_msg", .type = "TEXT" },
{ .name = "error_msg", .type = "TEXT" }, // 20
{ .name = "mcgstatus_msg", .type = "TEXT" },
{ .name = "mcistatus_msg", .type = "TEXT" }, // 20
{ .name = "mcistatus_msg", .type = "TEXT" },
{ .name = "mcastatus_msg", .type = "TEXT" },
{ .name = "user_action", .type = "TEXT" },
{ .name = "mc_location", .type = "TEXT" },
Expand Down Expand Up @@ -372,21 +374,23 @@ int ras_store_mce_record(struct ras_events *ras, struct mce_event *ev)
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_int (priv->stmt_mce_record, 18, ev->microcode);

sqlite3_bind_text(priv->stmt_mce_record, 19, ev->bank_name, -1, NULL);
sqlite3_bind_text(priv->stmt_mce_record, 20, ev->error_msg, -1, NULL);
sqlite3_bind_text(priv->stmt_mce_record, 21, ev->mcgstatus_msg, -1, NULL);
sqlite3_bind_text(priv->stmt_mce_record, 22, ev->mcistatus_msg, -1, NULL);
sqlite3_bind_text(priv->stmt_mce_record, 23, ev->mcastatus_msg, -1, NULL);
sqlite3_bind_text(priv->stmt_mce_record, 24, ev->user_action, -1, NULL);
sqlite3_bind_text(priv->stmt_mce_record, 25, ev->mc_location, -1, NULL);

rc = sqlite3_step(priv->stmt_mce_record);
if (rc != SQLITE_OK && rc != SQLITE_DONE)
Expand Down
9 changes: 6 additions & 3 deletions util/ras-mc-ctl.in
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1329,7 +1329,7 @@ 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, $user_action, $mc_location);
my ($mcgcap,$mcgstatus, $status, $misc, $ip, $tsc, $walltime, $ppin, $cpu, $cpuid, $apicid, $socketid, $cs, $bank, $cpuvendor, $microcode, $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);
Expand Down Expand Up @@ -1497,17 +1497,18 @@ sub errors

# MCE mce_record errors
if ($has_mce == 1) {
$query = "select id, timestamp, mcgcap, mcgstatus, status, addr, misc, ip, tsc, walltime, cpu, cpuid, apicid, socketid, cs, bank, cpuvendor, bank_name, error_msg, mcgstatus_msg, mcistatus_msg, user_action, mc_location from mce_record$conf{opt}{since} order by id";
$query = "select id, timestamp, mcgcap, mcgstatus, status, addr, misc, ip, tsc, walltime, ppin, cpu, cpuid, apicid, socketid, cs, bank, cpuvendor, microcode, bank_name, error_msg, mcgstatus_msg, mcistatus_msg, mcastatus_msg, user_action, mc_location from mce_record$conf{opt}{since} 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));
$query_handle->bind_columns(\($id, $time, $mcgcap,$mcgstatus, $status, $addr, $misc, $ip, $tsc, $walltime, $ppin, $cpu, $cpuid, $apicid, $socketid, $cs, $bank, $cpuvendor, $microcode, $bank_name, $msg, $mcgstatus_msg, $mcistatus_msg, $mcastatus_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 .= ", mca $mcastatus_msg" if ($mcastatus_msg);
$out .= ", $mc_location" if ($mc_location);
$out .= ", $user_action" if ($user_action);
$out .= sprintf ", mcgcap=0x%08x", $mcgcap if ($mcgcap);
Expand All @@ -1518,12 +1519,14 @@ sub errors
$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);
$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 .= sprintf ", microcode=0x%08x", $microcode if ($microcode);

$out .= "\n";
}
Expand Down
Loading