@@ -73,6 +73,11 @@ endif
if WITH_CPU_FAULT_ISOLATION
rasdaemon_SOURCES += ras-cpu-isolation.c queue.c
endif
+
+if WITH_CXL
+ rasdaemon_SOURCES += ras-cxl-handler.c
+endif
+
rasdaemon_LDADD = -lpthread $(SQLITE3_LIBS) $(LIBTRACEEVENT_LIBS)
rasdaemon_CFLAGS = $(SQLITE3_CFLAGS) $(LIBTRACEEVENT_CFLAGS)
@@ -81,7 +86,7 @@ include_HEADERS = config.h ras-events.h ras-logger.h ras-mc-handler.h \
ras-extlog-handler.h ras-arm-handler.h ras-non-standard-handler.h \
ras-devlink-handler.h ras-diskerror-handler.h rbtree.h ras-page-isolation.h \
non-standard-hisilicon.h non-standard-ampere.h ras-memory-failure-handler.h \
- ras-cpu-isolation.h queue.h
+ ras-cxl-handler.h ras-cpu-isolation.h queue.h
# This rule can't be called with more than one Makefile job (like make -j8)
# I can't figure out a way to fix that
@@ -127,6 +127,16 @@ AS_IF([test "x$enable_memory_failure" = "xyes" || test "x$enable_all" = "xyes"],
AM_CONDITIONAL([WITH_MEMORY_FAILURE], [test x$enable_memory_failure = xyes || test x$enable_all = xyes])
AM_COND_IF([WITH_MEMORY_FAILURE], [USE_MEMORY_FAILURE="yes"], [USE_MEMORY_FAILURE="no"])
+AC_ARG_ENABLE([cxl],
+ AS_HELP_STRING([--enable-cxl], [enable CXL events (currently experimental)]))
+
+AS_IF([test "x$enable_cxl" = "xyes" || test "x$enable_all" == "xyes"], [
+ AC_DEFINE(HAVE_CXL,1,"have CXL events collect")
+ AC_SUBST([WITH_CXL])
+])
+AM_CONDITIONAL([WITH_CXL], [test x$enable_cxl = xyes || test x$enable_all == xyes])
+AM_COND_IF([WITH_CXL], [USE_CXL="yes"], [USE_CXL="no"])
+
AC_ARG_ENABLE([abrt_report],
AS_HELP_STRING([--enable-abrt-report], [enable report event to ABRT (currently experimental)]))
@@ -215,6 +225,7 @@ compile time options summary
DEVLINK : $USE_DEVLINK
Disk I/O errors : $USE_DISKERROR
Memory Failure : $USE_MEMORY_FAILURE
+ CXL events : $USE_CXL
Memory CE PFA : $USE_MEMORY_CE_PFA
AMP RAS errors : $USE_AMP_NS_DECODE
CPU fault isolation : $USE_CPU_FAULT_ISOLATION
new file mode 100644
@@ -0,0 +1,162 @@
+/*
+ * Copyright (c) Huawei Technologies Co., Ltd. 2023. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <traceevent/kbuffer.h>
+#include "ras-cxl-handler.h"
+#include "ras-record.h"
+#include "ras-logger.h"
+#include "ras-report.h"
+
+/* Poison List: Payload out flags */
+#define CXL_POISON_FLAG_MORE BIT(0)
+#define CXL_POISON_FLAG_OVERFLOW BIT(1)
+#define CXL_POISON_FLAG_SCANNING BIT(2)
+
+/* CXL poison - source types */
+enum cxl_poison_source {
+ CXL_POISON_SOURCE_UNKNOWN = 0,
+ CXL_POISON_SOURCE_EXTERNAL = 1,
+ CXL_POISON_SOURCE_INTERNAL = 2,
+ CXL_POISON_SOURCE_INJECTED = 3,
+ CXL_POISON_SOURCE_VENDOR = 7,
+};
+
+int ras_cxl_poison_event_handler(struct trace_seq *s,
+ struct tep_record *record,
+ struct tep_event *event, void *context)
+{
+ int len;
+ unsigned long long val;
+ struct ras_events *ras = context;
+ time_t now;
+ struct tm *tm;
+ struct ras_cxl_poison_event ev;
+
+ now = record->ts/user_hz + ras->uptime_diff;
+ tm = localtime(&now);
+ if (tm)
+ strftime(ev.timestamp, sizeof(ev.timestamp),
+ "%Y-%m-%d %H:%M:%S %z", tm);
+ else
+ strncpy(ev.timestamp, "1970-01-01 00:00:00 +0000", sizeof(ev.timestamp));
+ trace_seq_printf(s, "%s ", ev.timestamp);
+
+ ev.memdev = tep_get_field_raw(s, event, "memdev",
+ record, &len, 1);
+ if (!ev.memdev)
+ return -1;
+ trace_seq_printf(s, "memdev:%s ", ev.memdev);
+
+ ev.pcidev = tep_get_field_raw(s, event, "pcidev",
+ record, &len, 1);
+ if (!ev.pcidev)
+ return -1;
+ trace_seq_printf(s, "pcidev:%s ", ev.pcidev);
+
+ ev.region = tep_get_field_raw(s, event, "region",
+ record, &len, 1);
+ if (!ev.region)
+ return -1;
+ trace_seq_printf(s, "region:%s ", ev.region);
+
+ ev.uuid = tep_get_field_raw(s, event, "uuid",
+ record, &len, 1);
+ if (!ev.uuid)
+ return -1;
+ trace_seq_printf(s, "region_uuid:%s ", ev.uuid);
+
+ if (tep_get_field_val(s, event, "hpa", record, &val, 1) < 0)
+ return -1;
+ ev.hpa = val;
+ trace_seq_printf(s, "poison list: hpa:0x%llx ", (unsigned long long)ev.hpa);
+
+ if (tep_get_field_val(s, event, "dpa", record, &val, 1) < 0)
+ return -1;
+ ev.dpa = val;
+ trace_seq_printf(s, "dpa:0x%llx ", (unsigned long long)ev.dpa);
+
+ if (tep_get_field_val(s, event, "length", record, &val, 1) < 0)
+ return -1;
+ ev.length = val;
+ trace_seq_printf(s, "length:%d ", ev.length);
+
+ if (tep_get_field_val(s, event, "source", record, &val, 1) < 0)
+ return -1;
+
+ switch (val) {
+ case CXL_POISON_SOURCE_UNKNOWN:
+ ev.source = "Unknown";
+ break;
+ case CXL_POISON_SOURCE_EXTERNAL:
+ ev.source = "External";
+ break;
+ case CXL_POISON_SOURCE_INTERNAL:
+ ev.source = "Internal";
+ break;
+ case CXL_POISON_SOURCE_INJECTED:
+ ev.source = "Injected";
+ break;
+ case CXL_POISON_SOURCE_VENDOR:
+ ev.source = "Vendor";
+ break;
+ default:
+ ev.source = "Invalid";
+ }
+ trace_seq_printf(s, "source:%s ", ev.source);
+
+ if (tep_get_field_val(s, event, "flags", record, &val, 1) < 0)
+ return -1;
+ ev.flags = val;
+ trace_seq_printf(s, "flags:%d ", ev.flags);
+
+ if (ev.flags & CXL_POISON_FLAG_OVERFLOW) {
+ if (tep_get_field_val(s, event, "overflow_t", record, &val, 1) < 0)
+ return -1;
+ if (val) {
+ /* CXL Specification 3.0
+ * Overflow timestamp - The number of unsigned nanoseconds
+ * that have elapsed since midnight, 01-Jan-1970 UTC
+ */
+ time_t ovf_ts_secs = val / 1000000000ULL;
+
+ tm = localtime(&ovf_ts_secs);
+ if (tm) {
+ strftime(ev.overflow_ts, sizeof(ev.overflow_ts),
+ "%Y-%m-%d %H:%M:%S %z", tm);
+ }
+ }
+ if (!val || !tm)
+ strncpy(ev.overflow_ts, "1970-01-01 00:00:00 +0000",
+ sizeof(ev.overflow_ts));
+ } else
+ strncpy(ev.overflow_ts, "1970-01-01 00:00:00 +0000", sizeof(ev.overflow_ts));
+ trace_seq_printf(s, "overflow timestamp:%s ", ev.overflow_ts);
+ trace_seq_printf(s, "\n");
+
+ /* Insert data into the SGBD */
+#ifdef HAVE_SQLITE3
+ ras_store_cxl_poison_event(ras, &ev);
+#endif
+
+#ifdef HAVE_ABRT_REPORT
+ /* Report event to ABRT */
+ ras_report_cxl_poison_event(ras, &ev);
+#endif
+
+ return 0;
+}
new file mode 100644
@@ -0,0 +1,24 @@
+/*
+ * Copyright (c) Huawei Technologies Co., Ltd. 2023. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ */
+
+#ifndef __RAS_CXL_HANDLER_H
+#define __RAS_CXL_HANDLER_H
+
+#include "ras-events.h"
+#include <traceevent/event-parse.h>
+
+int ras_cxl_poison_event_handler(struct trace_seq *s,
+ struct tep_record *record,
+ struct tep_event *event, void *context);
+#endif
@@ -40,6 +40,7 @@
#include "ras-devlink-handler.h"
#include "ras-diskerror-handler.h"
#include "ras-memory-failure-handler.h"
+#include "ras-cxl-handler.h"
#include "ras-record.h"
#include "ras-logger.h"
#include "ras-page-isolation.h"
@@ -243,6 +244,10 @@ int toggle_ras_mc_event(int enable)
rc |= __toggle_ras_mc_event(ras, "ras", "memory_failure_event", enable);
#endif
+#ifdef HAVE_CXL
+ rc |= __toggle_ras_mc_event(ras, "cxl", "cxl_poison", enable);
+#endif
+
free_ras:
free(ras);
return rc;
@@ -951,6 +956,16 @@ int handle_ras_events(int record_events)
"ras", "memory_failure_event");
#endif
+#ifdef HAVE_CXL
+ rc = add_event_handler(ras, pevent, page_size, "cxl", "cxl_poison",
+ ras_cxl_poison_event_handler, NULL, CXL_POISON_EVENT);
+ if (!rc)
+ num_events++;
+ else
+ log(ALL, LOG_ERR, "Can't get traces from %s:%s\n",
+ "cxl", "cxl_poison");
+#endif
+
if (!num_events) {
log(ALL, LOG_INFO,
"Failed to trace all supported RAS events. Aborting.\n");
@@ -39,6 +39,7 @@ enum {
DEVLINK_EVENT,
DISKERROR_EVENT,
MF_EVENT,
+ CXL_POISON_EVENT,
NR_EVENTS
};
@@ -559,6 +559,67 @@ int ras_store_mf_event(struct ras_events *ras, struct ras_mf_event *ev)
}
#endif
+#ifdef HAVE_CXL
+/*
+ * Table and functions to handle cxl:cxl_poison
+ */
+static const struct db_fields cxl_poison_event_fields[] = {
+ { .name = "id", .type = "INTEGER PRIMARY KEY" },
+ { .name = "timestamp", .type = "TEXT" },
+ { .name = "memdev", .type = "TEXT" },
+ { .name = "pcidev", .type = "TEXT" },
+ { .name = "region", .type = "TEXT" },
+ { .name = "region_uuid", .type = "TEXT" },
+ { .name = "hpa", .type = "INTEGER" },
+ { .name = "dpa", .type = "INTEGER" },
+ { .name = "length", .type = "INTEGER" },
+ { .name = "source", .type = "TEXT" },
+ { .name = "flags", .type = "INTEGER" },
+ { .name = "overflow_ts", .type = "TEXT" },
+};
+
+static const struct db_table_descriptor cxl_poison_event_tab = {
+ .name = "cxl_poison_event",
+ .fields = cxl_poison_event_fields,
+ .num_fields = ARRAY_SIZE(cxl_poison_event_fields),
+};
+
+int ras_store_cxl_poison_event(struct ras_events *ras, struct ras_cxl_poison_event *ev)
+{
+ int rc;
+ struct sqlite3_priv *priv = ras->db_priv;
+
+ if (!priv || !priv->stmt_cxl_poison_event)
+ return 0;
+ log(TERM, LOG_INFO, "cxl_poison_event store: %p\n", priv->stmt_cxl_poison_event);
+
+ sqlite3_bind_text(priv->stmt_cxl_poison_event, 1, ev->timestamp, -1, NULL);
+ sqlite3_bind_text(priv->stmt_cxl_poison_event, 2, ev->memdev, -1, NULL);
+ sqlite3_bind_text(priv->stmt_cxl_poison_event, 3, ev->pcidev, -1, NULL);
+ sqlite3_bind_text(priv->stmt_cxl_poison_event, 4, ev->region, -1, NULL);
+ sqlite3_bind_text(priv->stmt_cxl_poison_event, 5, ev->uuid, -1, NULL);
+ sqlite3_bind_int64(priv->stmt_cxl_poison_event, 6, ev->hpa);
+ sqlite3_bind_int64(priv->stmt_cxl_poison_event, 7, ev->dpa);
+ sqlite3_bind_int(priv->stmt_cxl_poison_event, 8, ev->length);
+ sqlite3_bind_text(priv->stmt_cxl_poison_event, 9, ev->source, -1, NULL);
+ sqlite3_bind_int(priv->stmt_cxl_poison_event, 10, ev->flags);
+ sqlite3_bind_text(priv->stmt_cxl_poison_event, 11, ev->overflow_ts, -1, NULL);
+
+ rc = sqlite3_step(priv->stmt_cxl_poison_event);
+ if (rc != SQLITE_OK && rc != SQLITE_DONE)
+ log(TERM, LOG_ERR,
+ "Failed to do cxl_poison_event step on sqlite: error = %d\n", rc);
+ rc = sqlite3_reset(priv->stmt_cxl_poison_event);
+ if (rc != SQLITE_OK && rc != SQLITE_DONE)
+ log(TERM, LOG_ERR,
+ "Failed reset cxl_poison_event on sqlite: error = %d\n",
+ rc);
+ log(TERM, LOG_INFO, "register inserted at db\n");
+
+ return rc;
+}
+#endif
+
/*
* Generic code
*/
@@ -896,6 +957,16 @@ int ras_mc_event_opendb(unsigned cpu, struct ras_events *ras)
}
#endif
+#ifdef HAVE_CXL
+ rc = ras_mc_create_table(priv, &cxl_poison_event_tab);
+ if (rc == SQLITE_OK) {
+ rc = ras_mc_prepare_stmt(priv, &priv->stmt_cxl_poison_event,
+ &cxl_poison_event_tab);
+ if (rc != SQLITE_OK)
+ goto error;
+ }
+#endif
+
ras->db_priv = priv;
return 0;
@@ -1008,6 +1079,16 @@ int ras_mc_event_closedb(unsigned int cpu, struct ras_events *ras)
}
#endif
+#ifdef HAVE_CXL
+ if (priv->stmt_cxl_poison_event) {
+ rc = sqlite3_finalize(priv->stmt_cxl_poison_event);
+ if (rc != SQLITE_OK)
+ log(TERM, LOG_ERR,
+ "cpu %u: Failed to finalize cxl_poison_event sqlite: error = %d\n",
+ cpu, rc);
+ }
+#endif
+
rc = sqlite3_close_v2(db);
if (rc != SQLITE_OK)
log(TERM, LOG_ERR,
@@ -114,6 +114,20 @@ struct ras_mf_event {
const char *action_result;
};
+struct ras_cxl_poison_event {
+ char timestamp[64];
+ const char *memdev;
+ const char *pcidev;
+ const char *region;
+ const char *uuid;
+ uint64_t hpa;
+ uint64_t dpa;
+ uint32_t length;
+ const char *source;
+ uint8_t flags;
+ char overflow_ts[64];
+};
+
struct ras_mc_event;
struct ras_aer_event;
struct ras_extlog_event;
@@ -123,6 +137,7 @@ struct mce_event;
struct devlink_event;
struct diskerror_event;
struct ras_mf_event;
+struct ras_cxl_poison_event;
#ifdef HAVE_SQLITE3
@@ -155,6 +170,9 @@ struct sqlite3_priv {
#ifdef HAVE_MEMORY_FAILURE
sqlite3_stmt *stmt_mf_event;
#endif
+#ifdef HAVE_CXL
+ sqlite3_stmt *stmt_cxl_poison_event;
+#endif
};
struct db_fields {
@@ -182,6 +200,7 @@ int ras_store_arm_record(struct ras_events *ras, struct ras_arm_event *ev);
int ras_store_devlink_event(struct ras_events *ras, struct devlink_event *ev);
int ras_store_diskerror_event(struct ras_events *ras, struct diskerror_event *ev);
int ras_store_mf_event(struct ras_events *ras, struct ras_mf_event *ev);
+int ras_store_cxl_poison_event(struct ras_events *ras, struct ras_cxl_poison_event *ev);
#else
static inline int ras_mc_event_opendb(unsigned cpu, struct ras_events *ras) { return 0; };
@@ -195,6 +214,7 @@ static inline int ras_store_arm_record(struct ras_events *ras, struct ras_arm_ev
static inline int ras_store_devlink_event(struct ras_events *ras, struct devlink_event *ev) { return 0; };
static inline int ras_store_diskerror_event(struct ras_events *ras, struct diskerror_event *ev) { return 0; };
static inline int ras_store_mf_event(struct ras_events *ras, struct ras_mf_event *ev) { return 0; };
+static inline int ras_store_cxl_poison_event(struct ras_events *ras, struct ras_cxl_poison_event *ev) { return 0; };
#endif
@@ -331,6 +331,42 @@ static int set_mf_event_backtrace(char *buf, struct ras_mf_event *ev)
return 0;
}
+static int set_cxl_poison_event_backtrace(char *buf, struct ras_cxl_poison_event *ev)
+{
+ char bt_buf[MAX_BACKTRACE_SIZE];
+
+ if (!buf || !ev)
+ return -1;
+
+ sprintf(bt_buf, "BACKTRACE=" \
+ "timestamp=%s\n" \
+ "memdev=%s\n" \
+ "pcidev=%s\n" \
+ "region=%s\n" \
+ "uuid=%s\n" \
+ "hpa=0x%lx\n" \
+ "dpa=0x%lx\n" \
+ "length=%d\n" \
+ "source=%s\n" \
+ "flags=%d\n" \
+ "overflow_timestamp=%s\n" \
+ ev->timestamp, \
+ ev->memdev, \
+ ev->pcidev, \
+ ev->region, \
+ ev->uuid, \
+ ev->hpa, \
+ ev->dpa, \
+ ev->length, \
+ ev->source, \
+ ev->flags, \
+ ev->overflow_ts);
+
+ strcat(buf, bt_buf);
+
+ return 0;
+}
+
static int commit_report_backtrace(int sockfd, int type, void *ev){
char buf[MAX_BACKTRACE_SIZE];
char *pbuf = buf;
@@ -368,6 +404,9 @@ static int commit_report_backtrace(int sockfd, int type, void *ev){
case MF_EVENT:
rc = set_mf_event_backtrace(buf, (struct ras_mf_event *)ev);
break;
+ case CXL_POISON_EVENT:
+ rc = set_cxl_poison_event_backtrace(buf, (struct ras_cxl_poison_event *)ev);
+ break;
default:
return -1;
}
@@ -776,3 +815,47 @@ mf_fail:
else
return -1;
}
+
+int ras_report_cxl_poison_event(struct ras_events *ras, struct ras_cxl_poison_event *ev)
+{
+ char buf[MAX_MESSAGE_SIZE];
+ int sockfd = 0;
+ int done = 0;
+ int rc = -1;
+
+ memset(buf, 0, sizeof(buf));
+
+ sockfd = setup_report_socket();
+ if (sockfd < 0)
+ return -1;
+
+ rc = commit_report_basic(sockfd);
+ if (rc < 0)
+ goto cxl_poison_fail;
+
+ rc = commit_report_backtrace(sockfd, CXL_POISON_EVENT, ev);
+ if (rc < 0)
+ goto cxl_poison_fail;
+
+ sprintf(buf, "ANALYZER=%s", "rasdaemon-cxl-poison");
+ rc = write(sockfd, buf, strlen(buf) + 1);
+ if (rc < strlen(buf) + 1)
+ goto cxl_poison_fail;
+
+ sprintf(buf, "REASON=%s", "CXL poison");
+ rc = write(sockfd, buf, strlen(buf) + 1);
+ if (rc < strlen(buf) + 1)
+ goto cxl_poison_fail;
+
+ done = 1;
+
+cxl_poison_fail:
+
+ if (sockfd >= 0)
+ close(sockfd);
+
+ if (done)
+ return 0;
+ else
+ return -1;
+}
@@ -39,6 +39,7 @@ int ras_report_arm_event(struct ras_events *ras, struct ras_arm_event *ev);
int ras_report_devlink_event(struct ras_events *ras, struct devlink_event *ev);
int ras_report_diskerror_event(struct ras_events *ras, struct diskerror_event *ev);
int ras_report_mf_event(struct ras_events *ras, struct ras_mf_event *ev);
+int ras_report_cxl_poison_event(struct ras_events *ras, struct ras_cxl_poison_event *ev);
#else
@@ -50,6 +51,7 @@ static inline int ras_report_arm_event(struct ras_events *ras, struct ras_arm_ev
static inline int ras_report_devlink_event(struct ras_events *ras, struct devlink_event *ev) { return 0; };
static inline int ras_report_diskerror_event(struct ras_events *ras, struct diskerror_event *ev) { return 0; };
static inline int ras_report_mf_event(struct ras_events *ras, struct ras_mf_event *ev) { return 0; };
+static inline int ras_report_cxl_poison_event(struct ras_events *ras, struct ras_cxl_poison_event *ev) { return 0; };
#endif