rpms
/
rasdaemon
7
0
Fork 0
Code Issues Pull Requests Releases Activity
rasdaemon/SOURCES/2290d65b97311dd5736838f1e28...

539 lines
17 KiB
Diff
Raw Blame History

commit 2290d65b97311dd5736838f1e285355f7f357046
Author: Shiju Jose <shiju.jose@huawei.com>
Date: Mon Mar 8 16:57:26 2021 +0000
rasdaemon: add support for memory_failure events
Add support to log the memory_failure kernel trace
events.
Example rasdaemon log and SQLite DB output for the
memory_failure event,
=================================================
rasdaemon: memory_failure_event store: 0x126ce8f8
rasdaemon: register inserted at db
<...>-785 [000] 0.000024: memory_failure_event: 2020-10-02 13:27:13 -0400 pfn=0x204000000 page_type=free buddy page action_result=Delayed
CREATE TABLE memory_failure_event (id INTEGER PRIMARY KEY, timestamp TEXT, pfn TEXT, page_type TEXT, action_result TEXT);
INSERT INTO memory_failure_event VALUES(1,'2020-10-02 13:27:13 -0400','0x204000000','free buddy page','Delayed');
==================================================
Signed-off-by: Shiju Jose <shiju.jose@huawei.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
---
Makefile.am | 4
ras-events.c | 15 +++
ras-memory-failure-handler.c | 179 +++++++++++++++++++++++++++++++++++++++++++
ras-memory-failure-handler.h | 25 ++++++
ras-record.c | 56 +++++++++++++
ras-record.h | 13 +++
ras-report.c | 68 ++++++++++++++++
ras-report.h | 5 -
8 files changed, 364 insertions(+), 1 deletion(-)
--- /dev/null 1970-01-01 00:00:00.000000000 +0000
+++ b/ras-memory-failure-handler.c 2021-10-14 16:31:36.840657728 -0400
@@ -0,0 +1,179 @@
+/*
+ * Copyright (c) Huawei Technologies Co., Ltd. 2020. 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 "libtrace/kbuffer.h"
+#include "ras-memory-failure-handler.h"
+#include "ras-record.h"
+#include "ras-logger.h"
+#include "ras-report.h"
+
+/* Memory failure - various types of pages */
+enum mf_action_page_type {
+ MF_MSG_KERNEL,
+ MF_MSG_KERNEL_HIGH_ORDER,
+ MF_MSG_SLAB,
+ MF_MSG_DIFFERENT_COMPOUND,
+ MF_MSG_POISONED_HUGE,
+ MF_MSG_HUGE,
+ MF_MSG_FREE_HUGE,
+ MF_MSG_NON_PMD_HUGE,
+ MF_MSG_UNMAP_FAILED,
+ MF_MSG_DIRTY_SWAPCACHE,
+ MF_MSG_CLEAN_SWAPCACHE,
+ MF_MSG_DIRTY_MLOCKED_LRU,
+ MF_MSG_CLEAN_MLOCKED_LRU,
+ MF_MSG_DIRTY_UNEVICTABLE_LRU,
+ MF_MSG_CLEAN_UNEVICTABLE_LRU,
+ MF_MSG_DIRTY_LRU,
+ MF_MSG_CLEAN_LRU,
+ MF_MSG_TRUNCATED_LRU,
+ MF_MSG_BUDDY,
+ MF_MSG_BUDDY_2ND,
+ MF_MSG_DAX,
+ MF_MSG_UNSPLIT_THP,
+ MF_MSG_UNKNOWN,
+};
+
+/* Action results for various types of pages */
+enum mf_action_result {
+ MF_IGNORED, /* Error: cannot be handled */
+ MF_FAILED, /* Error: handling failed */
+ MF_DELAYED, /* Will be handled later */
+ MF_RECOVERED, /* Successfully recovered */
+};
+
+/* memory failure page types */
+static const struct {
+ int type;
+ const char *page_type;
+} mf_page_type[] = {
+ { MF_MSG_KERNEL, "reserved kernel page" },
+ { MF_MSG_KERNEL_HIGH_ORDER, "high-order kernel page"},
+ { MF_MSG_SLAB, "kernel slab page"},
+ { MF_MSG_DIFFERENT_COMPOUND, "different compound page after locking"},
+ { MF_MSG_POISONED_HUGE, "huge page already hardware poisoned"},
+ { MF_MSG_HUGE, "huge page"},
+ { MF_MSG_FREE_HUGE, "free huge page"},
+ { MF_MSG_NON_PMD_HUGE, "non-pmd-sized huge page"},
+ { MF_MSG_UNMAP_FAILED, "unmapping failed page"},
+ { MF_MSG_DIRTY_SWAPCACHE, "dirty swapcache page"},
+ { MF_MSG_CLEAN_SWAPCACHE, "clean swapcache page"},
+ { MF_MSG_DIRTY_MLOCKED_LRU, "dirty mlocked LRU page"},
+ { MF_MSG_CLEAN_MLOCKED_LRU, "clean mlocked LRU page"},
+ { MF_MSG_DIRTY_UNEVICTABLE_LRU, "dirty unevictable LRU page"},
+ { MF_MSG_CLEAN_UNEVICTABLE_LRU, "clean unevictable LRU page"},
+ { MF_MSG_DIRTY_LRU, "dirty LRU page"},
+ { MF_MSG_CLEAN_LRU, "clean LRU page"},
+ { MF_MSG_TRUNCATED_LRU, "already truncated LRU page"},
+ { MF_MSG_BUDDY, "free buddy page"},
+ { MF_MSG_BUDDY_2ND, "free buddy page (2nd try)"},
+ { MF_MSG_DAX, "dax page"},
+ { MF_MSG_UNSPLIT_THP, "unsplit thp"},
+ { MF_MSG_UNKNOWN, "unknown page"},
+};
+
+/* memory failure action results */
+static const struct {
+ int result;
+ const char *action_result;
+} mf_action_result[] = {
+ { MF_IGNORED, "Ignored" },
+ { MF_FAILED, "Failed" },
+ { MF_DELAYED, "Delayed" },
+ { MF_RECOVERED, "Recovered" },
+};
+
+static const char *get_page_type(int page_type)
+{
+ int i;
+
+ for (i = 0; i < ARRAY_SIZE(mf_page_type); i++)
+ if (mf_page_type[i].type == page_type)
+ return mf_page_type[i].page_type;
+
+ return "unknown page";
+}
+
+static const char *get_action_result(int result)
+{
+ int i;
+
+ for (i = 0; i < ARRAY_SIZE(mf_action_result); i++)
+ if (mf_action_result[i].result == result)
+ return mf_action_result[i].action_result;
+
+ return "unknown";
+}
+
+
+int ras_memory_failure_event_handler(struct trace_seq *s,
+ struct pevent_record *record,
+ struct event_format *event, void *context)
+{
+ unsigned long long val;
+ struct ras_events *ras = context;
+ time_t now;
+ struct tm *tm;
+ struct ras_mf_event ev;
+
+ /*
+ * Newer kernels (3.10-rc1 or upper) provide an uptime clock.
+ * On previous kernels, the way to properly generate an event would
+ * be to inject a fake one, measure its timestamp and diff it against
+ * gettimeofday. We won't do it here. Instead, let's use uptime,
+ * falling-back to the event report's time, if "uptime" clock is
+ * not available (legacy kernels).
+ */
+
+ if (ras->use_uptime)
+ now = record->ts/user_hz + ras->uptime_diff;
+ else
+ now = time(NULL);
+
+ tm = localtime(&now);
+ if (tm)
+ strftime(ev.timestamp, sizeof(ev.timestamp),
+ "%Y-%m-%d %H:%M:%S %z", tm);
+ trace_seq_printf(s, "%s ", ev.timestamp);
+
+ if (pevent_get_field_val(s, event, "pfn", record, &val, 1) < 0)
+ return -1;
+ sprintf(ev.pfn, "0x%llx", val);
+ trace_seq_printf(s, "pfn=0x%llx ", val);
+
+ if (pevent_get_field_val(s, event, "type", record, &val, 1) < 0)
+ return -1;
+ ev.page_type = get_page_type(val);
+ trace_seq_printf(s, "page_type=%s ", ev.page_type);
+
+ if (pevent_get_field_val(s, event, "result", record, &val, 1) < 0)
+ return -1;
+ ev.action_result = get_action_result(val);
+ trace_seq_printf(s, "action_result=%s ", ev.action_result);
+
+ /* Store data into the SQLite DB */
+#ifdef HAVE_SQLITE3
+ ras_store_mf_event(ras, &ev);
+#endif
+
+#ifdef HAVE_ABRT_REPORT
+ /* Report event to ABRT */
+ ras_report_mf_event(ras, &ev);
+#endif
+
+ return 0;
+}
--- /dev/null 1970-01-01 00:00:00.000000000 +0000
+++ b/ras-memory-failure-handler.h 2021-10-14 16:31:36.840657728 -0400
@@ -0,0 +1,25 @@
+/*
+ * Copyright (c) Huawei Technologies Co., Ltd. 2020. 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_MEMORY_FAILURE_HANDLER_H
+#define __RAS_MEMORY_FAILURE_HANDLER_H
+
+#include "ras-events.h"
+#include "libtrace/event-parse.h"
+
+int ras_memory_failure_event_handler(struct trace_seq *s,
+ struct pevent_record *record,
+ struct event_format *event, void *context);
+
+#endif
--- a/ras-record.c 2018-04-25 06:19:03.000000000 -0400
+++ b/ras-record.c 2021-10-14 16:31:36.840657728 -0400
@@ -404,6 +404,55 @@ sqlite3_bind_text(priv->stmt_mce_record,
}
#endif
+/*
+ * Table and functions to handle ras:memory_failure
+ */
+
+#ifdef HAVE_MEMORY_FAILURE
+static const struct db_fields mf_event_fields[] = {
+ { .name="id", .type="INTEGER PRIMARY KEY" },
+ { .name="timestamp", .type="TEXT" },
+ { .name="pfn", .type="TEXT" },
+ { .name="page_type", .type="TEXT" },
+ { .name="action_result", .type="TEXT" },
+};
+
+static const struct db_table_descriptor mf_event_tab = {
+ .name = "memory_failure_event",
+ .fields = mf_event_fields,
+ .num_fields = ARRAY_SIZE(mf_event_fields),
+};
+
+int ras_store_mf_event(struct ras_events *ras, struct ras_mf_event *ev)
+{
+ int rc;
+ struct sqlite3_priv *priv = ras->db_priv;
+
+ if (!priv || !priv->stmt_mf_event)
+ return 0;
+ log(TERM, LOG_INFO, "memory_failure_event store: %p\n", priv->stmt_mf_event);
+
+ sqlite3_bind_text(priv->stmt_mf_event, 1, ev->timestamp, -1, NULL);
+ sqlite3_bind_text(priv->stmt_mf_event, 2, ev->pfn, -1, NULL);
+ sqlite3_bind_text(priv->stmt_mf_event, 3, ev->page_type, -1, NULL);
+ sqlite3_bind_text(priv->stmt_mf_event, 4, ev->action_result, -1, NULL);
+
+ rc = sqlite3_step(priv->stmt_mf_event);
+ if (rc != SQLITE_OK && rc != SQLITE_DONE)
+ log(TERM, LOG_ERR,
+ "Failed to do memory_failure_event step on sqlite: error = %d\n", rc);
+
+ rc = sqlite3_reset(priv->stmt_mf_event);
+ if (rc != SQLITE_OK && rc != SQLITE_DONE)
+ log(TERM, LOG_ERR,
+ "Failed reset memory_failure_event on sqlite: error = %d\n",
+ rc);
+
+ log(TERM, LOG_INFO, "register inserted at db\n");
+
+ return rc;
+}
+#endif
/*
* Generic code
@@ -567,6 +616,13 @@ usleep(10000);
rc = ras_mc_prepare_stmt(priv, &priv->stmt_arm_record,
&arm_event_tab);
#endif
+#ifdef HAVE_MEMORY_FAILURE
+ rc = ras_mc_create_table(priv, &mf_event_tab);
+ if (rc == SQLITE_OK) {
+ rc = ras_mc_prepare_stmt(priv, &priv->stmt_mf_event,
+ &mf_event_tab);
+ }
+#endif
ras->db_priv = priv;
return 0;
--- a/ras-record.h 2018-04-25 06:19:03.000000000 -0400
+++ b/ras-record.h 2021-10-14 16:31:36.840657728 -0400
@@ -75,12 +75,20 @@ struct ras_arm_event {
int32_t psci_state;
};
+struct ras_mf_event {
+ char timestamp[64];
+ char pfn[30];
+ const char *page_type;
+ const char *action_result;
+};
+
struct ras_mc_event;
struct ras_aer_event;
struct ras_extlog_event;
struct ras_non_standard_event;
struct ras_arm_event;
struct mce_event;
+struct ras_mf_event;
#ifdef HAVE_SQLITE3
@@ -104,6 +112,9 @@ struct sqlite3_priv {
#ifdef HAVE_ARM
sqlite3_stmt *stmt_arm_record;
#endif
+#ifdef HAVE_MEMORY_FAILURE
+ sqlite3_stmt *stmt_mf_event;
+#endif
};
int ras_mc_event_opendb(unsigned cpu, struct ras_events *ras);
@@ -113,6 +124,7 @@ int ras_store_mce_record(struct ras_even
int ras_store_extlog_mem_record(struct ras_events *ras, struct ras_extlog_event *ev);
int ras_store_non_standard_record(struct ras_events *ras, struct ras_non_standard_event *ev);
int ras_store_arm_record(struct ras_events *ras, struct ras_arm_event *ev);
+int ras_store_mf_event(struct ras_events *ras, struct ras_mf_event *ev);
#else
static inline int ras_mc_event_opendb(unsigned cpu, struct ras_events *ras) { return 0; };
@@ -122,6 +134,7 @@ static inline int ras_store_mce_record(s
static inline int ras_store_extlog_mem_record(struct ras_events *ras, struct ras_extlog_event *ev) { return 0; };
static inline int ras_store_non_standard_record(struct ras_events *ras, struct ras_non_standard_event *ev) { return 0; };
static inline int ras_store_arm_record(struct ras_events *ras, struct ras_arm_event *ev) { return 0; };
+static inline int ras_store_mf_event(struct ras_events *ras, struct ras_mf_event *ev) { return 0; };
#endif
--- a/ras-report.c 2017-10-14 05:11:34.000000000 -0400
+++ b/ras-report.c 2021-10-14 16:31:36.840657728 -0400
@@ -255,6 +255,28 @@ "midr=0x%lx\n" \
return 0;
}
+static int set_mf_event_backtrace(char *buf, struct ras_mf_event *ev)
+{
+ char bt_buf[MAX_BACKTRACE_SIZE];
+
+ if (!buf || !ev)
+ return -1;
+
+ sprintf(bt_buf, "BACKTRACE=" \
+ "timestamp=%s\n" \
+ "pfn=%s\n" \
+ "page_type=%s\n" \
+ "action_result=%s\n", \
+ ev->timestamp, \
+ ev->pfn, \
+ ev->page_type, \
+ ev->action_result);
+
+ 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;
@@ -283,6 +305,9 @@ memset(buf, 0, MAX_BACKTRACE_SIZE);
case ARM_EVENT:
rc = set_arm_event_backtrace(buf, (struct ras_arm_event *)ev);
break;
+ case MF_EVENT:
+ rc = set_mf_event_backtrace(buf, (struct ras_mf_event *)ev);
+ break;
default:
return -1;
}
@@ -549,3 +574,46 @@ return 0;
return -1;
}
}
+
+int ras_report_mf_event(struct ras_events *ras, struct ras_mf_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 mf_fail;
+
+ rc = commit_report_backtrace(sockfd, MF_EVENT, ev);
+ if (rc < 0)
+ goto mf_fail;
+
+ sprintf(buf, "ANALYZER=%s", "rasdaemon-memory_failure");
+ rc = write(sockfd, buf, strlen(buf) + 1);
+ if (rc < strlen(buf) + 1)
+ goto mf_fail;
+
+ sprintf(buf, "REASON=%s", "memory failure problem");
+ rc = write(sockfd, buf, strlen(buf) + 1);
+ if (rc < strlen(buf) + 1)
+ goto mf_fail;
+
+ done = 1;
+
+mf_fail:
+ if (sockfd > 0)
+ close(sockfd);
+
+ if (done)
+ return 0;
+ else
+ return -1;
+}
--- a/ras-report.h 2017-10-14 05:11:34.000000000 -0400
+++ b/ras-report.h 2021-10-14 16:31:36.840657728 -0400
@@ -34,7 +34,8 @@ enum {
MCE_EVENT,
AER_EVENT,
NON_STANDARD_EVENT,
- ARM_EVENT
+ ARM_EVENT,
+ MF_EVENT,
};
#ifdef HAVE_ABRT_REPORT
@@ -44,6 +45,7 @@ int ras_report_aer_event(struct ras_even
int ras_report_mce_event(struct ras_events *ras, struct mce_event *ev);
int ras_report_non_standard_event(struct ras_events *ras, struct ras_non_standard_event *ev);
int ras_report_arm_event(struct ras_events *ras, struct ras_arm_event *ev);
+int ras_report_mf_event(struct ras_events *ras, struct ras_mf_event *ev);
#else
@@ -52,6 +54,7 @@ static inline int ras_report_aer_event(s
static inline int ras_report_mce_event(struct ras_events *ras, struct mce_event *ev) { return 0; };
static inline int ras_report_non_standard_event(struct ras_events *ras, struct ras_non_standard_event *ev) { return 0; };
static inline int ras_report_arm_event(struct ras_events *ras, struct ras_arm_event *ev) { return 0; };
+static inline int ras_report_mf_event(struct ras_events *ras, struct ras_mf_event *ev) { return 0; };
#endif
--- a/Makefile.am 2018-04-25 06:21:56.000000000 -0400
+++ b/Makefile.am 2021-10-14 16:37:42.423639762 -0400
@@ -41,12 +41,16 @@ endif
if WITH_EXTLOG
rasdaemon_SOURCES += ras-extlog-handler.c
endif
+if WITH_MEMORY_FAILURE
+ rasdaemon_SOURCES += ras-memory-failure-handler.c
+endif
if WITH_ABRT_REPORT
rasdaemon_SOURCES += ras-report.c
endif
if WITH_HISI_NS_DECODE
rasdaemon_SOURCES += non-standard-hisi_hip07.c
endif
+
rasdaemon_LDADD = -lpthread $(SQLITE3_LIBS) libtrace/libtrace.a
include_HEADERS = config.h ras-events.h ras-logger.h ras-mc-handler.h \
--- a/ras-events.c 2021-10-14 16:31:36.730658636 -0400
+++ b/ras-events.c 2021-10-14 16:37:11.043898809 -0400
@@ -33,6 +33,7 @@ * Foundation, Inc., 51 Franklin Street,
#include "ras-arm-handler.h"
#include "ras-mce-handler.h"
#include "ras-extlog-handler.h"
+#include "ras-memory-failure-handler.h"
#include "ras-record.h"
#include "ras-logger.h"
@@ -218,6 +219,10 @@ if (rc < 0) {
rc |= __toggle_ras_mc_event(ras, "ras", "arm_event", enable);
#endif
+#ifdef HAVE_MEMORY_FAILURE
+ rc |= __toggle_ras_mc_event(ras, "ras", "memory_failure_event", enable);
+#endif
+
free_ras:
free(ras);
return rc;
@@ -736,6 +741,16 @@ (void)open("/sys/kernel/debug/ras/daemon
"ras", "aer_event");
#endif
+#ifdef HAVE_MEMORY_FAILURE
+ rc = add_event_handler(ras, pevent, page_size, "ras", "memory_failure_event",
+ ras_memory_failure_event_handler);
+ if (!rc)
+ num_events++;
+ else
+ log(ALL, LOG_ERR, "Can't get traces from %s:%s\n",
+ "ras", "memory_failure_event");
+#endif
+
if (!num_events) {
log(ALL, LOG_INFO,
"Failed to trace all supported RAS events. Aborting.\n");
Reference in New Issue View Git Blame Copy Permalink