[V3,2/2] rasdaemon: add support for ARM events

[Mauro Carvalho Chehab <mchehab@osg.samsung.com>]

Em Mon, 12 Jun 2017 16:16:05 -0600
Tyler Baicar <tbaicar@codeaurora.org> escreveu:

> Add support to handle the ARM kernel trace events
> which cover RAS ARM processor errors.

I'm getting this gcc 7.1 warning with this patch applied:

ras-record.c:227:41: warning: ‘arm_event_tab’ defined but not used [-Wunused-const-variable=]
 static const struct db_table_descriptor arm_event_tab = {
                                         ^~~~~~~~~~~~~


> 
> Signed-off-by: Tyler Baicar <tbaicar@codeaurora.org>
> ---
>  Makefile.am       |  3 ++
>  configure.ac      |  9 ++++++
>  ras-arm-handler.c | 90 +++++++++++++++++++++++++++++++++++++++++++++++++++++++
>  ras-arm-handler.h | 24 +++++++++++++++
>  ras-events.c      | 15 ++++++++++
>  ras-record.c      | 52 ++++++++++++++++++++++++++++++++
>  ras-record.h      | 16 ++++++++++
>  ras-report.c      | 75 ++++++++++++++++++++++++++++++++++++++++++++++
>  ras-report.h      |  5 +++-
>  9 files changed, 288 insertions(+), 1 deletion(-)
>  create mode 100644 ras-arm-handler.c
>  create mode 100644 ras-arm-handler.h
> 
> diff --git a/Makefile.am b/Makefile.am
> index 1bc25e4..83868e2 100644
> --- a/Makefile.am
> +++ b/Makefile.am
> @@ -27,6 +27,9 @@ endif
>  if WITH_NON_STANDARD
>     rasdaemon_SOURCES += ras-non-standard-handler.c
>  endif
> +if WITH_ARM
> +   rasdaemon_SOURCES += ras-arm-handler.c
> +endif
>  if WITH_MCE
>     rasdaemon_SOURCES += ras-mce-handler.c mce-intel.c mce-amd-k8.c \
>  			mce-intel-p4-p6.c mce-intel-nehalem.c \
> diff --git a/configure.ac b/configure.ac
> index 2adb798..648a3ac 100644
> --- a/configure.ac
> +++ b/configure.ac
> @@ -53,6 +53,15 @@ AS_IF([test "x$enable_non_standard" = "xyes"], [
>  ])
>  AM_CONDITIONAL([WITH_NON_STANDARD], [test x$enable_non_standard = xyes])
>  
> +AC_ARG_ENABLE([arm],
> +    AS_HELP_STRING([--enable-arm], [enable ARM events (currently experimental)]))
> +
> +AS_IF([test "x$enable_arm" = "xyes"], [
> +  AC_DEFINE(HAVE_ARM,1,"have ARM events collect")
> +  AC_SUBST([WITH_ARM])
> +])
> +AM_CONDITIONAL([WITH_ARM], [test x$enable_arm = xyes])
> +
>  AC_ARG_ENABLE([mce],
>      AS_HELP_STRING([--enable-mce], [enable MCE events (currently experimental)]))
>  
> diff --git a/ras-arm-handler.c b/ras-arm-handler.c
> new file mode 100644
> index 0000000..a76470d
> --- /dev/null
> +++ b/ras-arm-handler.c
> @@ -0,0 +1,90 @@
> +/*
> + * Copyright (c) 2016, The Linux Foundation. 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 version 2 and
> + * only version 2 as published by the Free Software Foundation.
> +
> + * 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 "libtrace/kbuffer.h"
> +#include "ras-arm-handler.h"
> +#include "ras-record.h"
> +#include "ras-logger.h"
> +#include "ras-report.h"
> +
> +int ras_arm_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_arm_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\n", ev.timestamp);
> +
> +	if (pevent_get_field_val(s, event, "affinity", record, &val, 1) < 0)
> +		return -1;
> +	ev.affinity = val;
> +	trace_seq_printf(s, " affinity: %d", ev.affinity);
> +
> +	if (pevent_get_field_val(s, event, "mpidr", record, &val, 1) < 0)
> +		return -1;
> +	ev.mpidr = val;
> +	trace_seq_printf(s, "\n MPIDR: 0x%llx", (unsigned long long)ev.mpidr);
> +
> +	if (pevent_get_field_val(s, event, "midr", record, &val, 1) < 0)
> +		return -1;
> +	ev.midr = val;
> +	trace_seq_printf(s, "\n MIDR: 0x%llx", (unsigned long long)ev.midr);
> +
> +	if (pevent_get_field_val(s, event, "running_state", record, &val, 1) < 0)
> +		return -1;
> +	ev.running_state = val;
> +	trace_seq_printf(s, "\n running_state: %d", ev.running_state);
> +
> +	if (pevent_get_field_val(s, event, "psci_state", record, &val, 1) < 0)
> +		return -1;
> +	ev.psci_state = val;
> +	trace_seq_printf(s, "\n psci_state: %d", ev.psci_state);
> +
> +	/* Insert data into the SGBD */
> +#ifdef HAVE_SQLITE3
> +	ras_store_arm_record(ras, &ev);
> +#endif
> +
> +#ifdef HAVE_ABRT_REPORT
> +	/* Report event to ABRT */
> +	ras_report_arm_event(ras, &ev);
> +#endif
> +
> +	return 0;
> +}
> diff --git a/ras-arm-handler.h b/ras-arm-handler.h
> new file mode 100644
> index 0000000..eae10ec
> --- /dev/null
> +++ b/ras-arm-handler.h
> @@ -0,0 +1,24 @@
> +/*
> + * Copyright (c) 2016, The Linux Foundation. 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 version 2 and
> + * only version 2 as published by the Free Software Foundation.
> +
> + * 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_ARM_HANDLER_H
> +#define __RAS_ARM_HANDLER_H
> +
> +#include "ras-events.h"
> +#include "libtrace/event-parse.h"
> +
> +int ras_arm_event_handler(struct trace_seq *s,
> +			 struct pevent_record *record,
> +			 struct event_format *event, void *context);
> +
> +#endif
> diff --git a/ras-events.c b/ras-events.c
> index 96aa6f1..812d712 100644
> --- a/ras-events.c
> +++ b/ras-events.c
> @@ -30,6 +30,7 @@
>  #include "ras-mc-handler.h"
>  #include "ras-aer-handler.h"
>  #include "ras-non-standard-handler.h"
> +#include "ras-arm-handler.h"
>  #include "ras-mce-handler.h"
>  #include "ras-extlog-handler.h"
>  #include "ras-record.h"
> @@ -213,6 +214,10 @@ int toggle_ras_mc_event(int enable)
>  	rc |= __toggle_ras_mc_event(ras, "ras", "non_standard_event", enable);
>  #endif
>  
> +#ifdef HAVE_ARM
> +	rc |= __toggle_ras_mc_event(ras, "ras", "arm_event", enable);
> +#endif
> +
>  free_ras:
>  	free(ras);
>  	return rc;
> @@ -691,6 +696,16 @@ int handle_ras_events(int record_events)
>                      "ras", "non_standard_event");
>  #endif
>  
> +#ifdef HAVE_ARM
> +        rc = add_event_handler(ras, pevent, page_size, "ras", "arm_event",
> +                               ras_arm_event_handler);
> +        if (!rc)
> +                num_events++;
> +        else
> +                log(ALL, LOG_ERR, "Can't get traces from %s:%s\n",
> +                    "ras", "arm_event");
> +#endif
> +
>  	cpus = get_num_cpus(ras);
>  
>  #ifdef HAVE_MCE
> diff --git a/ras-record.c b/ras-record.c
> index 357ab61..5dfa00f 100644
> --- a/ras-record.c
> +++ b/ras-record.c
> @@ -209,6 +209,58 @@ int ras_store_non_standard_record(struct ras_events *ras, struct ras_non_standar
>  }
>  #endif
>  
> +/*
> + * Table and functions to handle ras:arm
> + */
> +
> +#ifdef HAVE_ARM
> +static const struct db_fields arm_event_fields[] = {
> +		{ .name="id",			.type="INTEGER PRIMARY KEY" },
> +		{ .name="timestamp",		.type="TEXT" },
> +		{ .name="error_count",		.type="INTEGER" },
> +		{ .name="affinity",		.type="INTEGER" },
> +		{ .name="mpidr",		.type="INTEGER" },
> +		{ .name="running_state",	.type="INTEGER" },
> +		{ .name="psci_state",		.type="INTEGER" },
> +};
> +
> +static const struct db_table_descriptor arm_event_tab = {
> +	.name = "arm_event",
> +	.fields = arm_event_fields,
> +	.num_fields = ARRAY_SIZE(arm_event_fields),
> +};
> +
> +int ras_store_arm_record(struct ras_events *ras, struct ras_arm_event *ev)
> +{
> +	int rc;
> +	struct sqlite3_priv *priv = ras->db_priv;
> +
> +	if (!priv || !priv->stmt_arm_record)
> +		return 0;
> +	log(TERM, LOG_INFO, "arm_event store: %p\n", priv->stmt_arm_record);
> +
> +	sqlite3_bind_text (priv->stmt_arm_record,  1,  ev->timestamp, -1, NULL);
> +	sqlite3_bind_int  (priv->stmt_arm_record,  2,  ev->error_count);
> +	sqlite3_bind_int  (priv->stmt_arm_record,  3,  ev->affinity);
> +	sqlite3_bind_int  (priv->stmt_arm_record,  4,  ev->mpidr);
> +	sqlite3_bind_int  (priv->stmt_arm_record,  5,  ev->running_state);
> +	sqlite3_bind_int  (priv->stmt_arm_record,  6,  ev->psci_state);
> +
> +	rc = sqlite3_step(priv->stmt_arm_record);
> +	if (rc != SQLITE_OK && rc != SQLITE_DONE)
> +		log(TERM, LOG_ERR,
> +		    "Failed to do arm_event step on sqlite: error = %d\n", rc);
> +	rc = sqlite3_reset(priv->stmt_arm_record);
> +	if (rc != SQLITE_OK && rc != SQLITE_DONE)
> +		log(TERM, LOG_ERR,
> +		    "Failed reset arm_event on sqlite: error = %d\n",
> +		    rc);
> +	log(TERM, LOG_INFO, "register inserted at db\n");
> +
> +	return rc;
> +}
> +#endif
> +
>  #ifdef HAVE_EXTLOG
>  static const struct db_fields extlog_event_fields[] = {
>  		{ .name="id",			.type="INTEGER PRIMARY KEY" },
> diff --git a/ras-record.h b/ras-record.h
> index 473ae40..12c2218 100644
> --- a/ras-record.h
> +++ b/ras-record.h
> @@ -65,10 +65,21 @@ struct ras_non_standard_event {
>  	uint32_t length;
>  };
>  
> +struct ras_arm_event {
> +	char timestamp[64];
> +	int32_t error_count;
> +	int8_t affinity;
> +	int64_t mpidr;
> +	int64_t midr;
> +	int32_t running_state;
> +	int32_t psci_state;
> +};
> +
>  struct ras_mc_event;
>  struct ras_aer_event;
>  struct ras_extlog_event;
>  struct ras_non_standard_event;
> +struct ras_arm_event;
>  struct mce_event;
>  
>  #ifdef HAVE_SQLITE3
> @@ -90,6 +101,9 @@ struct sqlite3_priv {
>  #ifdef HAVE_NON_STANDARD
>  	sqlite3_stmt	*stmt_non_standard_record;
>  #endif
> +#ifdef HAVE_ARM
> +	sqlite3_stmt	*stmt_arm_record;
> +#endif
>  };
>  
>  int ras_mc_event_opendb(unsigned cpu, struct ras_events *ras);
> @@ -98,6 +112,7 @@ int ras_store_aer_event(struct ras_events *ras, struct ras_aer_event *ev);
>  int ras_store_mce_record(struct ras_events *ras, struct mce_event *ev);
>  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);
>  
>  #else
>  static inline int ras_mc_event_opendb(unsigned cpu, struct ras_events *ras) { return 0; };
> @@ -106,6 +121,7 @@ static inline int ras_store_aer_event(struct ras_events *ras, struct ras_aer_eve
>  static inline int ras_store_mce_record(struct ras_events *ras, struct mce_event *ev) { return 0; };
>  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; };
>  
>  #endif
>  
> diff --git a/ras-report.c b/ras-report.c
> index 1eb9f79..d4beee0 100644
> --- a/ras-report.c
> +++ b/ras-report.c
> @@ -228,6 +228,33 @@ static int set_non_standard_event_backtrace(char *buf, struct ras_non_standard_e
>  	return 0;
>  }
>  
> +static int set_arm_event_backtrace(char *buf, struct ras_arm_event *ev){
> +	char bt_buf[MAX_BACKTRACE_SIZE];
> +
> +	if(!buf || !ev)
> +		return -1;
> +
> +	sprintf(bt_buf, "BACKTRACE="    \
> +						"timestamp=%s\n"	\
> +						"error_count=%d\n"	\
> +						"affinity=%d\n"	\
> +						"mpidr=0x%lx\n"	\
> +						"midr=0x%lx\n"	\
> +						"running_state=%d\n"	\
> +						"psci_state=%d\n",	\
> +						ev->timestamp,	\
> +						ev->error_count,	\
> +						ev->affinity,	\
> +						ev->mpidr,	\
> +						ev->midr,	\
> +						ev->running_state,	\
> +						ev->psci_state);
> +
> +	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;
> @@ -253,6 +280,9 @@ static int commit_report_backtrace(int sockfd, int type, void *ev){
>  	case NON_STANDARD_EVENT:
>  		rc = set_non_standard_event_backtrace(buf, (struct ras_non_standard_event *)ev);
>  		break;
> +	case ARM_EVENT:
> +		rc = set_arm_event_backtrace(buf, (struct ras_arm_event *)ev);
> +		break;
>  	default:
>  		return -1;
>  	}
> @@ -425,6 +455,51 @@ non_standard_fail:
>  	return rc;
>  }
>  
> +int ras_report_arm_event(struct ras_events *ras, struct ras_arm_event *ev){
> +	char buf[MAX_MESSAGE_SIZE];
> +	int sockfd = 0;
> +	int rc = -1;
> +
> +	memset(buf, 0, sizeof(buf));
> +
> +	sockfd = setup_report_socket();
> +	if(sockfd < 0){
> +		return rc;
> +	}
> +
> +	rc = commit_report_basic(sockfd);
> +	if(rc < 0){
> +		goto arm_fail;
> +	}
> +
> +	rc = commit_report_backtrace(sockfd, ARM_EVENT, ev);
> +	if(rc < 0){
> +		goto arm_fail;
> +	}
> +
> +	sprintf(buf, "ANALYZER=%s", "rasdaemon-arm");
> +	rc = write(sockfd, buf, strlen(buf) + 1);
> +	if(rc < strlen(buf) + 1){
> +		goto arm_fail;
> +	}
> +
> +	sprintf(buf, "REASON=%s", "ARM CPU report problem");
> +	rc = write(sockfd, buf, strlen(buf) + 1);
> +	if(rc < strlen(buf) + 1){
> +		goto arm_fail;
> +	}
> +
> +	rc = 0;
> +
> +arm_fail:
> +
> +	if(sockfd > 0){
> +		close(sockfd);
> +	}
> +
> +	return rc;
> +}
> +
>  int ras_report_mce_event(struct ras_events *ras, struct mce_event *ev){
>  	char buf[MAX_MESSAGE_SIZE];
>  	int sockfd = 0;
> diff --git a/ras-report.h b/ras-report.h
> index c2fcf42..6c466f5 100644
> --- a/ras-report.h
> +++ b/ras-report.h
> @@ -33,7 +33,8 @@ enum {
>  	MC_EVENT,
>  	MCE_EVENT,
>  	AER_EVENT,
> -	NON_STANDARD_EVENT
> +	NON_STANDARD_EVENT,
> +	ARM_EVENT
>  };
>  
>  #ifdef HAVE_ABRT_REPORT
> @@ -42,6 +43,7 @@ int ras_report_mc_event(struct ras_events *ras, struct ras_mc_event *ev);
>  int ras_report_aer_event(struct ras_events *ras, struct ras_aer_event *ev);
>  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);
>  
>  #else
>  
> @@ -49,6 +51,7 @@ static inline int ras_report_mc_event(struct ras_events *ras, struct ras_mc_even
>  static inline int ras_report_aer_event(struct ras_events *ras, struct ras_aer_event *ev) { return 0; };
>  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; };
>  
>  #endif
>