This patch refines the process of log space check.
Check a folder's size need to traverse all nodes of files and subdirs,
in the case of extreme memory strain, accessing these log file nodes
frequently will increase the IO pressure on the eMMC.
Previously, all log files will be accessed and counted in each space
check, this patch refines it by only calculating the increment of new logs.
Tracked-On: #1024
Signed-off-by: Liu, Xinwu <xinwu.liu@intel.com>
Reviewed-by: Liu, Xiaojing <xiaojing.liu@intel.com>
Acked-by: Chen, Gang <gang.c.chen@intel.com>
This patch unifies the event analyzing process in the
crashlog_sender functions:
a. analyze android events in android_events.c.
b. unify the analyzing output form of vm events and crash events.
c. concentrate the event analyzing process to the same function.
Tracked-On: #1024
Signed-off-by: Liu, Xinwu <xinwu.liu@intel.com>
Reviewed-by: Liu, Xiaojing <xiaojing.liu@intel.com>
Acked-by: Chen, Gang <gang.c.chen@intel.com>
Push the recorded vm events to event_queue, and let senders do the
collecting job in the event_handler thread.
Tracked-On: #1024
Signed-off-by: Liu, Xinwu <xinwu.liu@intel.com>
Reviewed-by: Liu, Xiaojing <xiaojing.liu@intel.com>
Acked-by: Chen, Gang <gang.c.chen@intel.com>
Function sender_id need to traverse all senders and then find the
matched id. This patch adds property "id" for each configuration object
and removes function sender_id.
Tracked-On: #1024
Signed-off-by: Liu, Xinwu <xinwu.liu@intel.com>
Reviewed-by: Liu, Xiaojing <xiaojing.liu@intel.com>
Acked-by: Chen, Gang <gang.c.chen@intel.com>
Crashlog would drop all events after the storaged log size exceeds the
configured parameter "foldersize" previously, that basically means all
activities are stopped.
This patch brings other activities back (e.g. the records in the file
"history_event" will be updated continually), and restricts the "foldersize" by
stopping log collection.
Tracked-On: #1024
Signed-off-by: Liu, Xinwu <xinwu.liu@intel.com>
Reviewed-by: Zhi Jin <zhi.jin@intel.com>
Acked-by: Chen, Gang <gang.c.chen@intel.com>
Dmesg collects too much log when kernel meets log storm.
This patch reads kernel log from "/dev/kmsg" intead of dmesg and limits
the generated log size.
Tracked-On: #1024
Signed-off-by: Liu, Xinwu <xinwu.liu@intel.com>
Reviewed-by: Zhi Jin <zhi.jin@intel.com>
Acked-by: Chen, Gang <gang.c.chen@intel.com>
1. remove watching TOMBSTONE in AaaG.
2. add a configurable parameter "foldersize" in MB, sender will drop
all events when the storaged log size exceeds this parameter.
Tracked-On:#1024
Signed-off-by: Liu, Xinwu <xinwu.liu@intel.com>
Reviewed-by: Chen Gang <gang.c.chen@intel.com>
Acked-by: Zhang Di <di.zhang@intel.com>
Acrnprobe is using SHA to generate ids for events. These ids are only used
to index events, not for cryptographic purpose.
This patch unify the generating algorithm of short and long ids to
SHA256.
Tracked-On: #1024
Signed-off-by: Liu, Xinwu <xinwu.liu@intel.com>
Reviewed-by: Zhi Jin <zhi.jin@intel.com>
Acked-by: Chen Gang <gang.c.chen@intel.com>
1. Reload the trigger file(s) of subcrash while doing crash reclassify.
2. Support simple format for trigger file.
Signed-off-by: Liu, Xinwu <xinwu.liu@intel.com>
Acked-by: Chen Gang <gang.c.chen@intel.com>
In the original design, acrnprobe marked all handled VMs'events as "synced"
in file vmrecordid(this patch changes the name to VM_eventsID.log).
Currently, the Android log events are not logged if the first attempt at
reading collecting them from the VM fails. This patch changes the logic
so that the acrn-crashlog tool will retry continuously.
This patch defines different tags for handled VMs'events, and only marks
VMs'events "synced" after it returns successfully.
Signed-off-by: Liu, Xinwu <xinwu.liu@intel.com>
Reviewed-by: xiaojin2 <xiaojing.liu@intel.com>
Reviewed-by: Jin Zhi <zhi.jin@intel.com>
Acked-by: Geoffroy Van Cutsem <geoffroy.vancutsem@intel.com>
Acked-by: Chen gang <gang.c.chen@intel.com>
This patch is to fix potential issues, which are reported by static
analysis tool, for acrnprobe.
Changes:
1. Check the return value of sender_id() and get_sender_by_name(),
since it could be -1 or NULL.
2. Remove the parameter len from create_event, take parameter path
as a NULL-terminated string as default.
3. Modify for_each_* functions to avoid overflow.
Signed-off-by: Liu Xinwu <xinwu.liu@intel.com>
Reviewed-by: Zhi Jin <zhi.jin@intel.com>
Acked-by: Chen Gang <gang.c.chen@intel.com>
This patch is to fix potential issues, which are reported by static
analysis tool, for load_conf module of acrnprobe.
Changes:
1. Check the range of id while loading configuration, to avoid
memory corruption.
2. Use strncpy instead of strcpy to avoid buf overflow.
Signed-off-by: Liu Xinwu <xinwu.liu@intel.com>
Reviewed-by: Zhi Jin <zhi.jin@intel.com>
Acked-by: Chen Gang <gang.c.chen@intel.com>
This patch implements the main thread of acrnprobe.
As a log collection mechanism to record critical events on the
platform, acrnprobe provides the following features:
1. detect event.
2. analyze event and determine the event type.
3. collect information for the detected events.
4. archive these information as logs, and generate records.
Signed-off-by: Liu Xinwu <xinwu.liu@intel.com>
Reviewed-by: Zhang Yanmin <yanmin.zhang@intel.com>
Reviewed-by: Liu Chuansheng <chuansheng.liu@intel.com>
Reviewed-by: Zhao Yakui <yakui.zhao@intel.com>
Reviewed-by: Geoffroy Van Cutsem <geoffroy.vancutsem@intel.com>
Acked-by: Eddie Dong <Eddie.dong@intel.com>
