ANR基础

转自：http://blog.sina.com.cn/s/blog_c0de2be70102wd1k.html

1、ANR
basic knowledge

ANR分类：

Key Dispatch Timeout

Service Timeout

Broadcast Timeout

关键词：

event log: “am_anr”

trace.txt  -java

     -native

CPU usage  ago

CPU usage  later

iowait

“Early ANR” = The first ANR

2、How
to read Event Log

一些信息没有在logcat
log中而是在event
log中，查看event
log对与解决ANR问题很有帮助。

1) event log定义:
 event.logtags

           (   in code path
/system/core/logcat/event.logtags   )

           (   in phone path /system/etc/event-log-tags
         )

event.logtags:

           30008 am_anr (pid|1|5),(Package
Name|3),(Flags|1|5),(reason|3)

   Source Code ActivityManagerService.java:

           EventLog.writeEvent(EventLogTags.AM_ANR,
app.pid, app.processName...))

2) Log Example:

    I/am_anr  ( 1776):
[3274,com.sonyericsson.home,572997,keyDispatchingTimedOut]

3) Event: am_anr

   system/core/logcat/event.logtags

52004 binder_sample (descriptor|3),(method_num|1|5),(time|1|3),(blocking_package|3),(sample_percent|1|6)

# Application Not Responding

   30008 am_anr (pid|1|5),(Package
Name|3),(Flags|1|5),(reason|3)

          I/am_anr  ( 1776):
[3274,com.sonyericsson.home,572997,keyDispatchingTimedOut]

4) 重要的event
log标志：

db_sample

content_query_sample

content_update_sample

binder sample

dvm_lock_sample

Activity Event Series：

am_proc_start

am_restart_activity

am_on_resume_called

am_finish_activity

am_pause_activity

am_on_paused_called

am_destroy_activity

5) Database related：

            定义：

     52000 db_sample
(db|3),(sql|3),(time|1|3),(blocking_package|3), (sample_percent|1|6)

           # Database operation samples.

   # db: 数据库文件名

   # sql: the executed query (without query args)

   # time: cpu time millis (not wall time),
including lock acquisition

   # blocking_package: if this is on a main
thread, the package name, otherwise ""

   # sample_percent: the percent likelihood this
query was logged (based on 500ms)

sample_percent?  If  time >= 500ms, then
sample_percent show 100.

例子1：

I/db_sample(27898):
[/data/data/com.facebook.katana/databases/fb.db,INSERT INTO
cache(value, timestamp, name) VALUES(?, ?,
?);,99,com.facebook.katana,20]

信息解析：

Database:
/data/data/com.facebook.katana/databases/fb.db

Sql: INSERT INTO cache(value, timestamp, name)
VALUES(?, ?, ?)

Time spent: 99ms

This database operation runs in
com.facebook.katana main thread.

例子2:

I/db_sample(26092):
[/data/data/com.google.android.gm/databases/XX@YY,COMMIT;DELETE FROM
messages WHERE conversation = ? AND synced = ,797,,100]

6) Events :   binder_sample

定义：

52004 binder_sample (descriptor|3),(method_num|1|5),(time|1|3),(blocking_package|3),(sample_percent|1|6)

descriptor：接口描述符（任何一个binder都实现一个接口，IPC调用用的API，任何一个接口都有一个描述符，即一个字符串）

method_num：调用的方法的序列号

time：该方法调用耗用的时间
(从client端发起到server端完成并返回）

blocking_package：从哪个进程发起的调用

例子：

I/binder_sample(  520):
[com.android.internal.telephony.ITelephonyRegistry,7,12518,com.android.phone,100]

含义：从com.android.phone主线程发起了一个ITelephonyRegistry接口的第7个方法的binder调用，耗时12s。

（从后面的源代码可知，系统只记录从主线程发起的binder通讯）

查可知有ITelephonyRegistry.aidl，系统编译后会自动生成ITelephonyRegistry.java文件，里面定义了方法的序号。

out/target/common/obj/JAVA_LIBRARIES/framework_intermediates/src/telephony/java/com/android/internal/telephony/ITelephonyRegistry.java

private static final java.lang.String DESCRIPTOR
= "com.android.internal.telephony.ITelephonyRegistry";

static final int TRANSACTION_listen =
(android.os.IBinder.FIRST_CALL_TRANSACTION + 0);

static final int TRANSACTION_notifyCallState =
(android.os.IBinder.FIRST_CALL_TRANSACTION + 1);

......

static final int
TRANSACTION_notifyDataConnection =
(android.os.IBinder.FIRST_CALL_TRANSACTION + 7);

由此知第7个方法是notifyDataConnection()

7) Events : dvm_lock_sample

定义：

dvm_lock_sample
(process|3),(main|1|5),(thread|3),(time|1|3),(file|3),(line|1|5),(ownerfile|3),(ownerline|1|5),(sample_percent|1|6)

dvm_lock_sample会记录所有线程在java层（即dvm层）获取锁的记录。该log在获取到锁之后输出。

包括：所属进程，是否为主线程，线程名称，获取锁所用的时间，获取锁代码所在的源文件，行号，锁被占有的代码所在的源文件，行号，耗用时间所占比例。

如果ownerfile与file相同，用"-"表示。

Example:

11-25 14:14:24.140 I/dvm_lock_sample(  296):
[system_server,0,Binder Thread #10,12515,

   PackageManagerService.java,1719,-,682,100]

含义：Binder
Thread #10 ，非主线程，属于system_server进程，获取锁耗用了12s,
PackageManagerService.java, line
1719获取锁，PackageManagerService.java，line
682占用了锁。

3、OOM

     Memory leak will cause ANR. Because when
memory is low, normal operations will be blocked for lack of memory
or wait GC to collect memory, this will make normal operations taking
much more time than before. Low memory will cause system into very
bad performance, ANR occurrence rate will increase greatly in this
condition. You will see many out of memory issues are along with ANR
issues.

    If you see frequent ANR occurs in short time
and system is in OOM condition. These ANRs often can be ignored. We
should put effert on fixing memory leak. When the memory leak is
fixed, these ANRs should disappear.

   There are 2 kinds of OOM. One is OOM in one
dvm process. Another is system overall memory is not enough.

1) OOM in one dvm process

One process, one dvm, one GC.

Dvm heap limit in ICS: Change from
dalvik.vm.heapsize to dalvik.vm.heapgrowthlimit (64Mb now in ginger
fuji ICS)

D/dalvikvm(  216): GC_CONCURRENT freed 4443K,
54% free 28546K/61347K, paused 4ms+20ms

 系统为进程216分配了61347K的dvm
heap，有54%
free，分母代表了历史上对Heap的最大需求。

D/dalvikvm( 3142): GC_BEFORE_OOM freed 10K, 2%
free 64421K/65571K, paused 258msE/dalvikvm-heap( 3142): Out of memory
on a 335888-byte allocation.

2) System overall OOM

Kernel oom killer – when there is no enough
memory for allocating.

内核kernel
log:

init invoked oom-killer: gfp_mask=0xd0, order=2,
oom_adj=-16, oom_score_adj=-941

[] (>] (pgd_alloc+0x14/0xe0)

...... unwind_backtrace+0x0/0x12c) from
[] (dump_header.clone.1+0xd8/0x22c)

]
(dump_header.clone.1+0xd8/0x22c) from []
(oom_kill_process.clone.0+0x40/0x33c)

[]
(oom_kill_process.clone.0+0x40/0x33c) from []
(out_of_memory+0x328/0x458)

[] (out_of_memory+0x328/0x458)
from [] (__alloc_pages_nodemask+0x4a0/0x608)

[]
(__alloc_pages_nodemask+0x4a0/0x608) from []
(__get_free_pages+0x10/0x24)

[] (__get_free_pages+0x10/0x24)
from [

[] (sys_fork+0x28/0x2c) from
[] (ret_fast_syscall+0x0/0x30)

……

Out of memory: Kill process 22591
(ckscreen.uxpnxt) score 754 or sacrifice child

Killed process 22591 (ckscreen.uxpnxt)
total-vm:466932kB, anon-rss:23168kB, file-rss:8396kB

Android  Low memory killer – when memory
decrease to some value. So, it doesn’t mean memory is used up.

select 1381 (android.browser), adj 7, size
22194, to kill

select 18759 (ndroid.crashsms), adj 8, size
7129, to kill