File "/var/lib/systemd/coredump/core.tsar.0.xxx.lz4" is not readable: Permission denied

我用进程跟踪器发现常出现 tsar coredump， 一查看 coredumpctl list 顿时吓鸟:

```
...
Tue 2019-07-16 02:18:02 +08 26818 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:19:01 +08 26875 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:20:02 +08 26934 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:21:01 +08 26981 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:22:02 +08 27042 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:23:01 +08 27120 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:24:02 +08 27158 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:25:01 +08 27193 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:26:02 +08 27242 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:27:01 +08 27300 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:28:02 +08 27343 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:29:01 +08 27390 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:30:02 +08 27434 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:31:01 +08 27480 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:32:02 +08 27545 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:33:01 +08 27583 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:34:02 +08 27624 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:35:01 +08 27800 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:36:02 +08 27843 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:37:01 +08 27950 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:38:02 +08 28025 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:39:01 +08 28077 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:40:02 +08 28205 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:41:01 +08 28286 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:42:02 +08 28387 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:43:01 +08 28428 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:44:02 +08 28499 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:45:01 +08 28696 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:46:02 +08 28747 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:47:01 +08 28818 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:48:02 +08 28915 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:49:01 +08 28954 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:50:02 +08 29049 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:51:01 +08 29153 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 02:52:02 +08 29233 0 0 8 error /usr/bin/tsar
Tue 2019-07-16 15:10:22 +08 11781 0 0 8 error /usr/bin/tsar
lines 6735-6770/6770 (END

```

我看见 parent 是 cron， 所以查看 /etc/cron.d/tsar:

```
# cron tsar collect once per minute
MAILTO=""
* * * * * root /usr/bin/tsar --cron > /dev/null 2>&1
```

直接执行的错误信息跟之前的 cron 的 coredump 同样显示没有权限的错误信息：

```
$ sudo /usr/bin/tsar --cron
Floating point exception
$ coredumpctl dump 11781 --output tsar.11781.alamak
PID: 11781 (tsar)
UID: 0 (root)
GID: 0 (root)
Signal: 8 (FPE)
Timestamp: Tue 2019-07-16 15:10:21 +08 (58s ago)
Command Line: /usr/bin/tsar --cron
Executable: /usr/bin/tsar
Control Group: /user.slice/user-1000.slice/session-2.scope
Unit: session-2.scope
Slice: user-1000.slice
Session: 2
Owner UID: 1000 (xiaobai)
...
Storage: /var/lib/systemd/coredump/core.tsar.0.4537918427684a69bcb0013d8b9d7172.11781.1563261021000000.lz4 (inaccessible)
Message: Process 11781 (tsar) of user 0 dumped core.

Stack trace of thread 11781:
#0 0x00007f1a51539d7f store_single_partition (mod_partition.so)
#1 0x00007f1a51539ea6 read_partition_stat (mod_partition.so)
#2 0x00005625a1125417 collect_record (tsar)
#3 0x00005625a1125b57 running_cron (tsar)
#4 0x00005625a1123250 main (tsar)
#5 0x00007f1a53180b97 __libc_start_main (libc.so.6)
#6 0x00005625a11232ea _start (tsar)
File "/var/lib/systemd/coredump/core.tsar.0.4537918427684a69bcb0013d8b9d7172.11781.1563261021000000.lz4" is not readable: Permission denied
$ ls -l "/var/lib/systemd/coredump/core.tsar.0.4537918427684a69bcb0013d8b9d7172.11781.1563261021000000.lz4"
-rw-r----- 1 root root 588594 Jul 16 15:10 /var/lib/systemd/coredump/core.tsar.0.4537918427684a69bcb0013d8b9d7172.11781.1563261021000000.lz4
$
```

我已暂时把 /etc/cron.d/tsar 移除。 我的平台是 Ubuntu 18.04.2 LTS

Start with the /etc/cron.d/tsar entry and reproduce /usr/bin/tsar --cron on Ubuntu 18.04. Trace the reported path through running_cron, collect_record, read_partition_stat, and store_single_partition in mod_partition.so. Done means the cron invocation no longer raises a floating-point exception or repeatedly creates tsar coredumps.

Written by the indexing model from the issue text.