按照SystemTap Beginners Guide的Installation and Setup部分安装了SystemTap,没想到竟然还有点曲折,在这里纪录一下。

 

  • Linux发行版本:CentOS Linux release 7.4.1708 (Core)
  • 内核版本:3.10.0-693.2.2.el7.x86_64
  • uname -a: Linux hostname 3.10.0-693.2.2.el7.x86_64 #1 SMP Tue Sep 12 22:26:13 UTC 2017 x86_64 x86_64 x86_64 GNU/Linux

 

先安装如下两个RPM包:

  • systemtap
  • systemtap-runtime

以root运行如下命令:

  1. # yum install systemtap systemtap-runtime

 

在运行SystemTap之间,还需要装必要的内核信息包。在现代系统上,可以运行如下stap-prep来安装这些包,如下:

  1. # stap-prep
  2. Need to install the following packages:
  3. kernel-devel-3.10.0-693.2.2.el7.x86_64
  4. kernel-debuginfo-3.10.0-693.2.2.el7.x86_64
  5. Loaded plugins: fastestmirror
  6. Loading mirror speeds from cached hostfile
  7. No package kernel-devel-3.10.0-693.2.2.el7.x86_64 available.
  8. No package kernel-debuginfo-3.10.0-693.2.2.el7.x86_64 available.
  9. Error: Nothing to do
  10. Loaded plugins: fastestmirror
  11. Loading mirror speeds from cached hostfile
  12. Could not find debuginfo for main pkg: kernel-3.10.0-693.2.2.el7.x86_64
  13. No debuginfo packages available to install
  14. package kernel-devel-3.10.0-693.2.2.el7.x86_64 is not installed
  15. package kernel-debuginfo-3.10.0-693.2.2.el7.x86_64 is not installed
  16. problem installing rpm(s) kernel-devel-3.10.0-693.2.2.el7.x86_64 kernel-debuginfo-3.10.0-693.2.2.el7.x86_64

运行stap-prep的时候,它探测出还要安装kernel-devel-3.10.0-693.2.2.el7.x86_64包和kernel-debuginfo-3.10.0-693.2.2.el7.x86_64包 (实际上还有kernel-debuginfo-common包),但是自动安装失败。我们可以按照如下方法手动安装。

 

SystemTap需要安装内核内核符号文件来probe内核。必要的内核信息包含在如下三个包中:

  • kernel-debuginfo
  • kernel-debuginfo-common
  • kernel-devel

一定要安装与当前内核版本一致的包。当前环境的内核版本是3.10.0-693.2.2.el7.x86_64,所以需要安装的包为:

  • kernel-debuginfo-3.10.0-693.2.2.el7.x86_64
  • kernel-debuginfo-common-3.10.0-693.2.2.el7.x86_64
  • kernel-devel-3.10.0-693.2.2.el7.x86_64

注意:其实在上一步运行stap-prep时,已经把需要的包的名称及其内核精准地打印在屏幕上了。

 

接下来安装这三个包,注意不要直接yum install kernel-debuginfo kernel-debuginfo-common kernel-devel, 即使能找到相应的包,也是安装的最新版本,不会自动匹配当前版本。所以我们下载RPM包,再用rpm命令依次安装。

 

对于CentOS来说,内核符号文件一版在http://debuginfo.centos.org上有各个版本非常完整的包,但是一般从境内下载都比较慢,特别是kernel-debuginfo,比较大下载可能非常慢。所以在debuginfo.centos.org上下了kernel-debuginfo-common包,另外领个包在Google上搜了一把,分别找了两个镜像。下了之后才发现这个地方有坑,这个坑在后面展开讲。

  1. wget https://ftp.sjtu.edu.cn/scientific/7/archive/debuginfo/kernel-debuginfo-3.10.0-693.2.2.el7.x86_64.rpm
  2. wget http://debuginfo.centos.org/7/x86_64/kernel-debuginfo-common-x86_64-3.10.0-693.2.2.el7.x86_64.rpm
  3. wget ftp://mirror.switch.ch/pool/4/mirror/scientificlinux/7.0/x86_64/updates/security/kernel-devel-3.10.0-693.2.2.el7.x86_64.rpm

下载之后,直接用rpm命令安装就好:

  1. # rpm -ivh kernel-debuginfo-common-x86_64-3.10.0-693.2.2.el7.x86_64.rpm
  2. # rpm -ivh kernel-debuginfo-3.10.0-693.2.2.el7.x86_64.rpm
  3. # rpm -ivh kernel-devel-3.10.0-693.2.2.el7.x86_64.rpm

至此安装步骤完毕,下面来测试SystemTap能不能正常运行。

 

为了测试stap是否能正常运行,用如下简单命令打印:

  1. # stap -e 'probe begin{printf("Hello, World"); exit();}'

运行失败…结果如下:

  1. # stap -e 'probe begin{printf("Hello, World"); exit();}'
  2. ERROR: module version mismatch (#1 SMP Tue Sep 12 10:10:26 CDT 2017 vs #1 SMP Tue Sep 12 22:26:13 UTC 2017), release 3.10.0-693.2.2.el7.x86_64
  3. WARNING: /usr/bin/staprun exited with status: 1
  4. Pass 5: run failed. [man error::pass5]

错误信息是:”ERROR: module version mismatch (#1 SMP Tue Sep 1210:10:26 CDT 2017 vs #1 SMP Tue Sep 1222:26:13 UTC 2017)“。

 

stap运行的时候加上-v参数,打印更多信息看看还有没有更多线索:

  1. # stap -e 'probe begin{printf("Hello, World"); exit();}' -v
  2. Pass 1: parsed user script and 470 library scripts using 228224virt/41280res/3348shr/38020data kb, in 330usr/20sys/346real ms.
  3. Pass 2: analyzed script: 1 probe, 1 function, 0 embeds, 0 globals using 229148virt/42332res/3536shr/38944data kb, in 0usr/0sys/6real ms.
  4. Pass 3: using cached /root/.systemtap/cache/0b/stap_0bc9e27aef7a1de50ea41889a27fc524_1010.c
  5. Pass 4: using cached /root/.systemtap/cache/0b/stap_0bc9e27aef7a1de50ea41889a27fc524_1010.ko
  6. Pass 5: starting run.
  7. ERROR: module version mismatch (#1 SMP Tue Sep 12 10:10:26 CDT 2017 vs #1 SMP Tue Sep 12 22:26:13 UTC 2017), release 3.10.0-693.2.2.el7.x86_64
  8. WARNING: /usr/bin/staprun exited with status: 1
  9. Pass 5: run completed in 0usr/10sys/38real ms.
  10. Pass 5: run failed. [man error::pass5]

查看c文件,vi /root/.systemtap/cache/0b/stap_0bc9e27aef7a1de50ea41889a27fc524_1010.c,搜错误信息”module version mismatch“,能搜到报错发生在下面的第13行,至于UTS_RELEASE和UTS_VERSION是在哪里设置的,直接Google一把。

  1. 1 #ifndef STP_NO_VERREL_CHECK
  2. 2 const char* release = UTS_RELEASE;
  3. 3 #ifdef STAPCONF_GENERATED_COMPILE
  4. 4 const char* version = UTS_VERSION;
  5. 5 #endif
  6. 6 might_sleep();
  7. 7 if (strcmp (release, "3.10.0-693.2.2.el7.x86_64")) {
  8. 8 _stp_error ("module release mismatch (%s vs %s)", release, "3.10.0-693.2.2.el7.x86_64");
  9. 9 rc = -EINVAL;
  10. 10 }
  11. 11 #ifdef STAPCONF_GENERATED_COMPILE
  12. 12 if (strcmp (utsname()->version, version)) {
  13. 13 _stp_error ("module version mismatch (%s vs %s), release %s", version, utsname()->version, release);
  14. 14 rc = -EINVAL;
  15. 15 }
  16. 16 #endif
  17. 17 #endif

 

有两篇文章里面提到了同样的坑,文章连接在底部的参考中。在kernel-devel包的所以文件中搜以下变量UTS_VERSION,

  1. # rpm -ql kernel-devel | xargs grep UTS_VERSION
  2. /usr/src/kernels/3.10.0-693.2.2.el7.x86_64/include/generated/compile.h:#define UTS_VERSION "#1 SMP Tue Sep 12 10:10:26 CDT 2017"

可以看到在compile.h中有#define UTS_VERSION “#1 SMP Tue Sep 12 10:10:26 CDT 2017”. 这个是不是很熟悉… 对比下上面运行stap的报错信息, module mismatch的时间就是这个。文件compile.h是自动生成的,可能和当时编译时的时间相关。但是stap要求这个也和当前系统uname -a里面的时间完全一直,如果下个CentOS原生的kernel-devel应该就没这个问题。

解决问题的另一个简单方法就是直接修改这个compile.h文件,原来的文件如下:

  1. # cat /usr/src/kernels/3.10.0-693.2.2.el7.x86_64/include/generated/compile.h
  2. /* This file is auto generated, version 1 */
  3. /* SMP */
  4. #define UTS_MACHINE "x86_64"
  5. #define UTS_VERSION "#1 SMP Tue Sep 12 10:10:26 CDT 2017"
  6. #define LINUX_COMPILE_BY "mockbuild"
  7. #define LINUX_COMPILE_HOST "sl7-uefisign.fnal.gov"
  8. #define LINUX_COMPILER "gcc version 4.8.5 20150623 (Red Hat 4.8.5-16) (GCC) "

修改define UTS_VERSION那一行,如下:

#define UTS_VERSION “#1 SMP Tue Sep 12 10:10:26 CDT 2017” -> #define UTS_VERSION “#1 SMP Tue Sep 12 22:26:13 UTC 2017”

 

再次运行stap:

  1. # stap -e 'probe begin{printf("Hello, World"); exit();}' -v
  2. Pass 1: parsed user script and 470 library scripts using 228220virt/41276res/3348shr/38016data kb, in 350usr/10sys/355real ms.
  3. Pass 2: analyzed script: 1 probe, 1 function, 0 embeds, 0 globals using 229144virt/42328res/3536shr/38940data kb, in 0usr/0sys/6real ms.
  4. Pass 3: using cached /root/.systemtap/cache/0b/stap_0bc9e27aef7a1de50ea41889a27fc524_1010.c
  5. Pass 4: using cached /root/.systemtap/cache/0b/stap_0bc9e27aef7a1de50ea41889a27fc524_1010.ko
  6. Pass 5: starting run.
  7. ERROR: module version mismatch (#1 SMP Tue Sep 12 10:10:26 CDT 2017 vs #1 SMP Tue Sep 12 22:26:13 UTC 2017), release 3.10.0-693.2.2.el7.x86_64
  8. WARNING: /usr/bin/staprun exited with status: 1
  9. Pass 5: run completed in 0usr/10sys/38real ms.
  10. Pass 5: run failed. [man error::pass5]

因为中间生成的C文件和ko模块都是用的cache (蓝色标注的部分),我们把上面的cache文件删除,再重新运行,这次可以成功了。

  1. # stap -e 'probe begin{printf("Hello, World"); exit();}'
  2. Hello, World

 

https://sourceware.org/systemtap/SystemTap_Beginners_Guide/using-systemtap.html#using-setup

ERROR: module version mismatch

https://groups.google.com/forum/#!topic/openresty/nlEc3qlDyOc

 

 

版权声明:本文为wipan原创文章,遵循 CC 4.0 BY-SA 版权协议,转载请附上原文出处链接和本声明。
本文链接:https://www.cnblogs.com/wipan/p/9333623.html