使用gdb调试python脚本

最新推荐文章于 2024-07-19 06:16:03 发布

zzhongcy

最新推荐文章于 2024-07-19 06:16:03 发布

阅读量4.5k

点赞数 1

分类专栏： Linux gdb dbx 文章标签： Linux gdb

Linux gdb dbx 专栏收录该内容

5 篇文章 0 订阅

订阅专栏

     调试python脚本一般可通过记录log和使用python自带的pdb模块完成，但凡事总有例外，在以下三种情况时上述方法就无能为力了。
   1 段错误
   2 运行中的daemon程序
   3 core dump
这个时候就需祭出gdb进行调试。python2.6的源码中提供了部分预定义函数以便大家使用gdb调试，我们只需将文件Python-2.6/Misc/gdbinit所包括的内容加入到用户目录下的.gdbinit文件中即可，这样每次启动gdb时会自动完成这些宏的定义。但可惜的是Python2.6.2 gdbini对于pylocals的定义居然有错误，看来是没有随着代码的更新而同步更新。我们只需将 while $_i < f->f_nlocals修改为 while $_i < f->f_code->co_nlocals即可。文章后面所附的几个宏建议也加入的.gdbinit文件中，更多的宏可参考

http://web.archive.org/web/20070915134837/

http://www.mashebali.com/?Python_GDB_macros:The_Macros。

我们首先需要构造一个会造成段错误的python脚本。老实说让python发生段错误并不容易，但通过其外部调用库就很简单了。我们将该文件命名为gdb_test.py
import sys, os, libxml2

def segv_test():
    s = "<html><body><div><a><a></a></a><a></a></div></body></html>"
    options = libxml2.HTML_PARSE_RECOVER + \
              libxml2.HTML_PARSE_NOERROR + \
              libxml2.HTML_PARSE_NOWARNING
    doc = libxml2.htmlReadDoc(s, None, 'utf-8', options).doc
    ctxt = doc.xpathNewContext()
    nodes = ctxt.xpathEval('//body/node()')
    nodes.reverse()
    for note in nodes:
        nexts = note.xpathEval('node()')
        note.unlinkNode()
        note.freeNode() //freeNode会将该节点及其子节点释放掉
        nexts[0].unlinkNode()
        nexts[0].freeNode() //资源已经释放，再次释放会造成段错误

def main():
segv_test()

if __name__ == "__main__":
main()

使用gdb运行该脚本,我们会得到段错误信息。
gdb python
r gdb_test.py

*** glibc detected *** double free or corruption (fasttop): 0x08104570 ***

Program received signal SIGABRT, Aborted.
[Switching to Thread -1208260928 (LWP 26159)]
0x00b987a2 in _dl_sysinfo_int80 () from /lib/ld-linux.so.2

   键入bt得到如下堆栈信息:
(gdb) bt
#0 0x00b987a2 in _dl_sysinfo_int80 () from /lib/ld-linux.so.2
#1 0x00c00825 in raise () from /lib/tls/libc.so.6
#2 0x00c02289 in abort () from /lib/tls/libc.so.6
#3 0x00c34cda in __libc_message () from /lib/tls/libc.so.6
#4 0x00c3b56f in _int_free () from /lib/tls/libc.so.6
#5 0x00c3b94a in free () from /lib/tls/libc.so.6
#6 0x009812c6 in xmlFreeNode () from /opt/sohumc/lib/libxml2.so.2
#7 0x0029d7f3 in libxml_xmlFreeNode () from /opt/sohumc/lib/python2.6/site-packages/libxml2mod.so
#8 0x00780bae in PyCFunction_Call (func=0x8104570, arg=0xd05820, kw=0x6) at Objects/methodobject.c:116
#9 0x007d8c79 in call_function (pp_stack=0xbff8c48c, oparg=0) at Python/ceval.c:3679
#10 0x007d6d2b in PyEval_EvalFrameEx (f=0x8124ef4, throwflag=0) at Python/ceval.c:2370
#11 0x007d8e36 in fast_function (func=0x6, pp_stack=0xbff8c5dc, n=1, na=1, nk=0) at Python/ceval.c:3765
#12 0x007d89cd in call_function (pp_stack=0xbff8c5dc, oparg=0) at Python/ceval.c:3700
#13 0x007d6d2b in PyEval_EvalFrameEx (f=0x81242fc, throwflag=0) at Python/ceval.c:2370
#14 0x007d8e36 in fast_function (func=0x6, pp_stack=0xbff8c72c, n=0, na=0, nk=0) at Python/ceval.c:3765
#15 0x007d89cd in call_function (pp_stack=0xbff8c72c, oparg=0) at Python/ceval.c:3700
#16 0x007d6d2b in PyEval_EvalFrameEx (f=0x810a7c4, throwflag=0) at Python/ceval.c:2370
#17 0x007d8e36 in fast_function (func=0x6, pp_stack=0xbff8c87c, n=0, na=0, nk=0) at Python/ceval.c:3765
#18 0x007d89cd in call_function (pp_stack=0xbff8c87c, oparg=0) at Python/ceval.c:3700
#19 0x007d6d2b in PyEval_EvalFrameEx (f=0x8091d0c, throwflag=0) at Python/ceval.c:2370
#20 0x007d76f9 in PyEval_EvalCodeEx (co=0xb7fa3728, globals=0x6, locals=0xb7f9902c, args=0x0, argcount=0, kws=0x0, kwcount=0, defs=0x0, defcount=0,
    closure=0x0) at Python/ceval.c:2942
#21 0x007d47cb in PyEval_EvalCode (co=0xb7fa3728, globals=0xb7f9902c, locals=0xb7f9902c) at Python/ceval.c:515
#22 0x007fbbce in run_mod (mod=0x80ea780, filename=0xbffc6be6 "gdb_test.py", globals=0xb7f9902c, locals=0xb7f9902c, flags=0xbff8ca8c, arena=0x807ef28)
    at Python/pythonrun.c:1330
#23 0x007fbb58 in PyRun_FileExFlags (fp=0x8091d00, filename=0xbffc6be6 "gdb_test.py", start=257, globals=0xb7f9902c, locals=0xb7f9902c, closeit=1,
    flags=0xbff8ca8c) at Python/pythonrun.c:1316
#24 0x007fb22d in PyRun_SimpleFileExFlags (fp=0x8091d00, filename=0xbffc6be6 "gdb_test.py", closeit=1, flags=0xbff8ca8c) at Python/pythonrun.c:926
#25 0x007facc9 in PyRun_AnyFileExFlags (fp=0x8091d00, filename=0xbffc6be6 "gdb_test.py", closeit=1, flags=0xbff8ca8c) at Python/pythonrun.c:731
#26 0x00808fea in Py_Main (argc=1, argv=0xbff8cbb4) at Modules/main.c:597
#27 0x080486ae in main (argc=2, argv=0xbff8cbb4) at Modules/python.c:23

   pystack和pystackv两个宏可用来查看python内部的栈情况；可以看到程序时执行到freeNode函数时结束，该函数位于libxml2.py的3141行。
(gdb) pystack
/opt/lib/python2.6/site-packages/libxml2.py (3141): freeNode
gdb_test.py (17): segv_test
gdb_test.py (21): main
gdb_test.py (24): <module>

   通过堆栈我们可以看到脚本内部各函数的调用关系，那么我们如何查看函数内变量情况呢？正如大家所， python内部堆栈和函数的调用由PyEval_EvalFrameEx完成的，一次PyEval_EvalFrameEx意味着一次函数调用，象上面的第19,13,10行分别对应于main, segv_test, freeNode函数，将gdb定位到对应行后，使用pylocals宏即可查看该函数内部变量的详细情况。
(gdb) up 13
#13 0x007d6d2b in PyEval_EvalFrameEx (f=0x81242fc, throwflag=0) at Python/ceval.c:2370
2370    in Python/ceval.c
(gdb) pylocals
s:
object : '<html><body><div><a><a></a></a><a></a></div></body></html>'
type    : str
refcount: 3
address : 0xb7f64440
options:
object : 97
type    : int
refcount: 7
address : 0x8082c20
doc:
object : <xmlDoc (None) object at 0xb7cc04ec>
type    : instance
refcount: 1
address : 0xb7cc04ec
ctxt:
object : <libxml2.xpathContext instance at 0xb7f70ccc>
type    : instance
refcount: 1
address : 0xb7f70ccc
nodes:
object : [<xmlNode ((儓X? object at 0xb7cc0cac>]
type    : list
refcount: 2
address : 0xb7f70a8c
note:
object : <xmlNode ((?圶? object at 0xb7cc0cac>
type    : instance
refcount: 2
address : 0xb7cc0cac
nexts:
object : [<xmlNode (hhX? object at 0xb7cc750c>, <xmlNode (HXX? object at 0xb7cc76cc>, <xmlNode (@XX? object at 0xb7c9348c>]
type    : list
refcount: 1
address : 0xb7f4ce4c

   脚本调试时断点的设置是个很麻烦的东西，我所能想到的有两种方法：1 根据函数的python源码进行断点设置；2 采用sleep函数和ctrl+c来中断程序的运行。无论怎么样使用逐条执行进行调试都是很痛苦的事情，因为这个时候python解释器本身要做很多工作。
   由于本身对于python源码不是很熟悉，因此对如何使用gdb对python脚本调试上也只是很粗略的理解，这里权当抛砖引玉，欢迎达人们给出分享的经验。

额外的宏：
define pbt
set $i = 0
set $j = 0
while $i < 1000
select $i
if $eip >= &PyEval_EvalFrameEx
if $eip < &PyEval_EvalCodeEx
echo c frame #
p $i
echo py frame #
p $j
set $j = $j+1
x/s ((PyStringObject*)f->f_code->co_filename)->ob_sval
x/s ((PyStringObject*)f->f_code->co_name)->ob_sval
echo line #
p f->f_lineno
end
end
set $i = $i+1
end
end
document pbt
show python backtrace
end

define pyattrlist
set $dict = *(PyDictObject**)(((int)$arg0)+$arg0.ob_type.tp_dictoffset)
set $i = 0
set $j = 0
while $i < $dict.ma_mask
if 0 != $dict.ma_table[$i].me_value
   echo \nattr#:
   p $j
   x/s ((PyStringObject*)$dict.ma_table[$i].me_key).ob_sval
   pyobjinfo $dict.ma_table[$i].me_value
   set $j = $j+1
end
set $i = $i+1
end
end
document pyattrlist
show pythonic object attributes list
end

define pyattr
set $dict = *(PyDictObject**)(((int)$arg0)+$arg0.ob_type.tp_dictoffset)
set $i = 0
set $j = 0
while $i < $dict.ma_mask
if 0 != $dict.ma_table[$i].me_value
   if $j == $arg1
    set $attr = $dict.ma_table[$i].me_value
    echo $attr:
    p $attr
   end
   set $j = $j+1
end
set $i = $i+1
end
end
document pyattr
get pythonic object attribute
usage: pyattr <pyobj> <attr#>
atr# correlates to those shown in pyattrlist
end