Skip to main content

关于在bochs的gdb-stub中加入对虚拟内存访问的支持

最近操统课需要用到bochs这个模拟器，我用的2.3.5. 但是实际出现了以个问题，就是gdb-stub这个功能不大正常，运行gdb后可以正常下断点，但是单步进可以，单步过却和单步进相同，这个问题不大比较恶心的是无法显示局部变量的值，总是显示0xffffffff
　　 　　 　 　　 　　 　　　 　 　　　　
google了半天没看到一点有用的。。。。。唉，只好自己搞
　　 　　 　 　　 　　 　　　 　 　　　　
参阅了gdb protocol后我修改了bochs的gdbstub.cc，让其显示出与gdb交互的内容，大致了解了内存这部分的步骤，主要是gdb给server(即bochs)发一个内存读取的请求，格式为m addr,length, 然后server返回十六进制的数据
　　 　　 　 　　 　　 　　　 　 　　　　
在bochs的gdbstub.cc 558行为处理这个指令的，不难看懂，我修改了它，让它输出access_linear的返回值，然后再调试，查看局部变量，果然valid为0，进一步，到access_linear, 经过痛苦地多次调试，发现问题处在valid=BX_MEM(0)->dgb_fetch_mem...一行，最后进到或则个函数，在memory/misc_mem.cc 512行下看到了这部分，原来是bochs发现读取的地址越界，就用0xff填充，然后返回0， 我修改了0xff，然后用gdb调试时证实了这一点。
　　 　　 　 　　 　　 　　　 　 　　　　
最后往上看，发现是addr比内存的长度还要大， 加入调试输出后发现gdb读的地址是0xf010
9fe0左右，而BX_MEM_THIS len仅有2000000，这显然会越界。
　　 　　 　 　　 　　 　　　 　 　　　　
起初我考虑也许是那里关于内存大小的设置没有弄好，但没有找到具体位置，但后来发现其实那BX_MEM_THIS len 是在.bochsrc里有定义的,0x2000000=2^25=32k，而那个0xf0109fe0是内核做的虚拟内存，gdb只跟bochs打交道，而虚拟内存是内核弄的，虚拟机又管不着。因此好像没什么好办法。。。

之后我查了很多关于虚拟内存的资料，发现其实是cpu直接支持的，于是我看到了希望。仍是从bochs的源码开刀。

一开始我是把目光放在了gdt和ldt上，但是未果。最后又发现了它里面的CPU类有一个get_segment_base函数，几经尝试，找到了临时解决方法：

修改bochs的gdbstub.cc
将567行附近的
access_linear(addr,len,BX_READ,mem);
改成
access_linear((BX_CPU(0)->get_segment_base(BX_SEG_REG_CS))+addr,len,BX_READ,mem)
;
然后重新编译，记得是用gdb stub版本(./configure --enable-gdb-stub)

思路就是在gdb服务器端自动对虚拟内存作个映射。目前发现常用的关于断点，堆栈，数据，单步等相关命令都基本正常（不过好像有同学说w命令不好使，我没确认）。

不知道这个算不算bochs的一个bug, 跟它反映了一下，至今没有回应。

不过这还是第一次半系统地阅读了一个开源软件的代码，另外涉及到gdb协议，虚拟内存，cpu等知识，混在一起，最后还弄成功了，实在很过瘾。
　　
Labels:

Comments

Post a Comment

Popular posts from this blog

Determine Perspective Lines With Off-page Vanishing Point

In perspective drawing, a vanishing point represents a group of parallel lines, in other words, a direction. For any point on the paper, if we want a line towards the same direction (in the 3d space), we simply draw a line through it and the vanishing point. But sometimes the vanishing point is too far away, such that it is outside the paper/canvas. In this example, we have a point P and two perspective lines L1 and L2. The vanishing point VP is naturally the intersection of L1 and L2. The task is to draw a line through P and VP, without having VP on the paper. I am aware of a few traditional solutions: 1. Use extra pieces of paper such that we can extend L1 and L2 until we see VP. 2. Draw everything in a smaller scale, such that we can see both P and VP on the paper. Draw the line and scale everything back. 3. Draw a perspective grid using the Brewer Method. #1 and #2 might be quite practical. #3 may not guarantee a solution, unless we can measure distances/p...
3 comments
Read More »

Chasing an IO Phantom

My home server has been weird since months ago, it just becomes unresponsive occassionally. It is annoying but it happens only rarely, so normally I'd just wait or reboot it. But weeks ago I decided to get to the bottom of it. What's Wrong My system set up is: Root: SSD, LUKS + LVM + Ext4 Data: HDD, LUKS + ZFS 16GB RAM + 1GB swap Rootless dockerd The system may become unresponsive, when the IO on HDD  is persistantly high for a while. Also: Often kswapd0 has high CPU High IO on root fs (SSD) From dockerd and some containers RAM usage is high, swap usage is low It is very strange that IO on HDD can affect SSD. Note that when this happens, even stopping the IO on HDD does not always help. Usually restarting dockerd does not help, but rebooting helps. Investigation: Swap An obvious potential root cause is the swap. High CPU on kswapd0 usually means the free memory is low and the kernel is busy exchanging data between disk and swap. However, I tried the following steps, none of the...
Post a Comment
Read More »

Fix Google Security Code

Google Security Code (http://g.co/sc) is one type of 2-step verification. This is particularly useful when security keys and passkeys are not available. I have been using it in my LXC containers, until today I found out that it stopped working. It just kept saying "The code is invalid". It is easy to rule out some factors: The code works on other browsers on my laptop. The code works on other devices that are directly connected to the router. So it appears that Google also checks IP addresses besides the security code. Recently I have IPv6 enabled, so most devices that are directly connected to the router have both IPv4 and IPv6 addresses. But  I only enabled IPv4 for my LXC containers. So I guess when a code is generated by device A and used by device B, Google should be able to check that device A and device B are closely located. But in my case, IPv6 address appears on device A but not on device B, which may look suspicious. To fix the problem, I just needed to disable IPv...
Post a Comment
Read More »