ulimit 设置正确的分段错误 [英] Segmentation fault with ulimit set correctly

问题描述

我试图在 这个问题.

我发现即使堆栈设置为 2000 KB，像下面这样的代码也会随机导致分段错误.

I found out that a code like the one below causes segmentation fault randomly even if the stack is set to 2000 Kbytes.

int main ()
{
   int a[510000];
   a[509999] = 1;
   printf("%d", a[509999]);
   return 0;
}

如您所见，数组为 510000 x 4 字节 = 2040000 字节.

As you can see the array is 510000 x 4 bytes = 2040000 bytes.

使用 ulimit 命令将堆栈设置为 2000 KB(2048000 字节):

The stack is set to 2000 Kbytes (2048000 bytes) using ulimit command:

• ulimit -s 2000
• ulimit -Ss 2000

基于这些数字，应用程序有空间存储数组，但随机返回分段错误.

Based on those numbers the application has room to store the array, but randomly it return segmentation fault.

有什么想法吗?

推荐答案

你不能这样做有几个原因.有些东西已经在使用你的堆栈.

There's a few reasons why you can't do this. There are things that are already using parts of your stack.

main 不是堆栈中的第一件事.真正入口点、动态链接器等调用的函数在 main 之前，它们可能都在使用一些堆栈.

main is not the first thing on your stack. There are functions called by the real entry point, dynamic linker, etc. that are before main and they are all probably using some of the stack.

此外，通常可以将一些东西放在堆栈顶部以设置执行.我知道的许多系统都将所有字符串放在 argv 中，并将所有环境变量放在堆栈顶部(这就是为什么 main 不是入口点的原因，通常在 main 之前运行的代码会为 main 设置环境变量和 argv).

Additionally, there can be things that are usually put on the top of the stack to set up execution. Many systems I know put all the strings in argv and all environment variables on top of the stack (which is why main is not the entry point, there's usually code that runs before main that sets up environment variables and argv for main).

如果您的系统这样做，可能会故意浪费一部分堆栈以增加 ASLR 的随机性.

And to top it off a part of the stack can be deliberately wasted to increase the randomness of ASLR if your system does that.

在调试器中运行您的程序，在 main 处添加一个断点，查找堆栈寄存器的值并检查它上面的内存(请记住，除非您使用的是奇怪的架构，否则您的堆栈很可能会变小).我打赌你会在那里找到很多指针和字符串.我只是在 linux 系统上做的，因为我怀疑我所有的环境变量都在那里.

Run you program in the debugger, add a breakpoint at main, look up the value of the stack register and examine the memory above it (remember that most likely your stack grows down unless you're on a weird architecture). I bet you'll find lots of pointers and strings there. I just did this on a linux system and as I suspected all my environment variables were there.

Unix 上的资源限制 (ulimit) 的目的从未真正将事物微观管理到一个字节/微秒，它们只是为了阻止你的程序完全疯狂并用它来破坏整个系统.不要将它们视为正确道路上的红灯和停车标志，而应将它们视为跑道上的跑​​道和防撞护栏.

The purpose of resource limits (ulimit) on Unix has never really been to micromanage things down to a byte/microsecond, they are there just to stop your program from going completely crazy and taking down the whole system with it. See them not as red lights and stop signs on a proper road, see them as run-off areas and crash barriers on a racetrack.

这篇关于ulimit 设置正确的分段错误的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！