Calling memset causes segmentation fault

python_newbie picture python_newbie · May 28, 2013 · Viewed 16.8k times · Source

This program causes a seg fault on my UNIX machine. I narrowed the cause down to the second call of memset().

Why is this behaviour occurring? The first "chunk" of code is almost the same as the second, isn't it? Why didn't the first call of memset segfault if the second one does?

I looked over other threads concerning segfaulting memset calls, but none of them were similar to this.

In case you were wondering why I wrote such a trivial program, it's adapted from another program I've been writing that I was using to teach myself how to apply memcpy() to structures.

#include <stdio.h>
#include <stdlib.h>

typedef struct{
    int x;
    char text;
} FilesStruct;

int
main(int argc, char** argv)
{
    FilesStruct *old_dir;
    memset(old_dir,0,sizeof(FilesStruct));
    old_dir->x = 3;
    old_dir->text = 'c';
    printf("old dir: %d,%c\n",old_dir->x,old_dir->text);

    FilesStruct *new_dir;
    memset(new_dir,0,sizeof(FilesStruct));
    new_dir->x = 7;
    new_dir->text = 'g';
    printf("new dir: %d,%c\n",new_dir->x,new_dir->text);

    return 0;
}

Answer

simonc picture simonc · May 28, 2013
FilesStruct *old_dir;
memset(old_dir,0,sizeof(FilesStruct));

attempts to write to an uninitialised pointer. This results in undefined behaviour, including possibly a crash. Its just luck (good or bad, depending on how you look at it) that the first instance of this behaviour didn't crash.

You need to allocate memory for old_dir. The easiest way to do this is to declare it on the stack

FilesStruct old_dir;
memset(&old_dir,0,sizeof(old_dir));

or you could dynamically allocate on the heap (making sure to call free when you no longer need the object)

FilesStruct *old_dir = calloc(1, sizeof(*old_dir);
/* use old_dir */
free(old_dir);

The same applies to new_dir further down your code.