How to tokenize a string using strsep()

Question

I have a kernel module, and in the module, I'm trying to split a string using strsep() . I have the following decrypt variable that contains the string I'm trying to split.

unsigned char decrypt[KEY_SIZE];
printk(KERN_DEBUG "%s\n", decrypt);

output:

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
E = 010001
D = 21AFE07431CE47269083F8F8B7ABCBCEDA6DCB975457BE6662942C64091586FEE755C9A3832EAA0868665DB507A41A15F1EDF12E44ECF03D0E61111D457D730FA700D0FB0B6C13607C0F5F1DDDEB61AE9019E53A9C998F2AD5924430EEA3E9DA1B0E5F2B575DDBE86C4096B5C87661F7A7E7F7F21D0701509BBA881B4AE463F6F18C7F04AB742319E2D7319EECA136EEB0CF7B2BFA87E3A0E69FBC0E5FDC7EE6271EB2CA09DDBF7C8B57D951762708D76890E62858C1D5FC5B7E40D50913CE7797BD80F6A398FB92703FBDD33FBCB129B86E54F13EC14DA68BE139634DD1E9C01F01751
...
...

I'm using the following code to extract the values. My goal is to get the value of N , E , D in each case. When I call this module, my machine freeze. However, when I use gdb to debug the loop, it works.

As @John Bollinger asked, I have the following line to make the string null-terminated before using strsep() .

size_t lenght = strlen(decrypt);
int N = lenght - 2361; // 2361 is the original size
decrypt[lenght - N] = '\0';

Code:

char *s3 = decrypt;
int k = 0;
int size = 0;
char *test;

while (s3 != NULL) {
    test = strsep(&s3, " ");
    test = strsep(&s3, " ");
    test = strsep(&s3, "\n");

    switch (k) {
      case 0:
        size = strlen(test);
        printk(KERN_DEBUG "token id %d: size %d, token is %s\n", k, size, test);
        break;

      case 1:
        size = strlen(test);
        printk(KERN_DEBUG "token id %d: size %d, token is %s\n", k, size, test);
        break;
        
      case 2:
        size = strlen(test);
        printk(KERN_DEBUG "token id %d: size %d, token is %s\n", k, size, test);
        break;
  
      ........
      ........
    }    
    k = k + 1;
} 

Can someone please tell me, what am I doing incorrectly here? Or is there any other thread-safe function available to split a string? Thanks in advance.

Kernel version: Linux 4.15.0-142-generic

Answer 1

Can someone please tell me, what am I doing incorrectly here?

The one thing I see that you are clearly doing wrong is not validating your results before using them. If the input happens not to be formatted exactly as you expect, then it is easily possible for test to be null at entry to the switch , and the code does not appear to anticipate that possibility.

Of course, "not formatted exactly as you expect" might also mean having the wrong expectations. For example, one way to trigger the aforementioned null value of test would be for the last line of the input to end with a newline. In that case, strsep would not notice that it had reached the end of the string until the next cycle.

Also, this declaration is suspicious:

unsigned char decrypt[KEY_SIZE];

If the data can be up to KEY_SIZE in length, then that leaves no room for a string terminator. You must have a string terminator if you are processing it as a string via strsep() or outputting it or a tail of it as a string via printk . If the data don't naturally have a terminator then you need to make sure to add one (for which you must leave space).

Additionally, not wrong per se , but wasteful: calling strlen() , except for the last token. You can get the length of each other token by simple pointer difference: s3 - test - 1 .