Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

What is a glibc free/malloc/realloc invalid next size/invalid pointer error and how to fix it?

Tags:

c++

c

memory

glibc

You are most likely seeing this question because your question has been closed as a duplicate of this. For a moderately complete list of related questions, please see A long list of possible duplicates — C memory allocation and overrunning bounds on Meta Stack Overflow.


Example Question

From free char*: invalid next size (fast) asked by noobie on 2014-04-11.

I am freeing a char* after a concatenation process, but I receive this error:

free(): invalid next size (fast): 0x0000000001b86170

This is my code:

void concat(stringList *list) {
    char *res = (char*)malloc(sizeof(char*));

    strcpy(res, list->head->string);

    list->tmp = list->head->next;
    while (list->tmp != NULL) {
        strcat(res, ",");
        strcat(res, list->tmp->string);
        list->tmp = list->tmp->next;
    }

    printf("%s\n", res);
    free(res);
}

Generic Question

When running my program, I see an error message like this:

*** glibc detected *** ./a.out: free(): corrupted unsorted chunks: 0x12345678 ***

The detailed information can contain any of the following after the *** glibc detected *** and the program name, and the message is followed by a hexadecimal address (shown as 0x12345678) and another ***:

  • free(): corrupted unsorted chunks: 0x12345678
  • free(): invalid next size (fast): 0x12345678
  • free(): invalid next size (normal): 0x12345678
  • free(): invalid pointer: 0x12345678
  • free(): invalid size: 0x12345678
  • malloc(): corrupted unsorted chunks: 0x12345678
  • malloc(): corrupted unsorted chunks 2: 0x12345678
  • malloc(): memory corruption: 0x12345678
  • malloc(): memory corruption (fast): 0x12345678
  • malloc(): smallbin double linked list corrupted: 0x12345678
  • munmap_chunk(): invalid pointer: 0x12345678
  • realloc(): invalid next size (fast): 0x12345678
  • realloc(): invalid old size (fast): 0x12345678
  • realloc(): invalid pointer: 0x12345678
  • corrupted double-linked list: 0x12345678

This happens while calling the frobnicate() function; what is wrong with that function?

like image 839
PlasmaHH Avatar asked May 15 '14 13:05

PlasmaHH


1 Answers

Answer for Example Question

unwind gave the accepted answer to the example question:

Your code is wrong.

You are allocating space for a single pointer (malloc(sizeof(char*))), but no characters. You are overwriting your allocated space with all the strings, causing undefined behavior (in this particular case, corrupting malloc()'s book-keeping data).

You don't need to allocate space for the pointer (res); it's a local variable. You must allocate space for all the characters you wish to store at the address held by the pointer.

Since you're going to be traversing a list to find strings to concatenate, you can't know the total size upfront. You're going to have to do two passes over the list: one to sum the strlen() of each string, then allocate that plus space for the separator and terminator, then another pass when you actually do the concatenation.

Generic Answer

What you are seeing is the result of a corruption in the internal structures of the glibc allocator. When you are allocating or freeing dynamic memory, the allocator has to manage the memory it reserved from the OS and, depending on the action requested by you, find a new chunk to hand out, sort a freed chunk into the list of those that it can hand out later again, or give the memory back to the operating system. These error messages show that the data structures it uses to manage this functionality are corrupted.

These errors all mean that some of your code has modified memory that it was not given to use, invoking undefined behaviour. This is most likely the result of overwriting some memory quite a bit earlier in your program, and it is totally possible that the error does not lie in the frobnicate() function.

Yes, this means that the error can be anywhere in your program or 3rd party libraries you use.

This is probably not a good question for Stack Overflow. Unless you have a good simple reproduction of your problem, this community may be unable to help you very much. The cause of the error can be anywhere in your code (and is very often not in the function where the error is spotted), and it may be in code that we cannot see. Stack Overflow is not a collaborative debugging site. Even when someone can find the flaw in your code, it is unlikely that your specific question will ever help any future visitor.

Common causes

  • Use after free. You have freed/deleted some memory and writing into it afterwards, overwriting the structures glibc needs for bookkeeping.
  • Off-by-N error. You are writing N bytes after an allocated chunk into unallocated memory that glibc uses internally for its bookkeeping.
  • Uninitialized pointers. You are not initializing a pointer. By coincidence it points to some memory reserved by glibc but not allocated by your program and you write to it.
  • Allocating the wrong amount of space. This can be because you wrote long *data = malloc(number * 4) instead of long *data = malloc(number * sizeof(long)); or (better) long *data = malloc(number * sizeof(*data));. There are many other ways to get the size calculation wrong. Another common one is to forget to account for the null terminator character at the end of a string: char *copy = malloc(strlen(str)); instead of char *copy = malloc(strlen(str)+1);.

What you need to do now is to roll up your sleeves and debug that problem

There is no simple answer what to look for, or what to fix. No single syntactical construct that you were using wrong. The cause of this bug can come in literally thousands of varieties.

Tools

  • valgrind A tool created mostly for the purpose of finding exactly this kinds of errors. If it can't find anything make sure you are using the latest version, and you are also trying out the included exp-sgcheck tool. If you are running multithreaded code, the cause might also be related to a race condition so you might want to try the included race condition checkers drd and helgrind for more insight. At the point of writing this, valgrind supports the following platforms:
    • X86/Linux,
    • AMD64/Linux,
    • ARM/Linux,
    • PPC32/Linux,
    • PPC64/Linux,
    • S390X/Linux,
    • MIPS32/Linux,
    • MIPS64/Linux,
    • ARM/Android (2.3.x and later),
    • X86/Android (4.0 and later),
    • X86/Darwin and
    • AMD64/Darwin (Mac OS X 10.7, with limited support for 10.8).
  • purify A similar tool to valgrind, but commercial and aimed at a different set of platforms.
  • AddressSanitizer A similar tool, but integrated into the compiler toolchain (gcc and clang).
  • efence A drop in allocator replacement that will try to crash your program earlier, so that you can find out with a normal debugger where the write to invalid memory happened.
  • dmalloc a library with a similar purpose as efence.

Needing more assistance

If you can't solve your problem using one these tools, you should try to create an MCVE (How to create a Minimal, Complete, and Verifiable Example?) or, equivalently, an SSCCE (Short, Self Contained, Correct (Compilable), Example).

Remember to work on a copy of your code because creating an MCVE requires you to ruthlessly remove code that does not help reproduce the problem. Using a VCS (version control system) to assist is a good idea; you can record intermediate stages in reducing the problem to a minimum. It might be a new throw-away repository just for reducing your problem to a manageable size.

With a good modular design to your code, it should be relatively easy to create the MCVE. Maybe you also already have a unit test that is better suited to be fed into one of the above tools. You also might just want to create one that can later serve as a regression test for this bug.

like image 129
PlasmaHH Avatar answered Nov 13 '22 01:11

PlasmaHH