gvannest / ft_malloc Goto Github PK

bonuses

• ft_calloc
• making malloc thread safe
• defrag
• print memory
• les 2 fonctions du collegue d' Aziz

Main documents

https://sourceware.org/glibc/wiki/MallocInternals#Overview_of_Malloc

https://sploitfun.wordpress.com/2015/02/10/understanding-glibc-malloc/

https://www.blackhat.com/presentations/bh-usa-07/Ferguson/Whitepaper/bh-usa-07-ferguson-WP.pdf

https://wiki-prog.infoprepa.epita.fr/images/0/04/Malloc_tutorial.pdf

Decisions

• One global variable : pointer to first free chunk in the free list.

• Two header types: one for free and one for allocated chunks (cf doc above)

• One call mmap = one heap (for ex 132Kb for tiny) corresponding to one allocaiton type (tiny or small). Once the heap is full -> call mmap again and link the two heaps with a pointer at the end of the first heap

• Large allocations => direct call to mmap (no heap) and direct call to unmmap when free is called on them

• when a chunk is free'd : check flag (P) to see if previous chunk is free, and combine them in the free list in case it is. However change header in heap block

• WATCH OUT: the free list is only pointers to free chunks. indeed all the info (pointers to next and previous bloc + size) are in the header which has been changed when the program calls free on an allocated chunk. IT IS NOT A NEW DATA STRUCTURE! We only have the first pointer of the free list in a global variable, and then in the headers of the free chunks we have the necessary data to jump to the next free chunk + size

• WATCH OUT: ALIGNMENT. chunks must be multiples of 8 bytes. So minimum allocated chunk is 16 bytes : 2 times 4 bytes for the 2 size_t variables in the header, + minimum 8 bytes for user data.

• for defragmentation: when triggger (len of free list > threshold?) => defrag

• Errors management: check what fbabin has done on ftp. Using an enum with error codes?

To be decided

• Do we need a footer on the heap to know where the end of the allocated heap lies?
• when do we call unmmap on allocated heaps? WATH OUT: when unmmap is called do not forget to clean free list with free chunks in the unmmap'd region! SHould we wait for two heaps to be empty berfore unmmapping the last one?
• threshold for defragmentation
• no decision on realloc. to be taken

Main sub topics

mandatory part

• ft_malloc
• ft_free
• ft_realloc

Child issues

• ft_malloc: #2 (comment)

To check

• what is the behavior of the real free when the address does not exist?
• munmap les heaps vides
• checker les doubles free

Main issue for the ft_malloc function which deals with the allocation of memory

Note: free is in another issue, please refer to #1

Tasks

• Create a heap: call mmap, add header, unmmap

specific documentation for implementation

• analysis of malloc source code
  https://code.woboq.org/userspace/glibc/malloc/malloc.c.html#malloc_chunk

• making a global variable available in several .c files: https://stackoverflow.com/questions/1433204/how-do-i-use-extern-to-share-variables-between-source-files
  https://stackoverflow.com/questions/1410563/what-is-the-difference-between-a-definition-and-a-declaration
  https://www.tutorialspoint.com/extern-keyword-in-c

• chunk headers:
  https://reverseengineering.stackexchange.com/questions/19293/heap-chunk-structure-does-not-contain-previous-section-info

• malloc minimum chunk size (from malloc source code):

If n is zero, malloc returns a minumum-sized chunk. (The minimum
--
size is 16 bytes on most 32bit systems, and 24 or 32 bytes on 64bit
systems.)  On most systems, size_t is an unsigned type, so calls
with negative arguments are interpreted as requests for huge amounts
of space, which will often fail. The maximum supported value of n
differs across systems, but is in all cases less than the maximum
representable value of a size_t.

Main decisions on implementation

• @capaquet I suggest for simplicity that we keep the previous size in the header of every chunk, and not only those where the previous chunk is free. The goal of the true implementation is not to loose expensive memory, but I think it over complicates the job her