[issue38106] Race in PyThread_release_lock - can lead to memory corruption and deadlock

[Kirill Smelkov]

Kirill Smelkov  added the comment:

Ok, I did https://github.com/python/cpython/pull/16047.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue38106] Race in PyThread_release_lock - can lead to memory corruption and deadlock

[Kirill Smelkov]

Change by Kirill Smelkov :


--
keywords: +patch
pull_requests: +15669
stage:  -> patch review
pull_request: https://github.com/python/cpython/pull/16047

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue38106] Race in PyThread_release_lock - can lead to memory corruption and deadlock

[Armin Rigo]

Armin Rigo  added the comment:

I agree with your analysis.  I guess you (or someone) needs to write an 
explicit pull request, even if it just contains 187aa545165d cherry-picked.  
(I'm not a core dev any more nowadays)

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue38106] Race in PyThread_release_lock - can lead to memory corruption and deadlock

[Kirill Smelkov]

Kirill Smelkov  added the comment:

I agree it seems like a design mistake. Not only it leads to suboptimal
implementations, but what is more important, it throws misuse risks onto the 
user.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue38106] Race in PyThread_release_lock - can lead to memory corruption and deadlock

[Kristján Valur Jónsson]

Kristján Valur Jónsson  added the comment:

imho posix made a mistake in allowing signal/broadcast outside the mutex.  
Otherwise an implementation could rely on the mutex for internal state 
manipulation.  I have my own fast condition variable lib implemented using 
semaphores and it is simple to do if one requires the mutex to be held for the 
signal event.

Condition variables semantics are otherwise quite brilliant. For example, 
allowing for spurious wakeups to occur allows, again, for much simpler 
implementation.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue38106] Race in PyThread_release_lock - can lead to memory corruption and deadlock

[Kirill Smelkov]

Kirill Smelkov  added the comment:

And it is indeed better to always do pthread_cond_signal() from under mutex.
Many pthread libraries delay the signalling to associated mutex unlock, so
there should be no performance penalty here and the correctness is much more
easier to reason about.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue38106] Race in PyThread_release_lock - can lead to memory corruption and deadlock

[Kirill Smelkov]

Kirill Smelkov  added the comment:

Thanks for feedback. Yes, since for Python-level lock, PyThread_release_lock() 
is called with GIL held:

https://github.com/python/cpython/blob/v2.7.16-129-g58d61efd4cd/Modules/threadmodule.c#L69-L82

the GIL effectively serves as the synchronization device in between T2
releasing the lock, and T1 proceeding after second sema.acquire() when it gets 
to
execute python-level code with `del sema`.

However

a) there is no sign that this aspect - that release must be called under GIL -
   is being explicitly relied upon by PyThread_release_lock() code, and

b) e.g. _testcapimodule.c already has a test which calls
   PyThread_release_lock() with GIL released:

   
https://github.com/python/cpython/blob/v2.7.16-129-g58d61efd4cd/Modules/_testcapimodule.c#L1972-L2053
   
https://github.com/python/cpython/blob/v2.7.16-129-g58d61efd4cd/Modules/_testcapimodule.c#L1998-L2002

Thus, I believe, there is a bug in PyThread_release_lock() and we were just
lucky not to hit it due to GIL and Python-level usage.

For the reference, I indeed started to observe the problem when I moved locks
and other code that implement channels in Pygolang from Python to C level:

https://lab.nexedi.com/kirr/pygolang/commit/69db91bf
https://lab.nexedi.com/kirr/pygolang/commit/3b241983?expand_all_diffs=1

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue38106] Race in PyThread_release_lock - can lead to memory corruption and deadlock

[Kristján Valur Jónsson]

Kristján Valur Jónsson  added the comment:

Interesting.  Yet another reason to always do condition signalling with the 
lock held, such as is good practice to avoid race conditions.  That's the whole 
point of condition variables.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue38106] Race in PyThread_release_lock - can lead to memory corruption and deadlock

[Armin Rigo]

Armin Rigo  added the comment:

I may be wrong, but I believe that the bug requires using the C API (not just 
pure Python code).  This is because Python-level lock objects have their own 
lifetime, and should never be freed while another thread is in 
PyThread_release_lock() with them.

Nevertheless, the example shows that using this C API "correctly" is very hard. 
 Most direct users of the C API could run into the same problem in theory.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue38106] Race in PyThread_release_lock - can lead to memory corruption and deadlock

[Guido van Rossum]

Change by Guido van Rossum :


--
assignee:  -> benjamin.peterson
nosy:  -gvanrossum
title: Race in PyThread_release_lock - can lead to MEMORY CORRUPTION and 
DEADLOCK -> Race in PyThread_release_lock - can lead to memory corruption and 
deadlock

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue38106] Race in PyThread_release_lock - can lead to MEMORY CORRUPTION and DEADLOCK

[Kirill Smelkov]

New submission from Kirill Smelkov :

Hello up there. I believe I've discovered a race in PyThread_release_lock on
Python2.7 that, on systems where POSIX semaphores are not available and Python
locks are implemented with mutexes and condition variables, can lead to MEMORY
CORRUPTION and DEADLOCK. The particular system I've discovered the bug on is
macOS Mojave 10.14.6.

The bug is already fixed on Python3 and the fix for Python2 is easy:

git cherry-pick 187aa545165d

Thanks beforehand,
Kirill


Bug description

( Please see attached pylock_bug.pyx for the program that triggers the bug for 
real. )

On Darwin, even though this is considered as POSIX, Python uses
mutex+condition variable to implement its lock, and, as of 20190828, Py2.7
implementation, even though similar issue was fixed for Py3 in 2012, contains
synchronization bug: the condition is signalled after mutex unlock while the
correct protocol is to signal condition from under mutex:

  
https://github.com/python/cpython/blob/v2.7.16-127-g0229b56d8c0/Python/thread_pthread.h#L486-L506
  https://github.com/python/cpython/commit/187aa545165d (py3 fix)

PyPy has the same bug for both pypy2 and pypy3:

  
https://bitbucket.org/pypy/pypy/src/578667b3fef9/rpython/translator/c/src/thread_pthread.c#lines-443:465
  
https://bitbucket.org/pypy/pypy/src/5b42890d48c3/rpython/translator/c/src/thread_pthread.c#lines-443:465

Signalling condition outside of corresponding mutex is considered OK by
POSIX, but in Python context it can lead to at least memory corruption if we
consider the whole lifetime of python level lock. For example the following
logical scenario:

  T1  T2

  sema = Lock()
  sema.acquire()

  sema.release()

  sema.acquire()
  free(sema)

  ...


can translate to the next C-level calls:

  T1  T2

  # sema = Lock()
  sema = malloc(...)
  sema.locked = 0
  pthread_mutex_init()
  pthread_cond_init (_released)

  # sema.acquire()
  pthread_mutex_lock()
  # sees sema.locked == 0
  sema.locked = 1
  pthread_mutex_unlock()


  # sema.release()
  pthread_mutex_lock()
  sema.locked = 0
  pthread_mutex_unlock()

  # OS scheduler gets in and relinquishes control from T2
  # to another process
  ...

  # second sema.acquire()
  pthread_mutex_lock()
  # sees sema.locked == 0
  sema.locked = 1
  pthread_mutex_unlock()

  # free(sema)
  pthread_mutex_destroy()
  pthread_cond_destroy (_released)
  free(sema)


  # ...
  e.g. malloc() which returns memory where sema was

  ...
  # OS scheduler returns control to T2
  # sema.release() continues
  #
  # BUT sema was already freed and writing to anywhere
  # inside sema block CORRUPTS MEMORY. In particular if
  # _another_ python-level lock was allocated where sema
  # block was, writing into the memory can have effect on
  # further synchronization correctness and in particular
  # lead to deadlock on lock that was next allocated.
  
pthread_cond_signal(_released)

Note that T2.pthread_cond_signal(_released) CORRUPTS MEMORY as it
is called when sema memory was already freed and is potentially
reallocated for another object.

The fix is to move pthread_cond_signal to be done under corresponding mutex:

  # sema.release()
  pthread_mutex_lock()
  sema.locked = 0
  pthread_cond_signal(_released)
  pthread_mutex_unlock()

by cherry-picking commit 187aa545165d ("Signal condition variables with the
mutex held. Destroy condition variables before their mutexes").


Bug history

The bug was there since 1994 - since at least [1]. It was discussed in 2001
with original code author[2], but the code was still considered to be
race-free. In 2010 the place where pthread_cond_signal should be - before or
after pthread_mutex_unlock - was discussed with the rationale to avoid
threads bouncing[3,4,5], and in 2012 pthread_cond_signal was moved to be
called from under mutex, but only for CPython3[6,7].

In 2019 the bug was (re-)discovered while testing Pygolang[8] on macOS with
CPython2 and PyPy2 and PyPy3.

[1] https://github.com/python/cpython/commit/2c8cb9f3d240
[2] https://bugs.python.org/issue433625
[3] https://bugs.python.org/issue8299#msg103224
[4] https://bugs.python.org/issue8410#msg103313
[5] https://bugs.python.org/issue8411#msg113301
[6] https://bugs.python.org/issue15038#msg163187
[7] https://github.com/python/cpython/commit/187aa545165d
[8]