The python logging module uses a lock to surround many operations, in 
particular.  This causes deadlocks in programs that use logging, fork 
and threading simultaneously.

1) spawn one or more threads in your program
2) have at least one of those threads make logging calls that will be 
emitted.
3) have your main thread or another thread use os.fork() to run some 
python code in a child process.
4) If the fork happened while one of your threads was within the 
logging.Handler.handle() critical section (or anywhere else where 
handler.lock is acquired), your child process will deadlock as soon as 
it tries to log anything.  It inherited a held lock.

The deadlock is more likely to happen on a highly loaded system which 
tends to widen the deadlock opportunity window due to context switching.

A demo of the problem simplified into one file is attached.

The Python standard library should not be the cause of these deadlocks.  
We need a way for all standard library locks to be cleaned up when 
forking.  By doing one of the following:

A) acquire all locks before forking, release them immediately after.
B) forceably release all standard library locks after forking in the 
child process.

Code was added to call some cleanups after forking in 
http://bugs.python.org/issue874900 but there are more things that also 
need this same sort of cleanup (logging for example).

Rather than having to manually add after fork code hooks into every file 
in the standard library that uses locks, a more general solution to 
track and manage locks across fork would be a good idea.

I've started a project to patch this and similar messes up for Python
2.4 and later here:

 http://code.google.com/p/python-atfork/

I'd like to take ideas or implementations from that when possible for
future use in the python standard library.

issue 6923 has been opened to provide a C API for an atfork mechanism for 
use by extension modules.

Rather than having a kind of global module registry, locks could keep
track of what was the last PID, and reinitialize themselves if it changed.
This is assuming getpid() is fast :-)

> Antoine Pitrou <pitrou@free.fr> added the comment:
>
> Rather than having a kind of global module registry, locks could keep
> track of what was the last PID, and reinitialize themselves if it changed.
> This is assuming getpid() is fast :-)

Locks can't blindly release themselves because they find themselves
running in another process.

If anything if a lock is held and finds itself running in a new
process any attempt to use the lock should raise an exception so that
the bug is noticed.

I'm not sure a PID check is good enough.  old linux using linuxthreads
had a different pid for every thread, current linux with NPTL is more
like other oses with the same pid for all threads.

I was suggesting "reinitialize", rather than "release". That is, create
a new lock (mutex, semaphore, etc.) and let the old one die (or occupy
some tiny bit of memory).

no need for that.  the problem is that they're held by a thread that
does not exist in the newly forked child process so they will never be
released in the new process.

example: if you fork while another thread is in the middle of logging
something and then try to log something yourself in the child, your
child process will deadlock on the logging module's lock.

locks are not shared between processes so reinitializing them with a new
object wouldn't do anything.

I'm not sure that releasing the mutex is enough, it can still lead to a segfault, as is probably the case in this issue :
http://bugs.python.org/issue11148

Quoting pthread_atfork man page :

To understand the purpose of pthread_atfork, recall that fork duplicates the whole memory space, including mutexes in their current locking state, but only the calling thread: other threads are not running in the child process. The mutexes are not usable after the fork and must be initialized with pthread_mutex_init in the child process. This is a limitation of the current implementation and might or might not be present in future versions.

To avoid this, install handlers with pthread_atfork as follows: have the prepare handler lock the mutexes (in locking order), and the parent handler unlock the mutexes. The child handler should reset the mutexes using pthread_mutex_init, as well as any other synchronization objects such as condition variables.

Locking the global mutexes before the fork ensures that all other threads are locked out of the critical regions of code protected by those mutexes. Thus when fork takes a snapshot of the parent's address space, that snapshot will copy valid, stable data. Resetting the synchronization objects in the child process will ensure they are properly cleansed of any artifacts from the threading subsystem of the parent process. For example, a mutex may inherit a wait queue of threads waiting for the lock; this wait queue makes no sense in the child process. Initializing the mutex takes care of this.

pthread_atfork might be worth looking into

fwiw http://bugs.python.org/issue6643 recently fixed on issue where a mutex was being closed instead of reinitialized after a fork.  more are likely needed.

Are you suggesting to use pthread_atfork to call pthread_mutex_init on all mutexes created by Python in the child after a fork?  I'll have to ponder that some more.  Given the mutexes are all useless post fork it does not strike me as a bad idea.

> Are you suggesting to use pthread_atfork to call pthread_mutex_init on > all mutexes created by Python in the child after a fork?  I'll have to > ponder that some more.  Given the mutexes are all useless post fork it > does not strike me as a bad idea.

I don't really understand. It's quite similar to the idea you shot down in msg94135. Or am I missing something?

Yeah, I'm trying to figure out what I was thinking then or if I was just plain wrong. :)

I was clearly wrong about a release being done in the child being the right thing to do (issue6643 proved that, the state held by a lock is not usable to another process on all platforms such that release even works).

Part of it looks like I wanted a way to detect it was happening as any lock that is held during a fork indicates a _potential_ bug (the lock wasn't registered anywhere to be released before the fork) but not everything needs to care about that.

> I was clearly wrong about a release being done in the child being the right thing to do (issue6643 proved that, the state held by a lock is not usable to another process on all platforms such that release even works).

Yeah, apparently OS-X is one of them, the reporter in #11148 is
experiencing segfaults under OS-X.

> Are you suggesting to use pthread_atfork to call pthread_mutex_init on all mutexes created by Python in the child after a fork?  I'll have to ponder that some more.  Given the mutexes are all useless post fork it does not strike me as a bad idea.

Yes, that's what I was thinking. Instead of scattering the
lock-reclaiming code all over the place, try to use a more standard
API specifically designed with that in mind.
Note the base issue is that we're authorizing things which are
forbidden : in a multi-threaded process, only sync-safe calls are
authorized between fork and exec.

2011/2/10 Gregory P. Smith <report@bugs.python.org>:
>
> Gregory P. Smith <greg@krypto.org> added the comment:
>
> Yeah, I'm trying to figure out what I was thinking then or if I was just plain wrong. :)
>
> I was clearly wrong about a release being done in the child being the right thing to do (issue6643 proved that, the state held by a lock is not usable to another process on all platforms such that release even works).
>
> Part of it looks like I wanted a way to detect it was happening as any lock that is held during a fork indicates a _potential_ bug (the lock wasn't registered anywhere to be released before the fork) but not everything needs to care about that.
>
> ----------
> versions: +Python 3.3
>
> _______________________________________
> Python tracker <report@bugs.python.org>
> <http://bugs.python.org/issue6721>
> _______________________________________
>

I encountered this issue while debugging some multiprocessing code; fork() would be called from one thread while sys.stdout was in use in another thread (simply because of a couple of debugging statements). As a result the IO lock would be already "taken" in the child process and any operation on sys.stdout would deadlock.

This is definitely something that can happen more easily than I thought.

Here is a patch with tests for the issue (some of which fail of course).
Do we agree that these tests are right?

Those tests make sense to me.

# A lock taken from the current thread should stay taken in the
        # child process.

Note that I'm not sure of how to implement this.
After a fork, even releasing the lock can be unsafe, it must be re-initialized, see following comment in glibc's malloc implementation:
/* In NPTL, unlocking a mutex in the child process after a
   fork() is currently unsafe, whereas re-initializing it is safe and
   does not leak resources.  Therefore, a special atfork handler is
   installed for the child. */

Note that this means that even the current code allocating new locks after fork (in Lib/threading.py, _after_fork and _reset_internal_locks) is unsafe, because the old locks will be deallocated, and the lock deallocation tries to acquire and release the lock before destroying it (in issue #11148 the OP experienced a segfault on OS-X when locking a mutex, but I'm not sure of the exact context).

Also, this would imply keeping track of the thread currently owning the lock, and doesn't match the typical pthread_atfork idiom (acquire locks just before fork, release just after in parent and child, or just reinit them in the child process)

Finally, IMHO, forking while holding a lock and expecting it to be usable after fork doesn't make much sense, since a lock is acquired by a thread, and this threads doesn't exist in the child process. It's explicitely described as "undefined" by POSIX, see http://pubs.opengroup.org/onlinepubs/007908799/xsh/sem_init.html :
"""
The use of the semaphore by threads other than those created in the same process is undefined.
"""

So I'm not sure whether it's feasable/wise to provide such a guarantee.

> Also, this would imply keeping track of the thread currently owning
> the lock,

Yes, we would need to keep track of the thread id and process id inside
the lock. We also need a global variable of the main thread id after
fork, and a per-lock "taken" flag.

Synopsis:

    def _reinit_if_needed(self):
        # Call this before each acquire() or release()
        if self.pid != getpid():
            sem_init(self.sem, 0, 1)
            if self.taken:
                if self.tid == main_thread_id_after_fork:
                    # Lock was taken in forked thread, re-take it
                    sem_wait(self.sem)
                else:
                    # It's now released
                    self.taken = False
            self.pid = getpid()
            self.tid = current_thread_id()

> and doesn't match the typical pthread_atfork idiom (acquire locks just
> before fork, release just after in parent and child, or just reinit
> them in the child process)

Well, I fail to understand how that idiom can help us. We're not a
self-contained application, we're a whole programming language.
Calling fork() only when no lock is held is unworkable (for example, we
use locks around buffered I/O objects).

> Yes, we would need to keep track of the thread id and process id inside
> the lock. We also need a global variable of the main thread id after
> fork, and a per-lock "taken" flag.
>
> Synopsis:
>
>    def _reinit_if_needed(self):
>        # Call this before each acquire() or release()
>        if self.pid != getpid():
>            sem_init(self.sem, 0, 1)
>            if self.taken:
>                if self.tid == main_thread_id_after_fork:
>                    # Lock was taken in forked thread, re-take it
>                    sem_wait(self.sem)
>                else:
>                    # It's now released
>                    self.taken = False
>            self.pid = getpid()
>            self.tid = current_thread_id()
>

A couple remarks:
- with linuxthreads, different threads within the same process have
the same PID - it may be true for other implementations - so this
would lead to spurious reinitializations
- what's current_thread_id ? If it's thread_get_ident (pthread_self),
since TID is not guaranteed to be inherited across fork, this won't
work
- calling getpid at every acquire/release is expensive, even though
it's a trivial syscall (it'll have to measured though)
- imagine the following happens:

P1

lock.acquire()
fork()    ->       P2
                       start_new_thread T2
                       T1           T2
                                        lock.acquire()

The acquisition of lock by T2 will cause lock's reinitialization: what
happens to the lock wait queue ? who owns the lock ?
That why I don't think we can delay the reinitialization of locks, but
I could be wrong.

> Well, I fail to understand how that idiom can help us. We're not a
> self-contained application, we're a whole programming language.
> Calling fork() only when no lock is held is unworkable (for example, we
> use locks around buffered I/O objects).

Yes, but in that case, you don't have to reacquire the locks after fork.
In the deadlock you experienced above, the thread that forked wasn't
the one in the I/O code, so the corresponding lock can be
re-initialized anyway, since the thread in the I/O code at that time
won't exist after fork.
And it's true with every lock in the library code: they're only held
in short critical sections (typically acquired when entering a
function and released when leaving), and since it's not the threads
inside those libraries that fork, all those locks can simply be
reinitialized on fork, without having the reacquire them.

Oops, for liunxthreads, you should of course read "different PIDs", not "same PID".

> - what's current_thread_id ? If it's thread_get_ident (pthread_self),
> since TID is not guaranteed to be inherited across fork, this won't
> work

Ouch, then the approach I'm proposing is probably doomed.

> And it's true with every lock in the library code: they're only held
> in short critical sections (typically acquired when entering a
> function and released when leaving), and since it's not the threads
> inside those libraries that fork, all those locks can simply be
> reinitialized on fork, without having the reacquire them.

Well, this means indeed that *some* locks can be handled, but not all of
them and not automatically, right?
Also, how would you propose they be dealt with in practice? How do they
get registered, and how does the reinitialization happen?

(do note that library code can call arbitrary third-party code, by the
way: for example through encodings in the text I/O layer)

>> - what's current_thread_id ? If it's thread_get_ident (pthread_self),
>> since TID is not guaranteed to be inherited across fork, this won't
>> work
>
> Ouch, then the approach I'm proposing is probably doomed.
>

Well, it works on Linux with NPTL, but I'm not sure at all it holds
for other implementations (pthread_t it's only meaningful within the
same process).
But I'm not sure it's really the "killer" point: PID with linuxthreads
and lock being acquired by a second thread before the main thread
releases it in the child process also look like serious problems.

> Well, this means indeed that *some* locks can be handled, but not all of
> them and not automatically, right?
> Also, how would you propose they be dealt with in practice? How do they
> get registered, and how does the reinitialization happen?

When a lock object is allocated in Modules/threadmodule.c
(PyThread_allocate_lock/newlockobject), add the underlying lock
(self->lock_lock) to a linked list (since it's called with the GIL
held, we don't need to protect the linked list from concurrent
access). Each thread implementation (thread_pthread.h, thread_nt.h)
would provide a new PyThread_reinit_lock function that would do the
right thing (pthread_mutex_destroy/init, sem_destroy/init, etc).
Modules/threadmodule.c would provide a new PyThread_ReInitLocks that
would walk through the linked list and call PyThread_reinit_lock for
each lock.
PyOS_AfterFork would call this PyThread_ReInitLocks right after fork.
This would have the advantage of being consistent with what's already
done to reinit the TLS key and the import lock. So, we guarantee to be
in a consistent and usable state when PyOS_AfterFork returns. Also,
it's somewhat simpler because we're sure that at that point only one
thread is running (once again, no need to protect the linked-list
walk).
I don't think that the performance impact would be noticable (I know
it's O(N) where N is the number of locks), and contrarily to the
automatic approach, this wouldn't penalize every acquire/release.
Of course, this would solve the problem of threading's module locks,
so PyEval_ReInitThreads could be removed, along with threading.py's
_after_fork and _reset_internal_locks.
In short, this would reset every lock held so that they're usable in
the child process, even locks allocated e.g. from
Modules/_io/bufferedio.c.
But this wouldn't allow a lock's state to be inherited across fork for
the main thread (but like I said, I don't think that this makes much
sense anyway, and to my knowledge no implementation makes such a
guarantee - and definitely not POSIX).

Please disregard my comment on PyEval_ReInitThreads and _after_fork:
it will of course still be necessary, because it does much more than
just reinitializing locks (e.g. stop threads).
Also, note that both approaches don't handle synchronization
primitives other  than bare Lock and RLock. For example, Condition and
Event used in the threading module wouldn't be reset automatically:
that's maybe something that could be handled by Gregory's atfork
mechanism.

Thanks for the explanations. This sounds like an interesting path.

> Each thread implementation (thread_pthread.h, thread_nt.h)
> would provide a new PyThread_reinit_lock function that would do the
> right thing (pthread_mutex_destroy/init, sem_destroy/init, etc).
> Modules/threadmodule.c would provide a new PyThread_ReInitLocks that
> would walk through the linked list and call PyThread_reinit_lock for
> each lock.

Actually, I think the issue is POSIX-specific: Windows has no fork(),
and we don't care about other platforms anymore (they are, are being, or
will be soon deprecated).
It means only the POSIX implementation needs to register its locks in a
linked list.

> But this wouldn't allow a lock's state to be inherited across fork for
> the main thread (but like I said, I don't think that this makes much
> sense anyway, and to my knowledge no implementation makes such a
> guarantee - and definitely not POSIX).

Well, the big difference between Python locks and POSIX mutexes is that
Python locks can be released from another thread. They're a kind of
trivial semaphore really, and this makes them usable for other purpose
than mutual exclusion (you can e.g. use a lock as a simple event by
blocking on a second acquire() until another thread calls release()).

But even though we might not be "fixing" arbitrary Python code
automatically, fixing the interpreter's internal locks (especially the
IO locks) would be great already.

(we could also imagine that the creator of the lock decides whether it
should get reinitialized after fork)

Hi,

There seem to be two alternatives for atfork handlers:
1) acquire locks during prepare phase and unlock them in parent and child after fork.
2) reset library to some consistent state in child after fork.

http://pubs.opengroup.org/onlinepubs/009695399/functions/pthread_atfork.html

Option (2) makes sense but is probably not always applicable.
Option (1) depends on being able to acquire locks in locking order, but how can we determine correct locking order across libraries?

Initializing locks in child after fork without acquiring them before the fork may result in corrupted program state and so is probably not a good idea.

On a positive note, if I understand correctly, Python signal handler functions are actually run in the regular interpreter loop (as pending calls) after the signal has been handled and so os.fork() atfork handlers will not be restricted to async-signal-safe operations (since a Python fork is never done in a signal handler).

http://pubs.opengroup.org/onlinepubs/009695399/functions/xsh_chap02_04.html
http://pubs.opengroup.org/onlinepubs/009695399/functions/fork.html
"It is therefore undefined for the fork handlers to execute functions that are not async-signal-safe when fork() is called from a signal handler."

Opinion by Butenhof who was involved in the standardization effort of POSIX threads:
http://groups.google.com/group/comp.programming.threads/msg/3a43122820983fde

...so how can we establish correct (cross library) locking order during prepare stage?

Nir

@Nir Aides: *thanks* for this link:
   http://groups.google.com/group/comp.programming.threads/msg/3a43122820983fde
You made my day!

> ...so how can we establish correct (cross library) locking order
> during prepare stage?

That sounds like a lost battle, if it requires the libraries'
cooperation. I think resetting locks is the best we can do. It might not
work ok in all cases, but if it can handle simple cases (such as I/O and
logging locks), it is already very good.

> Hi,
>

Hello Nir,

> Option (2) makes sense but is probably not always applicable.
> Option (1) depends on being able to acquire locks in locking order, but how
> can we determine correct locking order across libraries?
>

There are indeed a couple problems with 1:
1) actually, releasing the mutex/semaphore from the child is not
guaranteed to be safe, see this comment from glibc's malloc:
/* In NPTL, unlocking a mutex in the child process after a
   fork() is currently unsafe, whereas re-initializing it is safe and
   does not leak resources.  Therefore, a special atfork handler is
   installed for the child. */
We could just destroy/reinit them, though.

2) acquiring locks just before fork is probably one of the best way to
deadlock (acquiring a lock we already hold, or acquiring a lock needed
by another thread before it releases its own lock). Apart from adding
dealock avoidance/recovery mechanisms - which would be far from
trivial - I don't see how we could solve this, given that each library
can use its own locks, not counting the user-created ones

3) there's another special lock we must take into account, the GIL:
contrarily to a typical C program, we can't have the thread forking
blindly try to acquire all locks just before fork, because since we
hold the GIL, other threads won't be able to proceed (unless of course
they're in a section where they don't run without the GIL held).

So, we would have to:
- release the GIL
- acquire all locks in the correct order
- re-acquire the GIL
- fork
- reinit all locks after fork

I think this is going to be very complicated.

4) Python locks differ from usual mutexes/semaphores in that they can
be held for quite some time (for example while performing I/O). Thus,
acquiring all the locks could take a long time, and users might get
irritated if fork takes 2 seconds to complete.

5) Finally, there's a fundamental problem with this approach, because
Python locks can be released by a thread other than the one that owns
it.
Imagine this happens:

T1                         T2
                          lock.acquire()
                          (do something without releasing lock)
fork()
lock.release()

This is perfectly valid with the current lock implementation (for
example, it can be used to implement a rendez-vous point so that T2
doesn't start processing before T1 forked worker processes, or
whatever).
But if T1 tries to acquire lock (held by T2) before fork, then it will
deadlock, since it will never be release by T2.

For all those reasons, I don't think that this approach is reasonable,
but I could be wrong :-)

> Initializing locks in child after fork without acquiring them before the
> fork may result in corrupted program state and so is probably not a good
> idea.

Yes, but in practise, I think that this shouldn't be too much of a
problem. Also note that you can very well have the same type of
problem with sections not protected explicitely by locks: for example,
if you have a thread working exclusively on an object (maybe part of a
threadpool), a fork can very well happen while the object is in an
inconsistent state. Acquiring locks before fork won't help that.
But I think this should eventually be addressed, maybe by specific
atfork handlers.

> On a positive note, if I understand correctly, Python signal handler
> functions are actually run in the regular interpreter loop (as pending
> calls) after the signal has been handled and so os.fork() atfork handlers
> will not be restricted to async-signal-safe operations (since a Python fork
> is never done in a signal handler).

That's correct.

In short, I think that we could first try to avoid common deadlocks by
just resetting locks in the child process. This is not panacea, but
this should solve the vast majority of deadlocks, and would open the
door to potential future refinements using atfork-like handlers.

Attached is a first draft for a such patch (with tests).
Synopsis:
- when a PyThread_type_lock is created, it's added to a linked-list,
when it's deleted, it's removed from the linked list
- PyOS_AfterFork() calls PyThread_ReinitLocks() which calls
PyThread_reinit_lock() for each lock in the linked list
- PyThread_reinit_lock() does the right thing (i.e. sem_destroy/init
for USE_SEMAPHORES and pthread_(mutex|cond)_destroy/init for emulated
semaphores).

Notes:
- since it's only applicable to POSIX (since other Unix thread
implementations will be dropped), I've only defined a
PyThread_ReinitLocks inside Python/thread_pthread.h, so it won't build
on other platforms. How should I proceed: like PyThread_ReInitTLS(),
add a stub function to all Python/thread_xxx.h, or guard the call to
PyThread_ReinitLocks() with #ifdef _POSIX_THREADS ?
- I'm not sure of how to handle sem_init/etc failures in the reinit
code: for now I just ignore this possibility, like what's done for the
import lock reset
- insertions/removals from the linked list are not protected from
concurrent access because I assume that locks are created/deleted with
the GIL held: is that a reasonable assumption, or should I add a mutex
to protect those accesses?

This fixes common deadlocks with threading.Lock, and
PyThread_type_lock (used for example by I/O code).

@ Charles-François Natali <report@bugs.python.org> wrote (2011-05-13 13:24+0200):
> I happily posted a reinit patch

I must say in advance that we have implemented our own thread
support 2003-2005 and i'm thus lucky not to need to use anything
else ever since.  So.  And of course i have no overview about
Python.  But i looked and saw no errors in the default path and
the tests run without errors.
Then i started to try your semaphore path which is a bit
problematic because Mac OS X doesn't offer anon sems ;).
(
By the way, in PyThread_acquire_lock_timed() these lines

    if (microseconds > 0)
                MICROSECONDS_TO_TIMESPEC(microseconds, ts);

result in these compiler warnings.

python/thread_pthread.h: In function ‘PyThread_acquire_lock_timed’:
Python/thread_pthread.h:424: warning: ‘ts.tv_sec’ may be used
uninitialized in this function
Python/thread_pthread.h:424: warning: ‘ts.tv_nsec’ may be used
uninitialized in this function
)

#ifdef USE_SEMAPHORES
#define broken_sem_init broken_sem_init
static int broken_sem_init(sem_t **sem, int shared, unsigned int value) {
    int ret;
    auto char buffer[32];
    static long counter = 3000;
    sprintf(buffer, "%016ld", ++counter);
    *sem = sem_open(buffer, O_CREAT, (mode_t)0600, (unsigned int)value);
    ret = (*sem == SEM_FAILED) ? -1 : 0;
    //printf("BROKEN_SEM_INIT WILL RETURN %d (value=%u)\n", ret,value);
    return ret;
}
static int sem_timedwait(sem_t *sem, struct timespec *ts) {
    int success = -1, iters = 1000;
    struct timespec now, wait;
    printf("STARTING LOOP\n");
    for (;;) {
        if (sem_trywait(sem) == 0) {
            printf("TRYWAIT OK\n");
            success = 0;
            break;
        }
        wait.tv_sec = 0, wait.tv_nsec = 200 * 1000;
        //printf("DOWN "); fflush(stdout);
        nanosleep(&wait, NULL);
        MICROSECONDS_TO_TIMESPEC(0, now);
        //printf("WOKE UP NOW=%ld:%ld END=%ld:%ld\n", now.tv_sec,now.tv_nsec, ts->tv_sec,ts->tv_nsec);
        if (now.tv_sec > ts->tv_sec ||
            (now.tv_sec == ts->tv_sec && now.tv_nsec >= ts->tv_nsec))
            break;
        if (--iters < 0) {
            printf("BREAKING OFF LOOP, 1000 iterations\n");
            errno = ETIMEDOUT;
            break;
        }
    }
    return success;
}
#define sem_destroy sem_close

typedef struct _pthread_lock {
    sem_t               *sem;
    struct _pthread_lock*next;
    sem_t               sem_buf;
} pthread_lock;
#endif

plus all the changes the struct change implies, say.
Yes it's silly, but i wanted to test.  And this is the result:

== CPython 3.3a0 (default:804abc2c60de+, May 14 2011, 01:09:53) [GCC 4.2.1 (Apple Inc. build 5666) (dot 3)]
==   Darwin-10.7.0-i386-64bit little-endian
==   /Users/steffen/src/cpython/build/test_python_19230
Testing with flags: sys.flags(debug=0, inspect=0, interactive=0, optimize=0, dont_write_bytecode=0, no_user_site=0, no_site=0, ignore_environment=1, verbose=0, bytes_warning=0, quiet=0)
Using random seed 1362049
[1/1] test_threading
STARTING LOOP
test_acquire_contended (test.test_threading.LockTests) ... ok
test_acquire_destroy (test.test_threading.LockTests) ... ok
test_acquire_release (test.test_threading.LockTests) ... ok
test_constructor (test.test_threading.LockTests) ... ok
test_different_thread (test.test_threading.LockTests) ... ok
test_reacquire (test.test_threading.LockTests) ... ok
test_state_after_timeout (test.test_threading.LockTests) ... ok
test_thread_leak (test.test_threading.LockTests) ... ok
test_timeout (test.test_threading.LockTests) ... STARTING LOOP
TRYWAIT OK
FAIL
test_try_acquire (test.test_threading.LockTests) ... ok
test_try_acquire_contended (test.test_threading.LockTests) ... ok
test_with (test.test_threading.LockTests) ... ok
test__is_owned (test.test_threading.PyRLockTests) ... ok
test_acquire_contended (test.test_threading.PyRLockTests) ... ok
test_acquire_destroy (test.test_threading.PyRLockTests) ... ok
test_acquire_release (test.test_threading.PyRLockTests) ... ok
test_constructor (test.test_threading.PyRLockTests) ... ok
test_different_thread (test.test_threading.PyRLockTests) ... ok
test_reacquire (test.test_threading.PyRLockTests) ... ok
test_release_unacquired (test.test_threading.PyRLockTests) ... ok
test_thread_leak (test.test_threading.PyRLockTests) ... ok
test_timeout (test.test_threading.PyRLockTests) ... STARTING LOOP
TRYWAIT OK
FAIL
test_try_acquire (test.test_threading.PyRLockTests) ... ok
test_try_acquire_contended (test.test_threading.PyRLockTests) ... ok
test_with (test.test_threading.PyRLockTests) ... ok
test__is_owned (test.test_threading.CRLockTests) ... ok
test_acquire_contended (test.test_threading.CRLockTests) ... ok
test_acquire_destroy (test.test_threading.CRLockTests) ... ok
test_acquire_release (test.test_threading.CRLockTests) ... ok
test_constructor (test.test_threading.CRLockTests) ... ok
test_different_thread (test.test_threading.CRLockTests) ... ok
test_reacquire (test.test_threading.CRLockTests) ... ok
test_release_unacquired (test.test_threading.CRLockTests) ... ok
test_thread_leak (test.test_threading.CRLockTests) ... BREAKING OFF LOOP, 1000 iterations
Timeout (1:00:00)!
Thread 0x00007fff70677ca0:
  File "/Users/steffen/src/cpython/Lib/test/lock_tests.py", line 17 in _wait
  File "/Users/steffen/src/cpython/Lib/test/lock_tests.py", line 52 in wait_for_finished
  File "/Users/steffen/src/cpython/Lib/test/lock_tests.py", line 152 in test_thread_leak
  File "/Users/steffen/src/cpython/Lib/unittest/case.py", line 407 in _executeTestPart
  File "/Users/steffen/src/cpython/Lib/unittest/case.py", line 462 in run
  File "/Users/steffen/src/cpython/Lib/unittest/case.py", line 514 in __call__
  File "/Users/steffen/src/cpython/Lib/unittest/suite.py", line 105 in run
  File "/Users/steffen/src/cpython/Lib/unittest/suite.py", line 67 in __call__
  File "/Users/steffen/src/cpython/Lib/unittest/suite.py", line 105 in run
  File "/Users/steffen/src/cpython/Lib/unittest/suite.py", line 67 in __call__
  File "/Users/steffen/src/cpython/Lib/unittest/runner.py", line 168 in run
  File "/Users/steffen/src/cpython/Lib/test/support.py", line 1187 in _run_suite
  File "/Users/steffen/src/cpython/Lib/test/support.py", line 1213 in run_unittest
  File "/Users/steffen/src/cpython/Lib/test/test_threading.py", line 748 in test_main
  File "/Users/steffen/src/cpython/Lib/test/regrtest.py", line 1044 in runtest_inner
  File "/Users/steffen/src/cpython/Lib/test/regrtest.py", line 838 in runtest
  File "/Users/steffen/src/cpython/Lib/test/regrtest.py", line 662 in main
  File "/Users/steffen/src/cpython/Lib/test/__main__.py", line 13 in <module>
  File "/Users/steffen/src/cpython/Lib/runpy.py", line 73 in _run_code
  File "/Users/steffen/src/cpython/Lib/runpy.py", line 160 in _run_module_as_main

Hope that helps a bit. Bâillement.

Hello Steffen,

First, thanks for testing this on OS-X: I only have access to Linux
systems (I tested both the semaphore and the emulated semaphore
paths).

If I understand correctly, the patch works fine with the default build
option on OS-X.
Then, you're saying that OS-X doesn't have POSIX unnamed semaphores:
this means that the default build uses the mutex+condition variable
version. Am I correct?
But I don't understand the last part of your message.
Do you mean that you implemented your own version of the semaphore
path using named semaphores, and that it fails with my patch (well,
your adapted version of it) ?
If yes, then you'll understand that I can't comment on this, since
it's not my code :-)
But after a quick look at the code you posted, I think that your
acquire code is broken:
sem_timedwait(timeout) if not the same as trying sem_trywait multiple
times until timeout expires: in case of contention, this will fail.

I think that generally it is better to deadlock than corrupt data.

> 2) acquiring locks just before fork is probably one of the best way to
> deadlock (acquiring a lock we already hold, or acquiring a lock needed
> by another thread before it releases its own lock). Apart from adding
> dealock avoidance/recovery mechanisms - which would be far from
> trivial - I don't see how we could solve this, given that each library
> can use its own locks, not counting the user-created ones

a) We know the correct locking order in Python's std libraries so the problem there is kind of solved.

b) We can put the burden of other locks on application developers and since currently no one registers atfork handlers, there is no problem there yet.

> 4) Python locks differ from usual mutexes/semaphores in that they can
> be held for quite some time (for example while performing I/O). Thus,
> acquiring all the locks could take a long time, and users might get
> irritated if fork takes 2 seconds to complete.

We only need a prepare handler to acquire locks that protect data from corruption.

A lock synchronizing IO which is held for long periods may possibly be initialized in child without being acquired in a prepare handler; for example, a lock serializing logging messages.

In other cases or in general an atfork handler may reset or reinitialize a library without acquiring locks in a prepare handler.

> 5) Finally, there's a fundamental problem with this approach, because
> Python locks can be released by a thread other than the one that owns
> it.
> Imagine this happens:
>
> T1                         T2
>                          lock.acquire()
>                          (do something without releasing lock)
> fork()
> lock.release()
>
> This is perfectly valid with the current lock implementation (for
> example, it can be used to implement a rendez-vous point so that T2
> doesn't start processing before T1 forked worker processes, or
> whatever).
> But if T1 tries to acquire lock (held by T2) before fork, then it will
> deadlock, since it will never be release by T2.

I think we do not need to acquire rendezvous locks in a prepare handler.

> > Initializing locks in child after fork without acquiring them before the
> > fork may result in corrupted program state and so is probably not a good
> > idea.
>
> Yes, but in practise, I think that this shouldn't be too much of a
> problem. Also note that you can very well have the same type of
> problem with sections not protected explicitely by locks: for example,
> if you have a thread working exclusively on an object (maybe part of a
> threadpool), a fork can very well happen while the object is in an
> inconsistent state. Acquiring locks before fork won't help that.

I think a worker thread that works exclusively on an object does not create the problem:
a) If the fork thread eventually needs to read the object then you need synchronization.
b) If the worker thread eventually writes data into file or DB then that operation will be completed at the parent process.

To summarize I think we should take the atfork path. An atfork handler does not need to acquire all locks, but only those required by library logic, which the handler is aware of, and as a bonus it can be used to do all sort of stuff such as cleaning up, reinitializing a library, etc...

> a) We know the correct locking order in Python's std libraries so the problem there is kind of solved.

I think that you're greatly under-estimating the complexity of lock ordering.
If we were just implementing a malloc implementation protected with a
single mutex, then yes, it would be simple.
But here, you have multiple libraries with each their own locks, locks
at the I/O layer, in the socket module (some name resolution libraries
are not thread-safe), and in many other places. And all those
interact.
For example, buffered I/O objects each have their own lock (Antoine,
correct me if I'm wrong).
It's a common cause of deadlock.
Now imagine I have a thread that logs information to a bz2 stream, so
that it's compressed on-the-fly. Sounds reasonable, no?
Well, the lock hierarchy is:

buffered stream lock
bz2-level lock
logging object I/O lock

Do you still think that getting the locking order right is easy?

Another example, with I/O locks (and if you're concerned with data
corruption, those are definitely the one you would want to handle with
atfork):
I have a thread blocking on a write (maybe the output pipe is full,
maybe it's a NFS file system and the server takes a long time to
respond, etc. Or maybe it's just waiting for someone to type something
on stdin.).
Another thread forks.
The atfork-handler will try to acquire the buffered I/O object's lock:
it won't succeed until the other threads finally manages to
write/read. It could take seconds, or forever.
And there are many other things that could go wrong, because
contrarily to a standalone and self-contained library, Python is made
of several components, at different layers, that can call each other
in an arbitrary order. Also, some locks can be held for arbitrarily
long.

That's why I still think that this can be fully handled by atfork handlers.

But don't get me wrong: like you, I think that we should definitely
have an atfork mechanism. I just think it won't be able to solve all
the issues, and that I can also bring its own set of troubles.

Concerning the risk of corruption (invariant broken), you're right.
But resetting the locks is the approach currently in use for the
threading module, and it seems to work reasonably well there.

Finally, I'd just like to insist on a point:
In a multi-threaded program, between fork and exec, the code must be
async-safe. This means that in theory, you can't call
pthread_mutex_release/pthread_mutex_destroy, fwrite, malloc, etc.
Period.
This means that in theory, we shouldn't be running Python code at all!
So if we really wanted to be safe, the only solution would be to
forbid fork() in a multi-threaded program.
Since it's not really a reasonable option, and that the underlying
platform (POSIX) doesn't allow to be safe either, I guess that the
only choice left is to provide a bet-try implementation, knowing
perfectly that there will always be some corner cases that can't be
handled.

@ Charles-François Natali wrote (2011-05-15 01:14+0200):
> So if we really wanted to be safe, the only solution would be to
> forbid fork() in a multi-threaded program.
> Since it's not really a reasonable option

But now - why this?  The only really acceptable thing if you have
control about what you are doing is the following:

class SMP::Process
    /*!
    * \brief Daemonize process.
    *[.]
    * \note
    * The implementation of this function is not trivial.
    * To avoid portability no-goes and other such problems,
    * you may \e not call this function after you have initialized
    * Thread::enableSMP(),
    * nor may there (have) be(en) Child objects,
    * nor may you have used an EventLoop!
    * I.e., the process has to be a single threaded, "synchronous" one.
    * [.]
    */
    pub static si32 daemonize(ui32 _daemon_flags=df_default);

namespace SMP::POSIX
    /*!
    * \brief \fn fork(2).
    *[.]
    * Be aware that this passes by all \SMP and Child related code,
    * i.e., this simply \e is the system-call.
    * Signal::resetAllSignalStates() and Child::killAll() are thus if
    * particular interest; thread handling is still entirely up to you.
    */
    pub static sir fork(void);

Which kind of programs cannot be written with this restriction?

Is it possible the following issue is related to this one?
http://bugs.python.org/issue10037 - "multiprocessing.pool processes started by worker handler stops working"

> Is it possible the following issue is related to this one?

It's hard to tell, the original report is rather vague.
But the comment about the usage of the maxtasksperchild argument reminds me of issue #10332 "Multiprocessing maxtasksperchild results in hang": basically, there's a race window in the Pool shutdown code where worker threads having completed their work can exit without being replaced.
So the connection with the current issue does not strike me, but you never know (that's the problem with those nasty race conditions ;-).

Concerning this issue, here's an updated patch.
I removed calls to pthread_mutex_destroy/pthread_condition_destroy/sem_destroy from the reinit functions: the reason is that I experienced a deadlock in test_concurrent_futures with the emulated semaphore code on Linux/NPTL inside pthread_condition_destroy: the new version strictly mimics what's done in glibc's malloc, and just calls pthrad_mutex_init and friends. It's safe, and shouldn't leak resources (and even if it does, it's way better than a deadlock).
I also placed the note on the interaction between locks and fork() at the top of Python/thread_pthread.h.

Steffen, can you explain in layman's terms?

On Sun, May 15, 2011 at 8:03 PM, Steffen Daode Nurpmeso <report@bugs.python.org> wrote:
>
> @ Charles-François Natali wrote (2011-05-15 01:14+0200):
>> So if we really wanted to be safe, the only solution would be to
>> forbid fork() in a multi-threaded program.
>> Since it's not really a reasonable option
>
> But now - why this?  The only really acceptable thing if you have
> control about what you are doing is the following:
>
> class SMP::Process
>    /*!
>    * \brief Daemonize process.
>    *[.]
>    * \note
>    * The implementation of this function is not trivial.
>    * To avoid portability no-goes and other such problems,
>    * you may \e not call this function after you have initialized
>    * Thread::enableSMP(),
>    * nor may there (have) be(en) Child objects,
>    * nor may you have used an EventLoop!
>    * I.e., the process has to be a single threaded, "synchronous" one.
>    * [.]
>    */
>    pub static si32 daemonize(ui32 _daemon_flags=df_default);
>
> namespace SMP::POSIX
>    /*!
>    * \brief \fn fork(2).
>    *[.]
>    * Be aware that this passes by all \SMP and Child related code,
>    * i.e., this simply \e is the system-call.
>    * Signal::resetAllSignalStates() and Child::killAll() are thus if
>    * particular interest; thread handling is still entirely up to you.
>    */
>    pub static sir fork(void);
>
> Which kind of programs cannot be written with this restriction?

@ Nir Aides wrote (2011-05-16 20:57+0200):
> Steffen, can you explain in layman's terms?

I am the layman here.
Charles-François has written a patch for Python which contradicted
his own proposal from msg135079, but he seems to have tested a lot
so that he then was even able to prove that his own proposal was
correct.  His new patch does implement that with a nice
introductional note.

He has also noticed that the only really safe solution is to
simply disallow multi-threading in programs which fork().  And
this either-or is exactly the conclusion we have taken and
implemented in our C++ library - which is not an embeddable
programming language that needs to integrate nicely in whatever
environment it is thrown into, but "even replaces main()".
And i don't know any application which cannot be implemented
regardless of fork()-or-threads instead of fork()-and-threads.
(You *can* have fork()+exec()-and-threads at any time!)

So what i tried to say is that it is extremely error-prone and
resource intensive to try to implement anything that tries to
achieve both.  I.e. on Solaris they do have a forkall() and it
seems they have atfork handlers for everything (and even document
that in the system manual).  atfork handlers for everything!!
And for what?  To implement a standart which is obviously
brain-dead because it is *impossible* to handle it - as your link
has shown this is even confessed by members of the committee.

And writing memory in the child causes page-faults.
That's all i wanted to say.
(Writing this mail required more than 20 minutes, the mentioned
one was out in less than one.  And it is much more meaningful
AFAIK.)

If there's agreement that the general problem is unsolvable (so fork and threads just don't get along with each other), what we could attempt is trying to limit the side effects in the standard library, so that fewest users as possible are affected by this problem.

For instance, having deadlocks just because of print statements sounds like a bad QoI that we could attempt to improve. Is there a reason while BufferedIO needs to hold its internal data-structure lock (used to make it thread-safe) while it's doing I/O and releasing the GIL? I would think that it's feasible to patch it so that its internal lock is only used to synchronize accesses to the internal data structures, but it is never held while I/O is performed (and thus the GIL is released -- at which point, if another threads forks, the problem appears).

> If there's agreement that the general problem is unsolvable (so fork and
> threads just don't get along with each other), what we could attempt is
> trying to limit the side effects in the standard library, so that fewest
> users as possible are affected by this problem.

Actually, I think Charles-François' suggested approach is a good one.

> For instance, having deadlocks just because of print statements sounds
> like a bad QoI that we could attempt to improve. Is there a reason while
> BufferedIO needs to hold its internal data-structure lock (used to make
> it thread-safe) while it's doing I/O and releasing the GIL? I would think
> that it's feasible to patch it so that its internal lock is only used to
> synchronize accesses to the internal data structures, but it is never
> held while I/O is performed (and thus the GIL is released -- at which
> point, if another threads forks, the problem appears).

Not really. Whether you update the internal structures depends on the
result of the I/O (so that e.g. two threads don't flush the same buffer
simultaneously).

Also, finer-grained locking is always a risky endeavour and we already have
a couple of bugs to fix in the current buffered I/O implementation :-/

The way I see it is that Charles-François' patch trades a possibility of a deadlock for a possibility of a child process running with inconsistent states due to forcibly reinitialized locks.

Personally, I would rather have an occasional deadlock than an occasional random crash.

I don't like increasing complexity with fine-grained locking either. While the general case is unsolvable what Giovanni proposed at least solves the specific case where only the basic IO code is involved after a fork. In hindsight the only real life use-case I can find that it would solve is doing an exec() right after a fork().

There are quite a few bugs in the tracker that seem to have this same root cause, so it appears the impossibility of cleanly handling threads and forks is not something people are generally aware of. Since I think we agree you can't just disable fork() when multiple threads are running, how about at least issuing a warning in that case? That would be a two-line change in threading.py.

> The way I see it is that Charles-François' patch trades a possibility of a deadlock for a possibility of a child process running with inconsistent states due to forcibly reinitialized locks.
>

Yeah, that's why I let this stale: that's really an unsolvable problem
in the general case. Don't mix fork() and threads, that's it.

> I don't like increasing complexity with fine-grained locking either. While the general case is unsolvable what Giovanni proposed at least solves the specific case where only the basic IO code is involved after a fork. In hindsight the only real life use-case I can find that it would solve is doing an exec() right after a fork().
>

Antoine seems to think that you can't release the I/O locks around I/O
syscalls (when the GIL is released). But I'm sure that if you come up
with a working patch it'll get considered for inclusion ;-)

> Since I think we agree you can't just disable fork() when multiple threads are running, how about at least issuing a warning in that case? That would be a two-line change in threading.py.

You mean a runtime warning? That would be ugly and clumsy.
A warning is probably a good idea, but it should be added somewhere in
os.fork() and threading documentation.

Well, I ping my view that we should:

1) Add general atfork() mechanism.
2) Dive into the std lib and add handlers one by one, that depending on case, either do the lock/init thing or just init the state of the library to some valid state in the child.

Once this mechanism is in place and committed with a few obvious handlers such as the one for the logging library, other handlers can be added over time.

Following this path we will slowly resolve the problem, handler by handler, without introducing the invalid state problem.

> You mean a runtime warning? That would be ugly and clumsy.
> A warning is probably a good idea, but it should be added somewhere in
os.fork() and threading documentation.

I was thinking about a run time warning that is emitted if you call os.fork() while multiple threads are active. It is ugly, but at least it tells you you are doing something that will in most cases not work correctly. I certainly agree that a warning should also be added to os.fork() documentation.

I'm attaching an example patch that adds it into _after_fork() in threading.py, but there are a number of other places where it might go instead. 

I believe that the comp.programming.threads post from David Butenhof linked above explains why adding atfork() handlers isn't going to solve this.

> I was thinking about a run time warning that is emitted if you call
> os.fork() while multiple threads are active. It is ugly, but at least
> it tells you you are doing something that will in most cases not work
> correctly.

The problem is that multiprocessing itself, by construction, uses fork()
with multiple threads. Perhaps there's a way to use only non-blocking
communication instead (rendering the helper threads useless), but that's
not a trivial change.

> I believe that the comp.programming.threads post from 
> David Butenhof linked above explains why adding atfork() 
> handlers isn't going to solve this.

In Python atfork() handlers will never run from signal handlers, and if I understood correctly, Charles-François described a way to "re-initialize" a Python lock safely under that assumption.

My suggestion to this would be that it should be outdated in the
same way that Georg Brandl has suggested for changing the default
encoding on python-dev [1], and add clear documentation on that,
also in respect to the transition phase ..

> The problem is that multiprocessing itself, by construction,
> uses fork() with multiple threads.

.. and maybe add some switches which allow usage of fork() for the
time being.

Today a '$ grep -Fir fork' does not show up threading.rst at all,
which seems to be little in regard to the great problem.
I would add a big fat note that multiprocessing.Process should be
used instead today, because how about those of us who are not
sophisticated enough to be appointed to standard committees?

But anyway we should be lucky: fork(2) is UNIX specific, and thus
it can be expected that all thread-safe libraries etc. are aware of
the fact that they may be cloned by it.  Except mine, of course.  ,~)

[1] http://mail.python.org/pipermail/python-dev/2011-June/112126.html
--
Ciao, Steffen
sdaoden(*)(gmail.com)
() ascii ribbon campaign - against html e-mail
/\ www.asciiribbon.org - against proprietary attachments

> I would add a big fat note that multiprocessing.Process should be
> used instead today, because how about those of us who are not
> sophisticated enough to be appointed to standard committees?

How do you think multiprocessing.Process launches a new process?

> How do you think multiprocessing.Process launches a new process?

But it's a single piece of code under control and even
multi-OS/multi-architecture test-coverage, not a general purpose
"Joe, you may just go that way and Python will handle it
correctly"?

What i mean is: ten years ago (or so), Java did not offer true
selection on sockets (unless i'm mistaken) - servers needed a 1:1
mapping of threads:sockets to handle connections?!
But then, a "this thread has finished the I/O, let's use it for
something different" seems to be pretty obvious.
This is ok if it's your professor who is forcefully misleading
you into the wrong direction, but otherwise you will have
problems, maybe sooner, maybe later (, maybe never).  And
currently there is not a single piece of documentation which
points you onto the problems.  (And there *are* really people
without Google.)

The problem is that it looks so simple and easy - but it's not.
In my eyes it's an unsolvable problem.  And for the sake of
resource usage, simplicity and execution speed i appreciate all
solutions which don't try to do the impossible.

I want to add that all this does not really help just as long just
*any* facility which is used by Python *itself* is not under control
of atfork.  Solaris e.g. uses atfork for it's memory allocator,
because that is surely needed if anything else but async-safe
facilities are used in the newly forked process.  Can Python give
that guarantee for all POSIX systems it supports?

Good night.
--
Ciao, Steffen
sdaoden(*)(gmail.com)
() ascii ribbon campaign - against html e-mail
/\ www.asciiribbon.org - against proprietary attachments

> > How do you think multiprocessing.Process launches a new process?
> 
> But it's a single piece of code under control and even
> multi-OS/multi-architecture test-coverage, not a general purpose
> "Joe, you may just go that way and Python will handle it
> correctly"?

Sorry, how does that make the problem any different?

> Well, I ping my view that we should:
>

Could you please detail the following points:
- what would be the API of this atfork() mechanism (with an example of
how it would be used in the library)?
- how do you find the correct order to acquire locks in the parent process?
- what do you do with locks that can be held for arbitrarily long
(e.g. I/O locks)?

> - what would be the API of this atfork() mechanism (with an example of how it would be used in the library)?

The atfork API is defined in POSIX and Gregory P. Smith proposed a Python one above that we can look into.
http://pubs.opengroup.org/onlinepubs/009695399/functions/pthread_atfork.html

We may need an API to reset a lock.

> - how do you find the correct order to acquire locks in the parent process?

One option is to use the import graph to determine call order of atfork handlers. 
If a current std library does not fit into this scheme we can possibly fix it when writing its handlers.

> - what do you do with locks that can be held for arbitrarily long (e.g. I/O locks)?

It is likely that such a lock does not need acquiring at the parent, and re-initializing the library in the child handler will do.
A  "critical section" lock that protects in-memory data should not be held for long.

>> - how do you find the correct order to acquire locks in the parent process?
>
> One option is to use the import graph to determine call order of atfork handlers.
> If a current std library does not fit into this scheme we can possibly fix it when writing its handlers.
>

Sorry, I fail to see how the "import graph" is related to the correct
lock acquisition order. Some locks are created dynamically, for
example.
That's why I asked for a specific API: when do you register a handler?
When are they called? When are they reset?

>> - what do you do with locks that can be held for arbitrarily long (e.g. I/O locks)?
>
> It is likely that such a lock does not need acquiring at the parent, and re-initializing the library in the child handler will do.

The whole point of atfork is to avoid breaking invariants and
introduce invalid state in the child process. If there is one thing we
want to avoid, it's precisely reading/writting corrupted data from/to
files, so eluding the I/O problem seems foolish to me.

> A  "critical section" lock that protects in-memory data should not be held for long.

Not necessarily. See for example I/O locks and logging module, which
hold locks until I/O completion.

> Sorry, I fail to see how the "import graph" is related to the correct
> lock acquisition order. Some locks are created dynamically, for
> example.

Import dependency is a reasonable heuristic to look into for inter-module locking order. 

The rational is explained in the following pthread_atfork man page:
http://pubs.opengroup.org/onlinepubs/009695399/functions/pthread_atfork.html
"A higher-level package may acquire locks on its own data structures before invoking lower-level packages. Under this scenario, the order specified for fork handler calls allows a simple rule of initialization for avoiding package deadlock: a package initializes all packages on which it depends before it calls the pthread_atfork() function for itself."

(The rational section is an interpretation which is not part of the standard)

A caveat is that since Python is an object oriented language it is more common than with C that code from a higher level module will be invoked by code from a lower level module, for example by calling an object method that was over-ridden by the higher level module - this actually happens in the logging module (emit method).

> That's why I asked for a specific API: when do you register a handler?
> When are they called? When are they reset?

Read the pthread_atfork man page.

> The whole point of atfork is to avoid breaking invariants and
> introduce invalid state in the child process. If there is one thing we
> want to avoid, it's precisely reading/writting corrupted data from/to
> files, so eluding the I/O problem seems foolish to me.

Please don't use insulting adjectives. 
If you think I am wrong, convincing me logically will do.

you can "avoid breaking invariants" using two different strategies:
1) Acquire locks before the fork and release/reset them after it.
2) Initialize the module to some known state after the fork.

For some (most?) modules it may be quite reasonable to initialize the module to a known state after the fork without acquiring its locks before the fork; this too is explained in the pthread_atfork man page:
"Alternatively, some libraries might be able to supply just a child routine that reinitializes the mutexes in the library and all associated states to some known value (for example, what it was when the image was originally executed)."

> > A  "critical section" lock that protects in-memory data should not be held for long.
>
> Not necessarily. See for example I/O locks and logging module, which
> hold locks until I/O completion.

Oops, I have always used the term "critical section" to describe a lock that protects data state as tightly as possible, ideally not even across function calls but now I see the Wikipedia defines one to protect any resource including IO.

The logging module locks the entire emit() function which I think is wrong. 
It should let the derived handler take care of locking when it needs to, if it needs to at all.

The logging module is an example for a module we should reinitialize after the fork without locking its locks before the fork.

[...]
> A caveat is that since Python is an object oriented language it is more common than with C that code from a higher level module will be invoked by code from a lower level module, for example by calling an object method that was over-ridden by the higher level module - this actually happens in the logging module (emit method).

Exactly. That's why registering atfork() handler in independent
modules can - and will - lead to deadlocks, if we get the order wrong.
Also, most locks are allocated dynamically (at the same time as the
object they protect), so the import order is not really relevant here.
Furthermore, there's not a strict ordering between the modules: how is
bz2 compared to loglib, for example?

>
>> That's why I asked for a specific API: when do you register a handler?
>> When are they called? When are they reset?
>
> Read the pthread_atfork man page.
>

No, it won't do it, since when an object - and its protecting lock -
is deallocated, the related atfork handler must be removed, for
example. You might handle this with wearefs, but that's definitely
something not accounted for by the pthread_atfork standard API.

>> The whole point of atfork is to avoid breaking invariants and
>> introduce invalid state in the child process. If there is one thing we
>> want to avoid, it's precisely reading/writting corrupted data from/to
>> files, so eluding the I/O problem seems foolish to me.
>
> Please don't use insulting adjectives.
> If you think I am wrong, convincing me logically will do.
>

I'm sorry if that sounded insulting to you, it was really
unintentional (English is not my mother tongue).

> you can "avoid breaking invariants" using two different strategies:
> 1) Acquire locks before the fork and release/reset them after it.
> 2) Initialize the module to some known state after the fork.
>
> For some (most?) modules it may be quite reasonable to initialize the module to a known state after the fork without acquiring its locks before the fork; this too is explained in the pthread_atfork man page:
> "Alternatively, some libraries might be able to supply just a child routine that reinitializes the mutexes in the library and all associated states to some known value (for example, what it was when the image was originally executed)."
>

The most problematic place is the I/O layer, since those are the locks
held longer (see for example issue #7123). And I'm not sure we can
simply reinit the I/O object after fork() without corrupting or losing
data.
But this approach (reinitializing after fork()) works well most of the
time, and is actually already used in multiple places (threading and
multiprocessing modules, and probably others).

> Oops, I have always used the term "critical section" to describe a lock that protects data state as tightly as possible, ideally not even across function calls but now I see the Wikipedia defines one to protect any resource including IO.
>

Yes, that's one peculiarity of Python locks.
Another one is that a lock can be released by a process other than the
one who acquired it.

> The logging module locks the entire emit() function which I think is wrong.
> It should let the derived handler take care of locking when it needs to, if it needs to at all.
>
> The logging module is an example for a module we should reinitialize after the fork without locking its locks before the fork.

It's possible.

Like I said earlier in this thread, I'm not at all opposed to the
atfork mechanism. I'm actually in favor of it, the first reason being
that we could factorize the various ad-hoc atfork handlers scattered
through the standard library.
My point is just that it's not as simple as it sounds because of
long-held locks, and that we've got to be really careful because of
inter-module dependencies.

Would you like to work on a patch to add an atfork mechanism?

Except for multiprocessing, does anyone know of any other module in the standard library that uses fork() and threads at the same time? After some grepping through the source I couldn't find any other cases.

I'm still in favor of just deprecating using fork() on a multithreaded process (with appropriate warnings and documentation) and I'm prepared to work on a patch that would remove the need for helper threads in the multiprocessing module.

I gather that having atfork would be useful beyond attempting to solve the locking problem, so this doesn't mean I'm opposed to it. However debugging rare problems in multithreaded/multiprocess applications is such a frustrating task that I really don't like a solution that only works in the common case.

> In Python atfork() handlers will never run from signal handlers, and 
> if I understood correctly, Charles-François described a way to 
> "re-initialize" a Python lock safely under that assumption.

Just to clarify: it's not that POSIX atfork() handlers run from signal handlers. It's that after a fork in a multithreaded process POSIX only guarantees calls to "safe" functions, which is the same set of functions as those that are safe to call from signal handlers. This fact does not change for Python's os.fork().

> Except for multiprocessing, does anyone know of any other module in
> the standard library that uses fork() and threads at the same time?
> After some grepping through the source I couldn't find any other
> cases.

It's quite common to launch a subprocess from a thread, so as to
communicate with the subprocess without blocking the main thread. I'm
not sure the stdlib itself does it, but the test suite does (when run in
parallel mode).

> I'm prepared to work on a patch that would remove the need for helper
> threads in the multiprocessing module.

Your contribution is welcome.

> Just to clarify: it's not that POSIX atfork() handlers run from signal
> handlers. It's that after a fork in a multithreaded process POSIX only
> guarantees calls to "safe" functions, which is the same set of
> functions as those that are safe to call from signal handlers.

For the record, I would consider POSIX totally broken from this point of
view. It seems most modern systems allow much more than that,
fortunately.

> Except for multiprocessing, does anyone know of any other module in the standard library that uses fork() and threads at the
> same time? After some grepping through the source I couldn't find any other cases.

The same problem arises in case of user-created threads, this problem
is not specific to the multiprocessing.

> Just to clarify: it's not that POSIX atfork() handlers run from signal handlers. It's that after a fork in a multithreaded process POSIX only guarantees calls to "safe" functions, which is the same set of functions as those that are safe to call from signal handlers. This fact does not change for Python's os.fork().
>

I think Nir knows perfectly that, he was just referring to a
limitation of pthread_atfork:
- fork() is async-safe, and thus can be called from a signal handler
- but if pthread_atfork handlers are installed, then fork() can become
non async-safe, if the handlers are not async-safe (and it's the case
when you're dealing with POSIX mutexes for example)
But since Python's user-defined signal handlers are actually called
synchronously (and don't run on behalf of the signal handler), there's
no risk of fork() being called from a signal handler.

> I'm still in favor of just deprecating using fork() on a multithreaded process (with appropriate warnings and documentation)

We can't do that, it would break existing code.
Furthermore, some libraries use threads behind the scene.

> I'm prepared to work on a patch that would remove the need for helper threads in the multiprocessing module.

What do you mean by helper threads?

> We can't do that, it would break existing code.

I would argue that such code is already broken.

> What do you mean by helper threads?

multiprocessing uses threads behind the scenes to handle queue traffic and such for individual forked processes. It's something I also wasn't aware of until Antoine pointed it out. It also has its own implementation of atfork hooks in an attempt to handle the locking issue.

>> We can't do that, it would break existing code.
>
> I would argue that such code is already broken.
>

- that's not necessarily true, if your code is carefully designed
- we can't forbid fork() in a multi-threaded application while it's
allowed by POSIX
- backward compatibility is *really* important

>> What do you mean by helper threads?
>
> multiprocessing uses threads behind the scenes to handle queue traffic and such for individual forked processes. It's something I also wasn't aware of until Antoine pointed it out. It also has its own implementation of atfork hooks in an attempt to handle the locking issue.
>

I'm curious as to how you'll manage to implement
multiprocessing.queues without threads.
Please open a dedicated issue for this.

> Would you like to work on a patch to add an atfork mechanism?

I will start with an attempt to generate a summary "report" on this rabbit hole of a problem, the views and suggestions raised by people here and what we may expect from atfork approach, its limitations, etc... I will also take a deeper look into the code.

Hopefully my brain will not deadlock or fork while I am at it.

more words, I know...

Well, my brain did not deadlock, but after spinning on the problem for a while longer, it now thinks Tomaž Šolc and Steffen are right.

We should try to fix the multiprocessing module so it does not deadlock single-thread programs and deprecate fork in multi-threaded programs.

Here is the longer version, which is a summary of what people said here in various forms, observations from diving into the code and Googling around:


1) The rabbit hole

a) In a multi-threaded program, fork() may be called while another thread is in a critical section. That thread will not exist in the child and the critical section will remain locked. Any attempt to enter that critical section will deadlock the child.

b) POSIX included the pthread_atfork() API in an attempt to deal with the problem:
http://pubs.opengroup.org/onlinepubs/009695399/functions/pthread_atfork.html

c) But it turns out atfork handlers are limited to calling async-signal-safe functions since fork may be called from a signal handler:
http://download.oracle.com/docs/cd/E19963-01/html/821-1601/gen-61908.html#gen-95948

This means atfork handlers may not actually acquire or release locks. See opinion by David Butenhof who was involved in the standardization effort of POSIX threads:
http://groups.google.com/group/comp.programming.threads/msg/3a43122820983fde

d) One consequence is that we can not assume third party libraries are safe to fork in multi-threaded program. It is likely their developers consider this scenario broken.

e) It seems the general consensus across the WWW concerning this problem is that it has no solution and that a fork should be followed by exec as soon as possible.

Some references:
http://pubs.opengroup.org/onlinepubs/009695399/functions/fork.html
http://austingroupbugs.net/view.php?id=62
http://sourceware.org/bugzilla/show_bug.cgi?id=4737
http://www.linuxprogrammingblog.com/threads-and-fork-think-twice-before-using-them


2) Python’s killer rabbit

The standard library multiprocessing module does two things that force us into the rabbit hole; it creates worker threads and forks without exec.

Therefore, any program that uses the multiprocessing module is a multi-threading forking program.

One immediate problem is that a multiprocessing.Pool may fork from its worker thread in Pool._handle_workers(). This puts the forked child at risk of deadlock with any code that was run by the parent’s main thread (the entire program logic).

More problems may be found with a code review.

Other modules to look at are concurrent.futures.process (creates a worker thread and uses multiprocessing) and socketserver (ForkingMixIn forks without exec).


3) God bless the GIL

a) Python signal handlers run synchronously in the interpreter loop of the main thread, so os.fork() will never be called from a POSIX signal handler.

This means Python atfork prepare and parent handlers may run any code. The code run at the child is still restricted though and may deadlock into any acquired locks left behind by dead threads in the standard library or lower level third party libraries.

b) Turns out the GIL also helps by synchronizing threads.

Any lock held for the duration of a function call while the GIL is held will be released by the time os.fork() is called. But if a thread in the program calls a function that yields the GIL we are in la la land again and need to watch out step.


4) Landing gently at the bottom

a) I think we should try to review and sanitize the worker threads of the multiprocessing module and other implicit worker threads in the standard library.

Once we do (and since os.fork() is never run from a POSIX signal handler) the multiprocessing library should be safe to use in programs that do not start other threads.

b) Then we should declare the user scenario of mixing the threading and multiprocessing modules as broken by design.

c) Finally, optionally provide atfork API

The atfork API can be used to refactor existing fork handlers in the standard library, provide handlers for modules such as the logging module that will reduce the risk of deadlock in existing programs, and can be used by developers who insist on mixing threading and forking in their programs.


5) Sanitizing worker threads in the multiprocessing module

TODO :) 

(will try to post some ideas soon)

Here is a morning reasoning exercise - please help find the flaws or refine it:


5) Sanitizing worker threads in the multiprocessing module

Sanitizing a worker thread in the context of this problem is to make sure it can not create a state that may deadlock another thread that calls fork(); or in other words fork-safe.
Keep in mind that in Python os.fork() is never called from a POSIX signal handler.
So what are examples of a fork-safe thread?

a) A thread that spins endlessly doing nothing in a C for(;;) loop is safe.

Another thread may call fork() without restrictions.

b) A Python thread that only calls function that do not yield the GIL and that does not acquire locks that are held beyond a Python tick is safe.

An example for such a lock is a critical-section lock acquired by a lower level third party library for the duration of a function call.
Such a lock will be released by the time os.fork() is called because of the GIL.

c) A Python thread that in addition to (2) also acquires a lock that is handled at fork is safe.

d) A Python thread that in addition to (2) and (3) calls function that yield the GIL but while the GIL is released only calls async-signal-safe code.

This is a bit tricky. We know that it is safe for thread A to fork and call async-signal-safe functions regardless of what thread B has been doing, but I do not know that thread A can fork and call non async-signal-safe functions if thread B was only calling async-signal-safe functions.

Nevertheless it makes sense: For example lets assume it isn't true, and that hypothetical thread A forked while thread B was doing the async-signal-safe function safe_foo(), and then thread A called non async-signal-safe function unsafe_bar() and deadlocked.

unsafe_bar() could have deadlocked trying to acquire a lock that was acquired by safe_foo(). But if this is so, then it could also happen the other way around.
Are there other practical possibilities?

Either way, we could double check and white list the async-signal-safe functions we are interested in, in a particular implementation.

e) Socket related functions such as bind() accept() send() and recv(), that Python calls without holding the GIL, are all async-signal-safe.

This means that in principle we can have a fork-safe worker thread for the purpose of communicating with a forked process using a socket.

No Python thread is ever fork safe because the Python interpreter itself can never be made fork safe. Nor should anyone try to make the interpreter itself safe. It is too complex and effectively impossible to guarantee.

There is no general solution to this, fork and threading is simply broken in POSIX and no amount of duct tape outside of the OS kernel can fix it. My only desire is that we attempt to do the right thing when possible with the locks we know about within the standard library.

If Nir's analysis is right, and Antoines comment pushes me into
this direction, (i personally have not looked at that code),
then multiprocessing is completely brain-damaged and has been
implemented by a moron.
And yes, I know this is a bug tracker, and even that of Python.

Nir should merge his last two messages into a single mail to
python-dev, and those guys should give Nir or Thomas or a group of
persons who have time and mental power a hg(1) repo clone and
committer access to that and multiprocessing should be rewritten,
maybe even from scratch, but i dunno.

For the extremely unlikely case that all that doesn't happen maybe
the patch of neologix should make it?

--Steffen
Ciao, sdaoden(*)(gmail.com)
() ascii ribbon campaign - against html e-mail
/\ www.asciiribbon.org - against proprietary attachments

Um, and just to add: i'm not watching out for anything, and it won't
and it can't be me:

?0%0[steffen@sherwood sys]$ grep -F smp CHANGELOG.svn -B3 | grep -E '^r[[:digit:]]+' | tail -n 1
r162 | steffen | 2006-01-18 18:29:58 +0100 (Wed, 18 Jan 2006) | 35 lines

--Steffen
Ciao, sdaoden(*)(gmail.com)
() ascii ribbon campaign - against html e-mail
/\ www.asciiribbon.org - against proprietary attachments

> then multiprocessing is completely brain-damaged and has been
> implemented by a moron.

Please do not use this kind of language. 
Being disrespectful to other people hurts the discussion.

P.S.:
I have to apologize, it's Tomaž, not Thomas.
(And unless i'm mistaken this is pronounced "TomAsch" rather than
the english "Tommes", so i was just plain wrong.)

--Steffen
Ciao, sdaoden(*)(gmail.com)
() ascii ribbon campaign - against html e-mail
/\ www.asciiribbon.org - against proprietary attachments

> > then multiprocessing is completely brain-damaged and has been
> > implemented by a moron.
> 
> Please do not use this kind of language. 
> Being disrespectful to other people hurts the discussion.

So i apologize once again.
'Still i think this should go to python-dev in the mentioned case.

(BTW: there are religions without "god", so whom shall e.g. i praise
for the GIL?)

--Steffen
Ciao, sdaoden(*)(gmail.com)
() ascii ribbon campaign - against html e-mail
/\ www.asciiribbon.org - against proprietary attachments

> (BTW: there are religions without "god", so whom shall e.g. i praise for the GIL?)

Guido? ;)

Hi Gregory,

> Gregory P. Smith <greg@krypto.org> added the comment:
> No Python thread is ever fork safe because the Python interpreter itself can never be made fork safe. 
> Nor should anyone try to make the interpreter itself safe. It is too complex and effectively impossible to guarantee.

a) I think the term "Guarantee" is not meaningful here since the interpreter is probably too complex to guarantee it does not contain other serious problems.
b) If no Python thread is ever fork safe, can you illustrate how a trivial Python thread spinning endlessly might deadlock a child forked by another Python thread?

I was not able to find reports of deadlocks clearly related to multiprocessing worker threads so they could be practically safe already, to the point other Python-Dev developers would be inclined to bury this as a theoretical problem :)

Anyway, there exists at least the problem of forking from the pool worker thread and possibly other issues, so the code should be reviewed.
Another latent problem is multiprocessing logging which is disabled by default?


> There is no general solution to this, fork and threading is simply broken in POSIX and no amount of duct tape outside of the OS kernel can fix it. 

This is why we should "sanitize" the multithreading module and deprecate mixing of threading and multiprocessing. 
I bet most developers using Python are not even aware of this problem. 
We should make sure they are through documentation.

Here is another way to look at the current situation:

1) Don't use threading for concurrency because of the GIL.
2) Don't mix threading with multiprocessing because threading and forking don't mix well.
3) Don't use multiprocessing because it can deadlock.

We should make sure developers are aware of (2) and can use (3) safely***.


> My only desire is that we attempt to do the right thing when possible with the locks we know about within the standard library.

Right, with an atfork() mechanism.

multiprocessing.util already has register_after_fork() which it uses for cleaning up certain things when a new process (launched by multiprocessing) is starting.  This is very similar to the proposed atfork mechanism.

Multiprocessing assumes that it is always safe to delete lock objects.  If reinit_locks.diff is committed then I guess this won't be a problem.

I will try to go through multiprocessing's use of threads: 

Queue
-----

Queue's have a feeder thread which pushes objects in to the underlying pipe as soon as possible.  The state which can be modified by this thread is a threading.Condition object and a collections.deque buffer.  Both of these are replaced by fresh copies by the after-fork mechanism.

However, because objects in the buffer may have __del__ methods or weakref callbacks associated, arbitrary code may be run by the background thread if the reference count falls to zero.

Simply pickling the argument of put() before adding it to the buffer fixes that problem -- see the patch for Issue 10886.  With this patch I think Queue's use of threads is fork-safe.

Pool
----

If a fork occurs while a pool is running then a forked process will get a copy of the pool object in an inconsistent state -- but that does not matter since trying to use a pool from a forked process will *never* work.

Also, some of a pool's methods support callbacks which can execute arbitrary code in a background thread.  This can create inconsistent state in a forked process

As with Queue.put, pool methods should pickle immediately for similar reasons.

I would suggest documenting clearly that a pool should only ever be used or deleted by the process which created it.  We can use register_after_fork to make all of a pool's methods raise an error after a fork.  

We should also document that callbacks should only be used if no more processes will be forked.

allow_connection_pickling
-------------------------

Currently multiprocessing.allow_connection_pickling() does not work because types are registered with ForkingPickler instead of copyreg -- see Issue 4892.  However, the code in multiprocessing.reduction uses a background thread to support the transfer of sockets/connections between processes.

If this code is ever resurrected I think the use of register_after_fork makes this safe.

Managers
--------

A manager uses a threaded server process.  This is not a problem unless you create a user defined manager which forks new processes.   The documentation should just say Don't Do That.


I think multiprocessing's threading issues are all fixable.

For the record, turns out there was a bit of misunderstanding. 

I used the term deprecate above to mean "warn users (through documentation) that they should not use (a feature)" and not in its Python-dev sense of "remove (a feature) after a period of warning".

I do not think the possibility to mix threading and multiprocessing together should be somehow forcibly disabled. 

Anyway, since my view does not seem to resonate with core developers I I'll give it a rest for now.

> Anyway, since my view does not seem to resonate with core developers I I'll
> give it a rest for now.

Well, the problem is that many views have been expressed in this
thread, which doesn't help getting a clear picture of what's needed to
make progress on this issue.
AFAIC, I think the following seems reasonable:
1) add an atfork module which provides a generic and
pthread_atfork-like mechanism to setup handlers that must be called
after fork (right now several modules use their own ad-hoc mechanism)
2) for multiprocessing, call exec() after fork() (issue #8713)
3) for buffered file objects locks, use the approach similar to the
patch I posted (reinit locks in the child process right after fork())

Does that sound reasonable?

Just wanted to say that I spent something like 8 hours debugging a subprocess + threading + logging deadlock on a real production system. 

I suspected one of my locks at first, but I couldn't find any. The post-fork code was very simple, and I didn't suspect that logging would be subject to the same issue.

The good news that I see a very clean solution for fixing this.

We can't free all locks across fork -- that is unsafe and mad, because the child might end up corrupting some shared (network) resource, for example/

However, extending RLock to provide ForkClearedRLock (this would be used by logging, i.e.) is quite straighforward.

The extended class would simply need to record the process ID, in which the lock was created, and the process ID, which is trying to acquire it.  Done!

> However, extending RLock to provide ForkClearedRLock (this would be used by logging, i.e.) is quite straighforward.
>
> The extended class would simply need to record the process ID, in which the lock was created, and the process ID, which is trying to acquire it.  Done!

There are at least two problems with this approach.
- with LinuxThreads, threads have different PIDs, so this would break.
 LinuxThreads have now been replaced with NPTL, so this may not be a
showstopper, though
- however, the other problem is more serious: it has the exact same
problem as the lock reinitialization upon fork(): locks are used to
protect critical sections, to make sure that threads see a consistent
state: if you simply proceed and reset/acquire the lock when the
process is not the last one that owned it, the invariants protected by
the lock will be broken.
The proper solution is to setup handlers to be called upon fork, that
not only reset locks, but also internal state of objects they protect.
However, this is a dull and boring task, and would require patching
dozens of different places. It's been on my todo list for some time...
Another "solution" would be to remove the only place in the standard
library where a bare fork() is used, in multiprocessing (issue #8713).
Then, it's the user's problem if he calls fork() :-)

A new lock type will NOT solve this.  It is ALWAYS okay to clear all thread/threading module locks after a fork.

They are and always have been process-local by definition so they are also by definition 100% invalid to any child process.

Anyone who has written code using them to "lock" an out-of-process resource has written code that is already broken today. Thread locks can't guard network resources.

Attached is a patch (without documentation) which creates an atfork module for Unix.

Apart from the atfork() function modelled on pthread_atfork() there is also a get_fork_lock() function.  This returns a recursive lock which is held whenever a child process is created using os.fork(), subprocess.Popen() or multiprocessing.Process().  It can be used like

    with atfork.get_fork_lock():
        r, w = os.pipe()
        pid = os.fork()
        if pid == 0:
            try:
                os.close(r)
                # do something with w
            finally:
                os._exit(0)
        else:
            os.close(w)

    # do something with r

This prevents processes forked by other threads from accidentally inheriting the writable end (which would potentially cause EOF to be delayed when reading from the pipe).  It can also be used to eliminate the potential race where you create an fd and then set the CLOEXEC flag on it.

The patch modifies Popen() and Process.start() to acquire the lock when they create their pipes.  (A race condition previously made Process.sentinel and Process.join() potentially unreliable in a multithreaded program.)

Note that using the deprecated os.popen?() and os.spawn?() functions can still cause accidental inheritance of fds.

(I have also done a hopefully complete patch to multiprocessing to optionally allow fork+exec on Unix -- see Issue 8713.)

Is there any particular reason not to merge Charles-François's reinit_locks.diff?

Reinitialising all locks to unlocked after a fork seems the only sane option.

> Is there any particular reason not to merge Charles-François's reinit_locks.diff?
> 
> Reinitialising all locks to unlocked after a fork seems the only sane option.

I agree with this. 
I haven't looked at the patch very closely. I think perhaps each lock
could have an optional callback for specific code to be run after
forking, but that may come in another patch.
(this would allow to make e.g. the C RLock fork-safe)