cpython/Lib/multiprocessing/heap.py
Antoine Pitrou e4679cd644
bpo-32759: Free unused arenas in multiprocessing.heap (GH-5827)
Large shared arrays allocated using multiprocessing would remain allocated
until the process ends.
2018-04-09 17:37:55 +02:00

338 lines
11 KiB
Python

#
# Module which supports allocation of memory from an mmap
#
# multiprocessing/heap.py
#
# Copyright (c) 2006-2008, R Oudkerk
# Licensed to PSF under a Contributor Agreement.
#
import bisect
from collections import defaultdict
import mmap
import os
import sys
import tempfile
import threading
from .context import reduction, assert_spawning
from . import util
__all__ = ['BufferWrapper']
#
# Inheritable class which wraps an mmap, and from which blocks can be allocated
#
if sys.platform == 'win32':
import _winapi
class Arena(object):
"""
A shared memory area backed by anonymous memory (Windows).
"""
_rand = tempfile._RandomNameSequence()
def __init__(self, size):
self.size = size
for i in range(100):
name = 'pym-%d-%s' % (os.getpid(), next(self._rand))
buf = mmap.mmap(-1, size, tagname=name)
if _winapi.GetLastError() == 0:
break
# We have reopened a preexisting mmap.
buf.close()
else:
raise FileExistsError('Cannot find name for new mmap')
self.name = name
self.buffer = buf
self._state = (self.size, self.name)
def __getstate__(self):
assert_spawning(self)
return self._state
def __setstate__(self, state):
self.size, self.name = self._state = state
# Reopen existing mmap
self.buffer = mmap.mmap(-1, self.size, tagname=self.name)
# XXX Temporarily preventing buildbot failures while determining
# XXX the correct long-term fix. See issue 23060
#assert _winapi.GetLastError() == _winapi.ERROR_ALREADY_EXISTS
else:
class Arena(object):
"""
A shared memory area backed by a temporary file (POSIX).
"""
if sys.platform == 'linux':
_dir_candidates = ['/dev/shm']
else:
_dir_candidates = []
def __init__(self, size, fd=-1):
self.size = size
self.fd = fd
if fd == -1:
# Arena is created anew (if fd != -1, it means we're coming
# from rebuild_arena() below)
self.fd, name = tempfile.mkstemp(
prefix='pym-%d-'%os.getpid(),
dir=self._choose_dir(size))
os.unlink(name)
util.Finalize(self, os.close, (self.fd,))
os.ftruncate(self.fd, size)
self.buffer = mmap.mmap(self.fd, self.size)
def _choose_dir(self, size):
# Choose a non-storage backed directory if possible,
# to improve performance
for d in self._dir_candidates:
st = os.statvfs(d)
if st.f_bavail * st.f_frsize >= size: # enough free space?
return d
return util.get_temp_dir()
def reduce_arena(a):
if a.fd == -1:
raise ValueError('Arena is unpicklable because '
'forking was enabled when it was created')
return rebuild_arena, (a.size, reduction.DupFd(a.fd))
def rebuild_arena(size, dupfd):
return Arena(size, dupfd.detach())
reduction.register(Arena, reduce_arena)
#
# Class allowing allocation of chunks of memory from arenas
#
class Heap(object):
# Minimum malloc() alignment
_alignment = 8
_DISCARD_FREE_SPACE_LARGER_THAN = 4 * 1024 ** 2 # 4 MB
_DOUBLE_ARENA_SIZE_UNTIL = 4 * 1024 ** 2
def __init__(self, size=mmap.PAGESIZE):
self._lastpid = os.getpid()
self._lock = threading.Lock()
# Current arena allocation size
self._size = size
# A sorted list of available block sizes in arenas
self._lengths = []
# Free block management:
# - map each block size to a list of `(Arena, start, stop)` blocks
self._len_to_seq = {}
# - map `(Arena, start)` tuple to the `(Arena, start, stop)` block
# starting at that offset
self._start_to_block = {}
# - map `(Arena, stop)` tuple to the `(Arena, start, stop)` block
# ending at that offset
self._stop_to_block = {}
# Map arenas to their `(Arena, start, stop)` blocks in use
self._allocated_blocks = defaultdict(set)
self._arenas = []
# List of pending blocks to free - see comment in free() below
self._pending_free_blocks = []
# Statistics
self._n_mallocs = 0
self._n_frees = 0
@staticmethod
def _roundup(n, alignment):
# alignment must be a power of 2
mask = alignment - 1
return (n + mask) & ~mask
def _new_arena(self, size):
# Create a new arena with at least the given *size*
length = self._roundup(max(self._size, size), mmap.PAGESIZE)
# We carve larger and larger arenas, for efficiency, until we
# reach a large-ish size (roughly L3 cache-sized)
if self._size < self._DOUBLE_ARENA_SIZE_UNTIL:
self._size *= 2
util.info('allocating a new mmap of length %d', length)
arena = Arena(length)
self._arenas.append(arena)
return (arena, 0, length)
def _discard_arena(self, arena):
# Possibly delete the given (unused) arena
length = arena.size
# Reusing an existing arena is faster than creating a new one, so
# we only reclaim space if it's large enough.
if length < self._DISCARD_FREE_SPACE_LARGER_THAN:
return
blocks = self._allocated_blocks.pop(arena)
assert not blocks
del self._start_to_block[(arena, 0)]
del self._stop_to_block[(arena, length)]
self._arenas.remove(arena)
seq = self._len_to_seq[length]
seq.remove((arena, 0, length))
if not seq:
del self._len_to_seq[length]
self._lengths.remove(length)
def _malloc(self, size):
# returns a large enough block -- it might be much larger
i = bisect.bisect_left(self._lengths, size)
if i == len(self._lengths):
return self._new_arena(size)
else:
length = self._lengths[i]
seq = self._len_to_seq[length]
block = seq.pop()
if not seq:
del self._len_to_seq[length], self._lengths[i]
(arena, start, stop) = block
del self._start_to_block[(arena, start)]
del self._stop_to_block[(arena, stop)]
return block
def _add_free_block(self, block):
# make block available and try to merge with its neighbours in the arena
(arena, start, stop) = block
try:
prev_block = self._stop_to_block[(arena, start)]
except KeyError:
pass
else:
start, _ = self._absorb(prev_block)
try:
next_block = self._start_to_block[(arena, stop)]
except KeyError:
pass
else:
_, stop = self._absorb(next_block)
block = (arena, start, stop)
length = stop - start
try:
self._len_to_seq[length].append(block)
except KeyError:
self._len_to_seq[length] = [block]
bisect.insort(self._lengths, length)
self._start_to_block[(arena, start)] = block
self._stop_to_block[(arena, stop)] = block
def _absorb(self, block):
# deregister this block so it can be merged with a neighbour
(arena, start, stop) = block
del self._start_to_block[(arena, start)]
del self._stop_to_block[(arena, stop)]
length = stop - start
seq = self._len_to_seq[length]
seq.remove(block)
if not seq:
del self._len_to_seq[length]
self._lengths.remove(length)
return start, stop
def _remove_allocated_block(self, block):
arena, start, stop = block
blocks = self._allocated_blocks[arena]
blocks.remove((start, stop))
if not blocks:
# Arena is entirely free, discard it from this process
self._discard_arena(arena)
def _free_pending_blocks(self):
# Free all the blocks in the pending list - called with the lock held.
while True:
try:
block = self._pending_free_blocks.pop()
except IndexError:
break
self._add_free_block(block)
self._remove_allocated_block(block)
def free(self, block):
# free a block returned by malloc()
# Since free() can be called asynchronously by the GC, it could happen
# that it's called while self._lock is held: in that case,
# self._lock.acquire() would deadlock (issue #12352). To avoid that, a
# trylock is used instead, and if the lock can't be acquired
# immediately, the block is added to a list of blocks to be freed
# synchronously sometimes later from malloc() or free(), by calling
# _free_pending_blocks() (appending and retrieving from a list is not
# strictly thread-safe but under CPython it's atomic thanks to the GIL).
if os.getpid() != self._lastpid:
raise ValueError(
"My pid ({0:n}) is not last pid {1:n}".format(
os.getpid(),self._lastpid))
if not self._lock.acquire(False):
# can't acquire the lock right now, add the block to the list of
# pending blocks to free
self._pending_free_blocks.append(block)
else:
# we hold the lock
try:
self._n_frees += 1
self._free_pending_blocks()
self._add_free_block(block)
self._remove_allocated_block(block)
finally:
self._lock.release()
def malloc(self, size):
# return a block of right size (possibly rounded up)
if size < 0:
raise ValueError("Size {0:n} out of range".format(size))
if sys.maxsize <= size:
raise OverflowError("Size {0:n} too large".format(size))
if os.getpid() != self._lastpid:
self.__init__() # reinitialize after fork
with self._lock:
self._n_mallocs += 1
# allow pending blocks to be marked available
self._free_pending_blocks()
size = self._roundup(max(size, 1), self._alignment)
(arena, start, stop) = self._malloc(size)
real_stop = start + size
if real_stop < stop:
# if the returned block is larger than necessary, mark
# the remainder available
self._add_free_block((arena, real_stop, stop))
self._allocated_blocks[arena].add((start, real_stop))
return (arena, start, real_stop)
#
# Class wrapping a block allocated out of a Heap -- can be inherited by child process
#
class BufferWrapper(object):
_heap = Heap()
def __init__(self, size):
if size < 0:
raise ValueError("Size {0:n} out of range".format(size))
if sys.maxsize <= size:
raise OverflowError("Size {0:n} too large".format(size))
block = BufferWrapper._heap.malloc(size)
self._state = (block, size)
util.Finalize(self, BufferWrapper._heap.free, args=(block,))
def create_memoryview(self):
(arena, start, stop), size = self._state
return memoryview(arena.buffer)[start:start+size]