micropython/extmod/asyncio/task.py

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

178 lines
5.5 KiB
Python
Raw Normal View History

# MicroPython asyncio module
# MIT license; Copyright (c) 2019-2020 Damien P. George
# This file contains the core TaskQueue based on a pairing heap, and the core Task class.
# They can optionally be replaced by C implementations.
from . import core
# pairing-heap meld of 2 heaps; O(1)
def ph_meld(h1, h2):
if h1 is None:
return h2
if h2 is None:
return h1
lt = core.ticks_diff(h1.ph_key, h2.ph_key) < 0
if lt:
if h1.ph_child is None:
h1.ph_child = h2
else:
h1.ph_child_last.ph_next = h2
h1.ph_child_last = h2
h2.ph_next = None
h2.ph_rightmost_parent = h1
return h1
else:
h1.ph_next = h2.ph_child
h2.ph_child = h1
if h1.ph_next is None:
h2.ph_child_last = h1
h1.ph_rightmost_parent = h2
return h2
# pairing-heap pairing operation; amortised O(log N)
def ph_pairing(child):
heap = None
while child is not None:
n1 = child
child = child.ph_next
n1.ph_next = None
if child is not None:
n2 = child
child = child.ph_next
n2.ph_next = None
n1 = ph_meld(n1, n2)
heap = ph_meld(heap, n1)
return heap
# pairing-heap delete of a node; stable, amortised O(log N)
def ph_delete(heap, node):
if node is heap:
child = heap.ph_child
node.ph_child = None
return ph_pairing(child)
# Find parent of node
parent = node
while parent.ph_next is not None:
parent = parent.ph_next
parent = parent.ph_rightmost_parent
# Replace node with pairing of its children
if node is parent.ph_child and node.ph_child is None:
parent.ph_child = node.ph_next
node.ph_next = None
return heap
elif node is parent.ph_child:
child = node.ph_child
next = node.ph_next
node.ph_child = None
node.ph_next = None
node = ph_pairing(child)
parent.ph_child = node
else:
n = parent.ph_child
while node is not n.ph_next:
n = n.ph_next
child = node.ph_child
next = node.ph_next
node.ph_child = None
node.ph_next = None
node = ph_pairing(child)
if node is None:
node = n
else:
n.ph_next = node
node.ph_next = next
if next is None:
node.ph_rightmost_parent = parent
parent.ph_child_last = node
return heap
# TaskQueue class based on the above pairing-heap functions.
class TaskQueue:
def __init__(self):
self.heap = None
def peek(self):
return self.heap
def push(self, v, key=None):
assert v.ph_child is None
assert v.ph_next is None
v.data = None
v.ph_key = key if key is not None else core.ticks()
self.heap = ph_meld(v, self.heap)
def pop(self):
v = self.heap
assert v.ph_next is None
self.heap = ph_pairing(v.ph_child)
v.ph_child = None
return v
def remove(self, v):
self.heap = ph_delete(self.heap, v)
# Task class representing a coroutine, can be waited on and cancelled.
class Task:
def __init__(self, coro, globals=None):
self.coro = coro # Coroutine of this Task
self.data = None # General data for queue it is waiting on
self.state = True # None, False, True, a callable, or a TaskQueue instance
self.ph_key = 0 # Pairing heap
self.ph_child = None # Paring heap
self.ph_child_last = None # Paring heap
self.ph_next = None # Paring heap
self.ph_rightmost_parent = None # Paring heap
def __iter__(self):
extmod/uasyncio: Fix race with cancelled task waiting on finished task. This commit fixes a problem with a race between cancellation of task A and completion of task B, when A waits on B. If task B completes just before task A is cancelled then the cancellation of A does not work. Instead, the CancelledError meant to cancel A gets passed through to B (that's expected behaviour) but B handles it as a "Task exception wasn't retrieved" scenario, printing out such a message (this is because finished tasks point their "coro" attribute to themselves to indicate they are done, and implement the throw() method, but that method inadvertently catches the CancelledError). The correct behaviour is for B to bounce that CancelledError back out. This bug is mainly seen when wait_for() is used, and in that context the symptoms are: - occurs when using wait_for(T, S), if the task T being waited on finishes at exactly the same time as the wait-for timeout S expires - task T will have run to completion - the "Task exception wasn't retrieved message" is printed with "<class 'CancelledError'>" as the error (ie no traceback) - the wait_for(T, S) call never returns (it's never put back on the uasyncio run queue) and all tasks waiting on this are blocked forever from running - uasyncio otherwise continues to function and other tasks continue to be scheduled as normal The fix here reworks the "waiting" attribute of Task to be called "state" and uses it to indicate whether a task is: running and not awaited on, running and awaited on, finished and not awaited on, or finished and awaited on. This means the task does not need to point "coro" to itself to indicate finished, and also allows removal of the throw() method. A benefit of this is that "Task exception wasn't retrieved" messages can go back to being able to print the name of the coroutine function. Fixes issue #7386. Signed-off-by: Damien George <damien@micropython.org>
2021-06-14 13:32:51 +01:00
if not self.state:
# Task finished, signal that is has been await'ed on.
self.state = False
elif self.state is True:
# Allocated head of linked list of Tasks waiting on completion of this task.
self.state = TaskQueue()
elif type(self.state) is not TaskQueue:
# Task has state used for another purpose, so can't also wait on it.
raise RuntimeError("can't wait")
return self
def __next__(self):
extmod/uasyncio: Fix race with cancelled task waiting on finished task. This commit fixes a problem with a race between cancellation of task A and completion of task B, when A waits on B. If task B completes just before task A is cancelled then the cancellation of A does not work. Instead, the CancelledError meant to cancel A gets passed through to B (that's expected behaviour) but B handles it as a "Task exception wasn't retrieved" scenario, printing out such a message (this is because finished tasks point their "coro" attribute to themselves to indicate they are done, and implement the throw() method, but that method inadvertently catches the CancelledError). The correct behaviour is for B to bounce that CancelledError back out. This bug is mainly seen when wait_for() is used, and in that context the symptoms are: - occurs when using wait_for(T, S), if the task T being waited on finishes at exactly the same time as the wait-for timeout S expires - task T will have run to completion - the "Task exception wasn't retrieved message" is printed with "<class 'CancelledError'>" as the error (ie no traceback) - the wait_for(T, S) call never returns (it's never put back on the uasyncio run queue) and all tasks waiting on this are blocked forever from running - uasyncio otherwise continues to function and other tasks continue to be scheduled as normal The fix here reworks the "waiting" attribute of Task to be called "state" and uses it to indicate whether a task is: running and not awaited on, running and awaited on, finished and not awaited on, or finished and awaited on. This means the task does not need to point "coro" to itself to indicate finished, and also allows removal of the throw() method. A benefit of this is that "Task exception wasn't retrieved" messages can go back to being able to print the name of the coroutine function. Fixes issue #7386. Signed-off-by: Damien George <damien@micropython.org>
2021-06-14 13:32:51 +01:00
if not self.state:
# Task finished, raise return value to caller so it can continue.
raise self.data
else:
# Put calling task on waiting queue.
self.state.push(core.cur_task)
# Set calling task's data to this task that it waits on, to double-link it.
core.cur_task.data = self
def done(self):
extmod/uasyncio: Fix race with cancelled task waiting on finished task. This commit fixes a problem with a race between cancellation of task A and completion of task B, when A waits on B. If task B completes just before task A is cancelled then the cancellation of A does not work. Instead, the CancelledError meant to cancel A gets passed through to B (that's expected behaviour) but B handles it as a "Task exception wasn't retrieved" scenario, printing out such a message (this is because finished tasks point their "coro" attribute to themselves to indicate they are done, and implement the throw() method, but that method inadvertently catches the CancelledError). The correct behaviour is for B to bounce that CancelledError back out. This bug is mainly seen when wait_for() is used, and in that context the symptoms are: - occurs when using wait_for(T, S), if the task T being waited on finishes at exactly the same time as the wait-for timeout S expires - task T will have run to completion - the "Task exception wasn't retrieved message" is printed with "<class 'CancelledError'>" as the error (ie no traceback) - the wait_for(T, S) call never returns (it's never put back on the uasyncio run queue) and all tasks waiting on this are blocked forever from running - uasyncio otherwise continues to function and other tasks continue to be scheduled as normal The fix here reworks the "waiting" attribute of Task to be called "state" and uses it to indicate whether a task is: running and not awaited on, running and awaited on, finished and not awaited on, or finished and awaited on. This means the task does not need to point "coro" to itself to indicate finished, and also allows removal of the throw() method. A benefit of this is that "Task exception wasn't retrieved" messages can go back to being able to print the name of the coroutine function. Fixes issue #7386. Signed-off-by: Damien George <damien@micropython.org>
2021-06-14 13:32:51 +01:00
return not self.state
def cancel(self):
# Check if task is already finished.
extmod/uasyncio: Fix race with cancelled task waiting on finished task. This commit fixes a problem with a race between cancellation of task A and completion of task B, when A waits on B. If task B completes just before task A is cancelled then the cancellation of A does not work. Instead, the CancelledError meant to cancel A gets passed through to B (that's expected behaviour) but B handles it as a "Task exception wasn't retrieved" scenario, printing out such a message (this is because finished tasks point their "coro" attribute to themselves to indicate they are done, and implement the throw() method, but that method inadvertently catches the CancelledError). The correct behaviour is for B to bounce that CancelledError back out. This bug is mainly seen when wait_for() is used, and in that context the symptoms are: - occurs when using wait_for(T, S), if the task T being waited on finishes at exactly the same time as the wait-for timeout S expires - task T will have run to completion - the "Task exception wasn't retrieved message" is printed with "<class 'CancelledError'>" as the error (ie no traceback) - the wait_for(T, S) call never returns (it's never put back on the uasyncio run queue) and all tasks waiting on this are blocked forever from running - uasyncio otherwise continues to function and other tasks continue to be scheduled as normal The fix here reworks the "waiting" attribute of Task to be called "state" and uses it to indicate whether a task is: running and not awaited on, running and awaited on, finished and not awaited on, or finished and awaited on. This means the task does not need to point "coro" to itself to indicate finished, and also allows removal of the throw() method. A benefit of this is that "Task exception wasn't retrieved" messages can go back to being able to print the name of the coroutine function. Fixes issue #7386. Signed-off-by: Damien George <damien@micropython.org>
2021-06-14 13:32:51 +01:00
if not self.state:
return False
# Can't cancel self (not supported yet).
if self is core.cur_task:
raise RuntimeError("can't cancel self")
# If Task waits on another task then forward the cancel to the one it's waiting on.
while isinstance(self.data, Task):
self = self.data
# Reschedule Task as a cancelled task.
if hasattr(self.data, "remove"):
# Not on the main running queue, remove the task from the queue it's on.
self.data.remove(self)
core._task_queue.push(self)
elif core.ticks_diff(self.ph_key, core.ticks()) > 0:
# On the main running queue but scheduled in the future, so bring it forward to now.
core._task_queue.remove(self)
core._task_queue.push(self)
self.data = core.CancelledError
return True