Reading received data from a closed TCP stream using StreamReader.read might hang forever

[vxgmichel]

BPO	35065
Nosy	@asvetlov, @1st1, @vxgmichel
PRs	bpo-35065: Remove StreamReaderProtocol._untrack_reader #10212
Files	stuck_on_py38.py: This test fails on 3.8 but passes on 3.7. patch-bpo-35065.diff: A possible fix

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2018-11-08.12:22:13.649>
created_at = <Date 2018-10-25.12:32:25.991>
labels = ['type-bug', '3.8', 'expert-asyncio']
title = 'Reading received data from a closed TCP stream using `StreamReader.read` might hang forever'
updated_at = <Date 2018-11-08.12:22:13.648>
user = 'https://github.com/vxgmichel'

bugs.python.org fields:

activity = <Date 2018-11-08.12:22:13.648>
actor = 'asvetlov'
assignee = 'none'
closed = True
closed_date = <Date 2018-11-08.12:22:13.649>
closer = 'asvetlov'
components = ['asyncio']
creation = <Date 2018-10-25.12:32:25.991>
creator = 'vxgmichel'
dependencies = []
files = ['47891', '47893']
hgrepos = []
issue_num = 35065
keywords = ['patch']
message_count = 5.0
messages = ['328430', '328442', '328528', '328547', '329467']
nosy_count = 3.0
nosy_names = ['asvetlov', 'yselivanov', 'vxgmichel']
pr_nums = ['10212']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue35065'
versions = ['Python 3.8']

[vxgmichel]

I'm not sure whether it is intended or not, but I noticed a change in the behavior of StreamReader between version 3.7 and 3.8.

Basically, reading some received data from a closed TCP stream using StreamReader.read might hang forever, under certain conditions.

I'm not sure what those conditions are but I managed to reproduce the issue consistently with the following workflow:

• server writes some data
• client reads a part of the data
• client closes the writer
• server closes the writer
• client tries to read the remaining data

The test attached implements the behavior. It fails on 3.8 but passes on 3.7

[asvetlov]

Hi Vincent!

No, the hang is not intended behavior. Thanks for the report.
I'll assign the issue to myself to don't miss the bug; but if you have a patch with a fix -- please don't hesitate to make a Pull Request for it.

[vxgmichel]

Hi Andrew!

I reverted the commit associated with the following PR, and the hanging issue disappeared:
#9201

I'll look into it.

[vxgmichel]

I found the culprit:

cpython/Lib/asyncio/streams.py

Line 350 in a05bef4

self._protocol._untrack_reader()

The call to _untrack_reader is performed too soon. Closing the transport causes protocol.connection_lost() to be "called soon" but by the time it is actually executed, the stream reader has been "untracked". Since the protocol doesn't know the stream reader anymore, it has not way to feed it the EOF.

The fix attached removes the _untrack_reader call and definition altogether. I don't really see the point of this method since one has to wait for connection_lost to be executed before untracking the reader, but connection_lost already gets rid of the reader reference.

With this fix, calling writer.close then awaiting writer.wait_closed (or awaiting writer.aclose) should:

• close the transport
• schedule protocol.connection_lost
• wait for the protocol to be closed
• run protocol.connection_lost
• feed the EOF to the reader
• set the protocol as closed
• get rid of the reader reference in the protocol
• return (making aclose causal and safe)
• the reader can then be safely garbage collected

But maybe I'm missing something about _untrack_reader?

[asvetlov]

New changeset fd512d7 by Andrew Svetlov (Vincent Michel) in branch 'master':
bpo-35065: Remove StreamReaderProtocol._untrack_reader (bpo-10212)
fd512d7