bpo-29406: asyncio SSL contexts leak sockets after calling close with certain servers #409

fafhrd91 · 2017-03-03T00:58:36Z

Some servers does not complete ssl shutdown procedure, which cause socket leak on asyncio side.

the-knights-who-say-ni · 2017-03-03T00:58:37Z

Hello, and thanks for your contribution!

I'm a bot set up to make sure that the project can legally accept your contribution by verifying you have signed the PSF contributor agreement (CLA).

Unfortunately we couldn't find an account corresponding to your GitHub username on bugs.python.org (b.p.o) to verify you have signed the CLA. This is necessary for legal reasons before we can look at your contribution. Please follow these steps to help rectify the issue:

If you don't have an account on b.p.o, please create one
Make sure your GitHub username is listed in "Your Details" at b.p.o
If you have not already done so, please sign the PSF contributor agreement. The "bugs.python.org username " requested by the form is the "Login name" field under "Your Details".
If you just signed the CLA, please wait at least one US business day and then check "Your Details" on bugs.python.org to see if your account has been marked as having signed the CLA (the delay is due to a person having to manually check your signed CLA)
Reply here saying you have completed the above steps

Thanks again to your contribution and we look forward to looking at it!

fafhrd91 · 2017-03-03T00:59:42Z

@1st1 PR for http://bugs.python.org/issue29406
I need to add tests though

fafhrd91 · 2017-03-03T01:00:34Z

I actually signed CLA before I started to contribute to tulip/asyncio

1st1 · 2017-03-03T01:01:48Z

I actually signed CLA before I started to contribute to tulip/asyncio

I think you need to go to the bugs.python.org and add your GH username to your profile.

fafhrd91 · 2017-03-03T01:08:25Z

I dont remembre account that I used when signed cla. I just re-signed it

1st1 · 2017-03-03T05:35:41Z

Maybe you used OAuth?

fafhrd91 · 2017-03-03T23:17:22Z

my bugs.python.org details page shows "Contributor Form Received: Yes"

1st1 · 2017-03-03T23:23:03Z

Did you list your GH account name in your account on bpo?

fafhrd91 · 2017-03-03T23:23:28Z

yes

fafhrd91 · 2017-03-03T23:24:57Z

http://bugs.python.org/user17178

1st1 · 2017-03-03T23:26:56Z

@brettcannon something weird is going on here with @the-knights-who-say-ni bot.

@fafhrd91 Can you rebase your PR against master branch? And then add "needs backport to 3.6" and 3.5 tags.

fafhrd91 · 2017-03-03T23:37:27Z

rebased, but I can not add labels

1st1 · 2017-03-03T23:40:33Z

Lib/asyncio/sslproto.py

+            self._shutdown_timeout_handle = self._loop.call_later(
+                self._shutdown_timeout, self._start_shutdown_timeout)
+
+    def _start_shutdown_timeout(self):


I would rename this method to _on_shutdown_timeout.

1st1 · 2017-03-03T23:43:47Z

Lib/asyncio/sslproto.py

+                self._shutdown_timeout, self._start_shutdown_timeout)
+
+    def _start_shutdown_timeout(self):
+        if self._transport is not None:


To make the code more clear I would set self._transport = None in _finalize and _fatal_error methods.

fafhrd91 · 2017-03-04T01:02:34Z

added simple unit tests

brettcannon · 2017-03-04T01:27:53Z

@1st1 there was nothing weird going on; you didn't remove the "CLA not signed" to retrigger the bot to check for the new CLA signing. In the end the synchronizing of the PR triggered the bot again and picked up on the signing.

1st1 · 2017-03-06T21:38:01Z

Lib/asyncio/sslproto.py


    def _finalize(self):
        if self._transport is not None:
            self._transport.close()
+            self._transport = None
+        if self._shutdown_timeout_handle is not None:


little nit: I'd add a newline before this if.

1st1 · 2017-03-06T21:39:09Z

Lib/asyncio/sslproto.py

@@ -412,7 +413,7 @@ class SSLProtocol(protocols.Protocol):

    def __init__(self, loop, app_protocol, sslcontext, waiter,
                 server_side=False, server_hostname=None,
-                 call_connection_made=True):
+                 call_connection_made=True, shutdown_timeout=5.0):


Why is it 5 seconds? Is this some timeout value defined in some SSL RFC?

1st1 · 2017-03-12T19:35:24Z

@fafhrd91 ping

fafhrd91 · 2017-03-12T19:37:03Z

5 seconds just some number

misg · 2017-04-03T17:12:09Z

You may want to see #981

fafhrd91 · 2017-04-03T17:15:05Z

@misg I think this is slightly different issue, #480 is related to #981

misg · 2017-04-03T18:55:53Z

@fafhrd91 I honestly don't know. In my case (with the server/client described in my PR), aiohttp.connector._cleanup_closed is always called and it is that function that calls asyncio.sslproto.SSLTransport._abort so I think I'm in the context where the ssl server doesn't complete ssh shutdown procedure (I actually didn't test with a ssl server that completes ssh shutdown procedure)

1st1 · 2017-06-09T22:23:25Z

@fafhrd91 Could you please rebase this PR on the most recent master? I'll merge it then.

fafhrd91 · 2017-06-09T23:02:32Z

@1st1 done

fafhrd91 · 2017-06-09T23:03:23Z

wait

fafhrd91 · 2017-06-09T23:28:34Z

@1st1 restored

1st1 · 2017-06-09T23:29:12Z

Can you add a NEWS entry?

fafhrd91 · 2017-06-09T23:37:24Z

added

1st1 · 2017-06-10T02:50:31Z

@1st1 restored

unittests don't seem to pass :(

1st1 · 2017-06-10T03:30:12Z

Lib/asyncio/sslproto.py

@@ -679,12 +692,18 @@ def _fatal_error(self, exc, message='Fatal error on transport'):
            })
        if self._transport:
            self._transport._force_close(exc)
+            self._transport = None


Should we also cleanup _shutdown_timeout_handle here?

Should fatal_error just call finalize?

I don't think that makes sense.

added cleanup code for _shutdown_timeout_handle, it is possible to get here while handle still active

…rvers

… certain servers (#409) (cherry picked from commit a608d2d)

…e with certain servers (GH-409) * asyncio SSL contexts leak sockets after calling close with certain servers * cleanup _shutdown_timeout_handle on _fatal_error. (cherry picked from commit a608d2d)

… certain servers (#409) (cherry picked from commit a608d2d)

…e with certain servers (GH-409) * asyncio SSL contexts leak sockets after calling close with certain servers * cleanup _shutdown_timeout_handle on _fatal_error. (cherry picked from commit a608d2d)

…e with certain servers (GH-409) (#2062) * bpo-29406: asyncio SSL contexts leak sockets after calling close with certain servers (#409) (cherry picked from commit a608d2d) * [3.6] bpo-29406: asyncio SSL contexts leak sockets after calling close with certain servers (GH-409) * asyncio SSL contexts leak sockets after calling close with certain servers * cleanup _shutdown_timeout_handle on _fatal_error. (cherry picked from commit a608d2d)

…e with certain servers (GH-409) (#2063) * bpo-29406: asyncio SSL contexts leak sockets after calling close with certain servers (#409) (cherry picked from commit a608d2d) * [3.5] bpo-29406: asyncio SSL contexts leak sockets after calling close with certain servers (GH-409) * asyncio SSL contexts leak sockets after calling close with certain servers * cleanup _shutdown_timeout_handle on _fatal_error. (cherry picked from commit a608d2d)

…ose with certain servers (#409)" This reverts commit a608d2d.

…ing close with certain servers (GH-409) (#2062)" This reverts commit 6e14fd2.

…ing close with certain servers (GH-409) (#2063)" This reverts commit 1395c58.

…ing close with certain servers (GH-409) (#2062)" (#2112) This reverts commit 6e14fd2.

…ing close with certain servers (GH-409) (#2063)" (#2113) This reverts commit 1395c58.

…ose with certain servers (#409)" (#2111) This reverts commit a608d2d.

Bumps [aiohttp](https://github.com/aio-libs/aiohttp) from 3.6.2 to 3.7.2. - [Release notes](https://github.com/aio-libs/aiohttp/releases) - [Changelog](https://github.com/aio-libs/aiohttp/blob/master/CHANGES.rst) - [Commits](aio-libs/aiohttp@v3.6.2...v3.7.2) Signed-off-by: dependabot-preview[bot] <[email protected]> Co-authored-by: dependabot-preview[bot] <27856297+dependabot-preview[bot]@users.noreply.github.com>

the-knights-who-say-ni added the CLA not signed label Mar 3, 2017

fafhrd91 changed the title ~~asyncio SSL contexts leak sockets after calling close with certain servers~~ bpo-29406: asyncio SSL contexts leak sockets after calling close with certain servers Mar 3, 2017

the-knights-who-say-ni added CLA signed and removed CLA not signed labels Mar 3, 2017

fafhrd91 changed the base branch from 3.6 to master March 3, 2017 23:36

1st1 added needs backport to 3.5 labels Mar 3, 2017

1st1 reviewed Mar 3, 2017

View reviewed changes

1st1 reviewed Mar 6, 2017

View reviewed changes

1st1 self-assigned this Jun 10, 2017

1st1 reviewed Jun 10, 2017

View reviewed changes

fafhrd91 added 2 commits June 9, 2017 20:43

asyncio SSL contexts leak sockets after calling close with certain se…

b4eb6ae

…rvers

cleanup _shutdown_timeout_handle on _fatal_error

31d625a

1st1 merged commit a608d2d into python:master Jun 10, 2017

1st1 pushed a commit that referenced this pull request Jun 10, 2017

bpo-29406: asyncio SSL contexts leak sockets after calling close with…

009abbe

… certain servers (#409) (cherry picked from commit a608d2d)

1st1 pushed a commit that referenced this pull request Jun 10, 2017

bpo-29406: asyncio SSL contexts leak sockets after calling close with…

304b885

… certain servers (#409) (cherry picked from commit a608d2d)

1st1 added a commit that referenced this pull request Jun 11, 2017

Revert "bpo-29406: asyncio SSL contexts leak sockets after calling cl…

eb9bd66

…ose with certain servers (#409)" This reverts commit a608d2d.

1st1 mentioned this pull request Jun 11, 2017

Revert "bpo-29406: asyncio SSL contexts leak sockets after calling cl… #2111

Merged

1st1 added a commit that referenced this pull request Jun 11, 2017

Revert "[3.6] bpo-29406: asyncio SSL contexts leak sockets after call…

8ffe094

…ing close with certain servers (GH-409) (#2062)" This reverts commit 6e14fd2.

1st1 mentioned this pull request Jun 11, 2017

Revert "[3.6] bpo-29406: asyncio SSL contexts leak sockets after call… #2112

Merged

1st1 added a commit that referenced this pull request Jun 11, 2017

Revert "[3.5] bpo-29406: asyncio SSL contexts leak sockets after call…

71ada0b

…ing close with certain servers (GH-409) (#2063)" This reverts commit 1395c58.

1st1 mentioned this pull request Jun 11, 2017

Revert "[3.5] bpo-29406: asyncio SSL contexts leak sockets after call… #2113

Merged

1st1 added a commit that referenced this pull request Jun 11, 2017

Revert "[3.6] bpo-29406: asyncio SSL contexts leak sockets after call…

83d30bd

…ing close with certain servers (GH-409) (#2062)" (#2112) This reverts commit 6e14fd2.

1st1 added a commit that referenced this pull request Jun 11, 2017

Revert "[3.5] bpo-29406: asyncio SSL contexts leak sockets after call…

4e9dfe2

…ing close with certain servers (GH-409) (#2063)" (#2113) This reverts commit 1395c58.

1st1 added a commit that referenced this pull request Jun 11, 2017

Revert "bpo-29406: asyncio SSL contexts leak sockets after calling cl…

09663de

…ose with certain servers (#409)" (#2111) This reverts commit a608d2d.

bpo-29406: asyncio SSL contexts leak sockets after calling close with certain servers #409

bpo-29406: asyncio SSL contexts leak sockets after calling close with certain servers #409

Conversation

fafhrd91 commented Mar 3, 2017

the-knights-who-say-ni commented Mar 3, 2017

fafhrd91 commented Mar 3, 2017

fafhrd91 commented Mar 3, 2017

1st1 commented Mar 3, 2017

fafhrd91 commented Mar 3, 2017

1st1 commented Mar 3, 2017

fafhrd91 commented Mar 3, 2017

1st1 commented Mar 3, 2017

fafhrd91 commented Mar 3, 2017

fafhrd91 commented Mar 3, 2017

1st1 commented Mar 3, 2017

fafhrd91 commented Mar 3, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

fafhrd91 commented Mar 4, 2017

brettcannon commented Mar 4, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

1st1 commented Mar 12, 2017

fafhrd91 commented Mar 12, 2017

misg commented Apr 3, 2017

fafhrd91 commented Apr 3, 2017

misg commented Apr 3, 2017

1st1 commented Jun 9, 2017

fafhrd91 commented Jun 9, 2017

fafhrd91 commented Jun 9, 2017

fafhrd91 commented Jun 9, 2017

1st1 commented Jun 9, 2017

fafhrd91 commented Jun 9, 2017

1st1 commented Jun 10, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment