[
https://issues.apache.org/jira/browse/HADOOP-12487?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14975022#comment-14975022
]
Colin Patrick McCabe commented on HADOOP-12487:
-----------------------------------------------
bq. The /dev/null & dup2 mechanism works on Solaris but it doesn't work on
Linux unfortunately, due to the brokenness in the handling of close() on a
socket that's being used in an accept().
Thanks for the link. Very interesting discussion. It's usually a good idea to
put background info on jiras in case the hyperlink goes away in the future. To
summarize the Linux kernel discussion in the thread you linked, it sounds like
{{close}} does not break out of {{accept}} on Linux, but {{shutdown}} does.
You refer to that as "brokenness," but some other people (including Eric
Duzamet and Al Viro) give reasons for the behavior and defend it.
bq. However having said that the /dev/null & dup2 mechanism works on Solaris I
can't come up with a race scenario where it's actually needed. DomainSocket
encapsulates the underlying FD and DomainSocket invalidates itself on close, so
I can't see how the FD can actually be used for anything, even if it is reused
by an open operation in a different thread. If you can come up with a scenario
involving DomainSocket then I'll investigate, thanks.
We have multiple threads operating on the same {{DomainSocket}} at once. If
thread #1 closes the domain socket, releasing the file descriptor number,
thread #2 in a separate, non-unix-file-descriptor thread opens a new
descriptor, and thread #3 does something with the old file descriptor number,
thread #3 may stomp on thread #2. I thought you understood this race based on
your comment here:
https://issues.apache.org/jira/browse/HADOOP-12487?focusedCommentId=14964205&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14964205
> DomainSocket.close() assumes incorrect Linux behaviour
> ------------------------------------------------------
>
> Key: HADOOP-12487
> URL: https://issues.apache.org/jira/browse/HADOOP-12487
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: net
> Affects Versions: 2.7.1
> Environment: Linux Solaris
> Reporter: Alan Burlison
> Assignee: Alan Burlison
> Attachments: shutdown.c
>
>
> I'm getting a test failure in TestDomainSocket.java, in the
> testSocketAcceptAndClose test. That test creates a socket which one thread
> waits on in DomainSocket.accept() whilst a second thread sleeps for a short
> time before closing the same socket with DomainSocket.close().
> DomainSocket.close() first calls shutdown0() on the socket before closing
> close0() - both those are thin wrappers around the corresponding libc socket
> calls. DomainSocket.close() contains the following comment, explaining the
> logic involved:
> {code}
> // Calling shutdown on the socket will interrupt blocking system
> // calls like accept, write, and read that are going on in a
> // different thread.
> {code}
> Unfortunately that relies on non-standards-compliant Linux behaviour. I've
> written a simple C test case that replicates the scenario above:
> # ThreadA opens, binds, listens and accepts on a socket, waiting for
> connections.
> # Some time later ThreadB calls shutdown on the socket ThreadA is waiting in
> accept on.
> Here is what happens:
> On Linux, the shutdown call in ThreadB succeeds and the accept call in
> ThreadA returns with EINVAL.
> On Solaris, the shutdown call in ThreadB fails and returns ENOTCONN. ThreadA
> continues to wait in accept.
> Relevant POSIX manpages:
> http://pubs.opengroup.org/onlinepubs/9699919799/functions/accept.html
> http://pubs.opengroup.org/onlinepubs/9699919799/functions/shutdown.html
> The POSIX shutdown manpage says:
> "The shutdown() function shall cause all or part of a full-duplex connection
> on the socket associated with the file descriptor socket to be shut down."
> ...
> "\[ENOTCONN] The socket is not connected."
> Page 229 & 303 of "UNIX System V Network Programming" say:
> "shutdown can only be called on sockets that have been previously connected"
> "The socket \[passed to accept that] fd refers to does not participate in the
> connection. It remains available to receive further connect indications"
> That is pretty clear, sockets being waited on with accept are not connected
> by definition. Nor is it the accept socket connected when a client connects
> to it, it is the socket returned by accept that is connected to the client.
> Therefore the Solaris behaviour of failing the shutdown call is correct.
> In order to get the required behaviour of ThreadB causing ThreadA to exit the
> accept call with an error, the correct way is for ThreadB to call close on
> the socket that ThreadA is waiting on in accept.
> On Solaris, calling close in ThreadB succeeds, and the accept call in ThreadA
> fails and returns EBADF.
> On Linux, calling close in ThreadB succeeds but ThreadA continues to wait in
> accept until there is an incoming connection. That accept returns
> successfully. However subsequent accept calls on the same socket return EBADF.
> The Linux behaviour is fundamentally broken in three places:
> # Allowing shutdown to succeed on an unconnected socket is incorrect.
> # Returning a successful accept on a closed file descriptor is incorrect,
> especially as future accept calls on the same socket fail.
> # Once shutdown has been called on the socket, calling close on the socket
> fails with EBADF. That is incorrect, shutdown should just prevent further IO
> on the socket, it should not close it.
> The real issue though is that there's no single way of doing this that works
> on both Solaris and Linux, there will need to be platform-specific code in
> Hadoop to cater for the Linux brokenness.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
