#4514: IO manager can deadlock if a file descriptor is closed behind its back
---------------------------------+------------------------------------------
Reporter: adept | Owner: bos
Type: bug | Status: patch
Priority: normal | Component: Runtime System
Version: 7.0.1 | Keywords:
Testcase: | Blockedby:
Os: Linux | Blocking:
Architecture: Unknown/Multiple | Failure: Incorrect result at runtime
---------------------------------+------------------------------------------
Comment(by bos):
Johan, I think you're right about the need to invoke the callbacks. If you
look at the {{{closeFd}}} code in
[http://hackage.haskell.org/trac/ghc/attachment/ticket/4514/base-4514.patch
base-4514.patch], you'll see that there are two places we could invoke the
callbacks:
1. While we have the {{{MVar}}} named {{{emFds}}} held.
1. After we've released it.
There's an unavoidable race here at present: another thread could have
opened a file or socket, getting the same file descriptor, and we've no
way to tell our callee that our file descriptor has been closed, so it
could read/write the wrong file descriptor.
To deal with this, I think we need to add the equivalents of {{{POLLHUP}}}
and {{{POLLRDHUP}}} to the events that we can send to threads, call 'em
{{{evtWriteHup}}} and {{{evtReadHup}}}. If you asked for {{{evtWrite}}}
and someone closed your file descriptor, we'll give you {{{evtWriteHup}}}.
If you asked for {{{evtRead}}}, we'll give you {{{evtReadHup}}}.
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/4514#comment:20>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler
_______________________________________________
Glasgow-haskell-bugs mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
