Re: kernel32: fix use-after-close in create_process

[Dmitry Timoshkov]

Dan Kegel <dank(a)kegel.com> wrote:
> Spotted by Guillaume ORTEGA.  He reports this fixes
> http://bugs.winehq.org/show_bug.cgi?id=24101
>
> Passes kernel32 tests.

That's not a use after close, client's fd passed in the server call is used
just as an association of the server side.

-- 
Dmitry.

kernel32: fix use-after-close in create_process

[Dan Kegel]

On Tue, Sep 21, 2010 at 5:30 AM, Dmitry Timoshkov
<dmitry(a)codeweavers.com> wrote:
> Dan Kegel <dank(a)kegel.com> wrote:
> > Spotted by Guillaume ORTEGA.  He reports this fixes
> > http://bugs.winehq.org/show_bug.cgi?id=24101
> >
> > Passes kernel32 tests.
>
> That's not a use after close, client's fd passed in the server call is used
> just as an association of the server side.

Guillaume claims in
http://bugs.winehq.org/show_bug.cgi?id=24101#c33
that wine_server_send_fd() returns before the fd is
actually sent, and delaying the close until after the
server call fixes the race.