Quoting Dwight Engen ([email protected]):
> On Fri, 7 Mar 2014 17:27:38 -0600
> Serge Hallyn <[email protected]> wrote:
> 
> > Quoting Dwight Engen ([email protected]):
> > > This came up as I was chasing an fd leak with valgrind that showed
> > > up in the fork()ed child. There is no reason we can't just close
> > > the fd before the fork which also simplifies the code.
> > 
> > There is though - the ongoing_create() functions checks the pid of the
> > task holding the lock on the /partial file.  If you close it
> > immediately, then a racing create or start will assume that the create
> > failed and remove the container.
> > 
> > So unfortunately we can't close that fd until we're done.
> 
> Got it, thanks, sorry for the noise. I missed the unlock on close
> thing. Do you think its worthwhile to close it just in the child since
> the lock won't be inherited across fork?

Yup, that sounds good, thanks.

-serge
_______________________________________________
lxc-devel mailing list
[email protected]
http://lists.linuxcontainers.org/listinfo/lxc-devel

Reply via email to