Re: [Xen-devel] [PATCH 23/24] libxl: child processes cleanups

[Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx>]

Ian Campbell writes ("Re: [Xen-devel] [PATCH 23/24] libxl: child processes 
cleanups"):
> On Mon, 2012-04-16 at 18:18 +0100, Ian Jackson wrote:
> > Abolish libxl_fork.  Its only callers were in xl.  Its functionality
> > is now moved elsewhere, as follows:
...
> >  static inline int libxl__ev_child_inuse(libxl__ev_child *childw_out)
> >                  { return childw_out->pid >= 0; }
> >  
> > +/* Useable (only) in the child to once more make the ctx useable for
> > + * xenstore operations.
> 
> Specifically "the child" is the middle child of a spawn? Otherwise the
> constraint must be something like "before any threads are created in the
> new process", or something like that?

The fact that raw fork() may be used in the child created by
libxl__ev_child_inuse isn't documented.  It would be possible to
document this but the set of restrictions on the behaviours of the
middle child and any resulting grandchildren are rather complex.

I think it might be better to draw a veil over this and leave it as a
special piece of knowledge implicit in the implementation of
libxl__spawn_*.

In practice the use of _xenstore_reopen in the middle child is fine
provided that the middle child doesn't then fork _and_ then use
libxl's xenstore functions in both the middle child and the
grandchild.

Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel