Segmentation fault due to double free for archetype.

Corinna Vinschen corinna-cygwin@cygwin.com
Mon Jan 17 12:11:46 GMT 2022


On Jan 17 20:41, Takashi Yano wrote:
> On Mon, 17 Jan 2022 12:01:51 +0100
> Corinna Vinschen wrote:
> > On Jan 15 19:20, Takashi Yano wrote:
> > > I also found the following patch fixes the issue. Is this the
> > > right thing?
> > > 
> > > diff --git a/winsup/cygwin/fhandler.cc b/winsup/cygwin/fhandler.cc
> > > index fc7c0422e..e51208117 100644
> > > --- a/winsup/cygwin/fhandler.cc
> > > +++ b/winsup/cygwin/fhandler.cc
> > > @@ -441,7 +441,7 @@ fhandler_base::open_with_arch (int flags, mode_t mode)
> > >  	|| open (flags, mode & 07777)))
> > >      {
> > >        if (archetype)
> > > -	delete archetype;
> > > +	cygheap->fdtab.delete_archetype (archetype);
> > >      }
> > >    else if (archetype)
> > >      {
> > 
> > Good catch!  I think this is basically ok, but you have to check the
> > usecount, i. e.
> > 
> >   if (archetype && archetype_usecount (-1) == 0)
> >     cygheap->fdtab.delete_archetype (archetype);
> > 
> > Does that sound right?
> 
> Where is archetype->usecount is incremented?

In fhandler_base::open_with_arch.

> It seems that
> archetype->usecount is zero here. archetype->usecount is not
> incremented around line 672 in dtable.cc when archetype is
> created by fh->clone().
> 
>   else
>     {
>       if (!fh->get_name ())
>         fh->set_name (fh->dev ().native ());
>       fh->archetype = fh->clone ();
>       debug_printf ("created an archetype (%p) for %s(%d/%d)", fh->archetype, fh->get_name (), fh->dev ().get_major (), fh->dev ().get_minor ());
>       fh->archetype->archetype = NULL;
>       *cygheap->fdtab.add_archetype () = fh->archetype;
>     }

Right, but if open isn't called, because you already have an archetype
at this point:

  if (!(res = (archetype && archetype->io_handle)
        || open (flags, mode & 07777)))

Then the archetype is one already created by a former open_with_arch
call and then you delete an archetype which is still in use, no?


Corinna


More information about the Cygwin mailing list