[Drbd-dev] [PATCH 2/2] exec: allow core_pipe recursion check to look for a value of 1 rather than 0 (v2)

[Neil Horman]

On Mon, Feb 01, 2010 at 11:29:36AM +0100, Oleg Nesterov wrote:
> On 01/31, Neil Horman wrote:
> >
> > On Sun, Jan 31, 2010 at 04:50:01PM +0100, Oleg Nesterov wrote:
> > > On 01/29, Neil Horman wrote:
> > > >
> > > >  void do_coredump(long signr, int exit_code, struct pt_regs *regs)
> > > >  {
> > > > ...
> > > > -		if (call_usermodehelper_pipe(helper_argv[0], helper_argv, NULL,
> > > > -				&cprm.file)) {
> > > > +		cprm.file = NULL;
> > >
> > > it is already NULL,
> > >
> > Are we sure, it was declared on the stack.
> 
> it must be NULL, or compiler is buggy. it was declared as "= { ... }".
> 
> > I think its safer to ensure that its
> > NULL.
> 
> OK, agreed. I mentioned this just in case.
> 
> > > > +		if (call_usermodehelper_fns(helper_argv[0], helper_argv, NULL,
> > > > +					    UMH_WAIT_EXEC, umh_pipe_setup,
> > > > +					    NULL, &cprm)) {
> > > > +			if (cprm.file)
> > > > +				filp_close(cprm.file, NULL);
> > >
> > > Hmm. Looks like this change fixes the bug by accident.
> > >
> > > Before this patch, I think we leak info->stdin if kernel_thread() fails
> > > in __call_usermodehelper() pathes.
> > >
> > I think we did that in call_usermodehelper_pipe.
> 
> Afaics, no. Well yes, call_usermodehelper_pipe() closes write_pipe,
> but I meant nobody closes read_pipe, info->stdin, if we fail before
> ____call_usermodehelper() is called.
> 
> > > Completely off-topic, but I think __call_usermodehelper(UMH_NO_WAIT) is
> > > buggy. if kernel_thread() failes it should do call_usermodehelper_freeinfo().
> > > Also, UMH_WAIT_EXEC should set ->retval in this case.
> > >
> > I went down that path last time I changed this code, Andrew and I decided that
> > yes it was buggy, but someone (can't recall how) smacked me around a bit and
> > explained how it worked (some odd artifact behavior of the scheduler).  Its in
> > the lkml archives if you want to get the whole story.
> 
> Hmm. I strongly believe this is buggy, and the scheduler can't help in any
> way. Fortunately, kernel_thread() must "never" fail...
> 
> Oh. And in theory, it is better to change wait_for_helper(). It should
> do allow_signal(SIGCHLD) after kernel_thread(). Otherwise, kernel_thread()
> can fail if user-space sends SIGCHLD to the forking thread.
> 

Commit 95e0d86badc410d525ea7218fd32df7bfbf9c837 has the discussion from the
previous time that I messed with this code.  Not sure how closely it relates,
but...

Neil

>