Re: [BUG] ptraced process waiting on syscall may return kernel internal errnos

[Oleg Nesterov]

On 06/14, Rafael J. Wysocki wrote:
>
> Sorry for being late, I've just realized that you are discussing the freezer
> here. ;-)

my fault, I was going to cc you but forgot, sorry!

> On Wednesday, 13 June 2007 17:15, Oleg Nesterov wrote:
>
> > > @@ -105,7 +105,11 @@ static int recalc_sigpending_tsk(struct 
> > >  		set_tsk_thread_flag(t, TIF_SIGPENDING);
> > >  		return 1;
> > >  	}
> > > -	clear_tsk_thread_flag(t, TIF_SIGPENDING);
> > > +	/*
> > > +	 * We must never clear the flag in another thread, or in current
> > > +	 * when it's possible the current syscall is returning -ERESTART*.
> > > +	 * So we don't clear it here, and only callers who know they should do.
> > > +	 */
> > >  	return 0;
> > >  }
> > 
> > This breaks cancel_freezing(). Somehow we should clear TIF_SIGPENDING for
> > kernel threads. Otherwise we may have subtle failures if try_to_freeze_tasks()
> > fails.
> 
> Well, the only code path in which we'd want to call cancel_freezing() for kernel
> threads is when the freezing of kernel threads.  However, this only happens if
> one of the kernel threads declares itself as freezable and the fails to call
> try_to_freeze(), which is a bug.

But this happens? We know a lot of reasons why try_to_freeze() can fail just
because some kthread waits for already frozen task.

>                                   Thus I don't think that we need to worry
> about that case too much.

Well, we can have very subtle problems because a kernel thread may run with
TIF_SIGPENDING forever. This means in particualar that any wait_event_interruptible()
can't succeed. I think this is worse than explicit failure (like -ERESSTART... leak),
because it is hard to reproduce/debug.

> Moreover, I'm not sure that it's a good idea at all to send signals to kernel
> threads from the freezer, since in fact we only need to wake them up to make
> them call try_to_freeze() (after we've set TIF_FREEZE for them).

Yes! I completely agree.

Oleg.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/