[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20070620111415.GA91@tv-sign.ru>
Date: Wed, 20 Jun 2007 15:14:15 +0400
From: Oleg Nesterov <oleg@...sign.ru>
To: Benjamin Herrenschmidt <benh@...nel.crashing.org>
Cc: Linus Torvalds <torvalds@...ux-foundation.org>,
Davide Libenzi <davidel@...ilserver.org>,
Nicholas Miell <nmiell@...cast.net>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: Fix signalfd interaction with thread-private signals
On 06/20, Benjamin Herrenschmidt wrote:
>
> On Tue, 2007-06-19 at 18:06 +0400, Oleg Nesterov wrote:
> > On 06/19, Benjamin Herrenschmidt wrote:
> > >
> > > On Tue, 2007-06-19 at 13:14 +0400, Oleg Nesterov wrote:
> > >
> > > > The commited "Fix signalfd interaction with thread-private signals"
> > > > (commit caec4e8dc85e0644ec24aeb36285e1ba02da58cc) doesn't implement
> > > > this.
> > >
> > > Indeed, if you want what Davide described, you need to also change
> > > signalfd side. The patch I did merely prevents another thread from
> > > dequeuing somebody else private signals.
> >
> > Yes I see, but why do we need this change? Yes, we can dequeue SIGSEGV
> > from another thread. Just don't do it if you have a handler for SIGSEGV?
>
> Well, for such synchronous signals, it's a fairly stupid idea,
> especially since you can't predict who will get it. Signals such as SEGV
> are forced-in, which means they are force-unblocked. Thus, you can't
> know for sure whome of signalfd or the target thread will get it first,
> depending on who catches the siglock first I suppose. In one case,
> you'll manage to steal it, in the other, you'll thread will be killed.
Yes. As I said, I think this falls into the "just don't do that" category.
But nothing bad happens from the kernel POV.
Also, suppose that some thread does
for (;;)
signal(SIGSEGV, SIG_IGN);
Now we have the same situation. do_sigaction() can steal SIGSEGV from
another thread.
Perhaps, the proposed behaviour
> Multiple threads can wait on the signalfd. Each one will dequeue either
> its own private signals (tsk->pending) or the process shared ones
> (tsk->signal->shared_pending).
is more convenient, I can't judge. If we implement this, sys_signalfd(-1) in
essence means "attach to the thread group, not current". This also makes sense.
But what we have now (with this patch applied) is a bit strange, imho.
Oleg.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists