Re: [BUG] tty: Userland can create hung tasks

[Tejun Heo <tj@kernel.org>]

Hello, Alan.

On Wed, Nov 01, 2017 at 05:09:08PM +0000, Alan Cox wrote:
> On Tue, 31 Oct 2017 19:06:51 -0700
> Tejun Heo <tj@kernel.org> wrote:
> 
> > Hello,
> > 
> > tty hangup code doesn't mark the console as being HUPed for, e.g.,
> > /dev/console and that can put the session leader trying to
> > disassociate from the controlling terminal in an indefinite D sleep.
> > 
> > Looking at the code, I have no idea why some tty devices are never
> > marked being hung up.  It *looks* intentional and dates back to the
> > git origin but I couldn't find any clue.  The following patch is a
> > workaround which fixes the observed problem but definitely isn't the
> > proper fix.
> 
> Two reasons
> 
> 1. It broke the serial consoles because they would hang up and close down
> the hardware. With tty_port that *should* be fixable properly for any
> cases remaining.
> 
> 2. The console layer was (and still is) completely broken and doens't
> refcount properly. So if you turn on console hangups it breaks (as indeed
> does freeing consoles and half a dozen other things).
> 
> Sadly it seems that nobody cares about the console.

Hmm... so, does something like the posted workaround look reasonable
to you?  It doesn't fix the actual hangup issue but gets rid of
processes getting stuck in D sleep indefinitely.

Thanks.

-- 
tejun