Re: [PATCH v2 09/19] staging/speakup: Use tty_ldisc_ref() for paste kworker

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Sun, 2016-01-10 at 21:40 -0800, Peter Hurley wrote:
> On 01/10/2016 04:25 PM, Peter Hurley wrote:
> > On 01/10/2016 03:16 PM, Ben Hutchings wrote:
> > > On Sat, 2016-01-09 at 20:41 -0800, Peter Hurley wrote:
> > > > As the function documentation for tty_ldisc_ref_wait() notes, it is
> > > > only callable from a tty file_operations routine; otherwise there
> > > > is no guarantee the ref won't be NULL.
> > > > 
> > > > The key difference with the VT's paste_selection() is that is an ioctl,
> > > > where __speakup_paste_selection() is completely asynch kworker, kicked
> > > > off from interrupt context.
> > > > 
> > > > Fixes: 28a821c30688 ("Staging: speakup: Update __speakup_paste_selection()
> > > >        tty (ab)usage to match vt")
> > > > Cc: <stable@xxxxxxxxxxxxxxx>
> > > > 
> > > > Signed-off-by: Peter Hurley <peter@xxxxxxxxxxxxxxxxxx>
> > > > ---
> > > >  drivers/staging/speakup/selection.c | 4 +++-
> > > >  1 file changed, 3 insertions(+), 1 deletion(-)
> > > > 
> > > > diff --git a/drivers/staging/speakup/selection.c b/drivers/staging/speakup/selection.c
> > > > index aa5ab6c..86c0b9a 100644
> > > > --- a/drivers/staging/speakup/selection.c
> > > > +++ b/drivers/staging/speakup/selection.c
> > > > @@ -142,7 +142,9 @@ static void __speakup_paste_selection(struct work_struct *work)
> > > >  	struct tty_ldisc *ld;
> > > >  	DECLARE_WAITQUEUE(wait, current);
> > > >  
> > > > -	ld = tty_ldisc_ref_wait(tty);
> > > > +	ld = tty_ldisc_ref(tty);
> > > > +	if (!ld)
> > > > +		return;
> > > >  	tty_buffer_lock_exclusive(&vc->port);
> > > >  
> > > >  	add_wait_queue(&vc->paste_wait, &wait);
> > > 
> > > This leaks a reference to the tty.  Instead of returning directly, I
> > > think you need to add a label and goto the tty_kref_put() at the bottom
> > > of the function.
> > 
> > Ugh, speakup_paste_selection() is a worse hack than I thought it was.
> 
> What if the kworker has already been scheduled but not run? Leaky reference
> anyway.

I don't think so - the cmpxchg() should ensure that only one paste is
outstanding.

> What guarantees that the kref is gettable to begin with and isn't incrementing
> from 0?

Surely it's a bug in the caller of speakup_paste_selection() if it
doesn't hold a reference?

Ben.

> This isn't how tty krefs work.
> 
> I'll fix the patch to drop the kref but this is broken anyway.
> 
> Regards,
> Peter Hurley
> 
-- 
Ben Hutchings
Q.  Which is the greater problem in the world today, ignorance or apathy?
A.  I don't know and I couldn't care less.

Attachment: signature.asc
Description: This is a digitally signed message part


[Index of Archives]     [Linux Kernel]     [Kernel Development Newbies]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite Hiking]     [Linux Kernel]     [Linux SCSI]