Re: [PATCH] staging: dgnc: check if kzalloc fails in dgnc_tty_init()

[Giedrius Statkevičius]

On Tue, 7 Apr 2015, Sudip Mukherjee wrote:

> On Tue, Apr 07, 2015 at 01:26:32PM +0300, Giedrius StatkeviÄius wrote:
> > kzalloc() could fail so add a check and return -ENOMEM if it does that gets
> > propogated to the pci layer
> > 
> > Signed-off-by: Giedrius StatkeviÄius <giedrius.statkevicius@xxxxxxxxx>
> > ---
> >  drivers/staging/dgnc/dgnc_tty.c | 2 ++
> >  1 file changed, 2 insertions(+)
> > 
> > diff --git a/drivers/staging/dgnc/dgnc_tty.c b/drivers/staging/dgnc/dgnc_tty.c
> > index 61d5a8e..23337da 100644
> > --- a/drivers/staging/dgnc/dgnc_tty.c
> > +++ b/drivers/staging/dgnc/dgnc_tty.c
> > @@ -311,6 +311,8 @@ int dgnc_tty_init(struct dgnc_board *brd)
> >  			 * interrupt context, and there are no locks held.
> >  			 */
> >  			brd->channels[i] = kzalloc(sizeof(*brd->channels[i]), GFP_KERNEL);
> > +			if (!brd->channels[i])
> > +				return -ENOMEM;
> won't this create memory leak ?
> suppose you have brd->nasync = 3 
> and kzalloc fails when i=2, and you return -ENOMEM,
> then what happens to the memory already allocated to brd->channels[0]
> and brd->channels[1] ?
> 
> regards
> sudip
> 
Didn't think about that, sorry. It will cause a memory leak indeed. I'll send a
v2 that creates a label that frees all successful kzalloc() before returning
-ENOMEM.

Su pagarba / Regards,
Giedrius