[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <Pine.LNX.4.64.0809011521230.9596@pc-004.diku.dk>
Date: Mon, 1 Sep 2008 15:23:25 +0200 (CEST)
From: Julia Lawall <julia@...u.dk>
To: Takashi Iwai <tiwai@...e.de>
Cc: Julien Brunel <brunel@...u.dk>, perex@...ex.cz,
linux-kernel@...r.kernel.org, kernel-janitors@...r.kernel.org
Subject: Re: [PATCH] sound/arm: Bad NULL test
On Mon, 1 Sep 2008, Takashi Iwai wrote:
> At Mon, 1 Sep 2008 14:30:29 +0200 (CEST),
> Julia Lawall wrote:
> >
> > On Mon, 1 Sep 2008, Takashi Iwai wrote:
> >
> > > At Mon, 1 Sep 2008 10:59:54 +0200,
> > > Julien Brunel wrote:
> > > >
> > > > From: Julien Brunel <brunel@...u.dk>
> > > >
> > > > In case of error, the function aaci_init_card returns an ERR pointer,
> > > > but never returns a NULL pointer. We have noticed a bad NULL test,
> > > > which comes after a call to this function. Rather than doing an IS_ERR
> > > > test, we suggest to duplicate the label out: one label for the case where
> > > > aaci_init_card returns a valid pointer, and another for the case where
> > > > aaci_init_card returns an ERR pointer.
> > > >
> > > > The semantic match that finds this problem is as follows:
> > > > (http://www.emn.fr/x-info/coccinelle/)
> > > >
> > > > // <smpl>
> > > > @match_bad_null_test@
> > > > expression x, E;
> > > > statement S1,S2;
> > > > @@
> > > > x = aaci_init_card(...)
> > > > ... when != x = E
> > > > * if (x != NULL)
> > > > S1 else S2
> > > > // </smpl>
> > > >
> > > > Signed-off-by: Julien Brunel <brunel@...u.dk>
> > > > Signed-off-by: Julia Lawall <julia@...u.dk>
> > >
> > > The fix below is simpler. Could you check whether it's OK?
> >
> > It is indeed simpler, and looks correct, but it seems a little odd to take
> > a value that can never be NULL and set it to NULL just to avoid changing a
> > test. Another alternative would be to leave the value as it is, and put
> > an IS_ERR test at the out label. But the value of the test is statically
> > determined by the goto that reaches it, so the original patch proposes
> > just getting rid of the test completely.
>
> OTOH, double labels are pretty ugly and hard to follow.
> Maybe the patch like below is a bit cleaner.
OK, this seems like a reasonable compromise. The function aaci_init_card
is indeed only used in one place, and -ENOMEM is the only possible error
value. So perhaps it is just as easy to create that where it is called
rather than in aaci_init_card itself.
julia
> Takashi
>
> diff --git a/sound/arm/aaci.c b/sound/arm/aaci.c
> index b0a4744..89096e8 100644
> --- a/sound/arm/aaci.c
> +++ b/sound/arm/aaci.c
> @@ -999,7 +999,7 @@ static struct aaci * __devinit aaci_init_card(struct amba_device *dev)
> card = snd_card_new(SNDRV_DEFAULT_IDX1, SNDRV_DEFAULT_STR1,
> THIS_MODULE, sizeof(struct aaci));
> if (card == NULL)
> - return ERR_PTR(-ENOMEM);
> + return NULL;
>
> card->private_free = aaci_free_card;
>
> @@ -1083,8 +1083,8 @@ static int __devinit aaci_probe(struct amba_device *dev, void *id)
> return ret;
>
> aaci = aaci_init_card(dev);
> - if (IS_ERR(aaci)) {
> - ret = PTR_ERR(aaci);
> + if (!aaci) {
> + ret = -ENOMEM;
> goto out;
> }
>
>
--
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