On Fri, 7 Aug 2009, Daniel K. wrote:
> Julia Lawall wrote:
> > --- a/arch/powerpc/sysdev/fsl_rio.c
> > +++ b/arch/powerpc/sysdev/fsl_rio.c
> > @@ -1057,6 +1057,10 @@ int fsl_rio_setup(struct of_device *dev)
> > law_start, law_size);
> >
> > ops = kmalloc(sizeof(struct rio_ops), GFP_KERNEL);
> > + if (!ops) {
> > + rc = -ENOMEM;
> > + goto err_ops;
> > + }
> > ops->lcread = fsl_local_config_read;
> > ops->lcwrite = fsl_local_config_write;
> > ops->cread = fsl_rio_config_read;
> > @@ -1064,6 +1068,10 @@ int fsl_rio_setup(struct of_device *dev)
> > ops->dsend = fsl_rio_doorbell_send;
> >
> > port = kzalloc(sizeof(struct rio_mport), GFP_KERNEL);
> > + if (!port) {
> > + rc = -ENOMEM;
> > + goto err_port;
> > + }
> > port->id = 0;
> > port->index = 0;
> >
> > @@ -1071,7 +1079,7 @@ int fsl_rio_setup(struct of_device *dev)
> > if (!priv) {
> > printk(KERN_ERR "Can't alloc memory for 'priv'\n");
> > rc = -ENOMEM;
> > - goto err;
> > + goto err_priv;
> > }
> >
> > INIT_LIST_HEAD(&port->dbells);
> > @@ -1169,13 +1177,15 @@ int fsl_rio_setup(struct of_device *dev)
> >
> > return 0;
> > err:
> > - if (priv)
> > - iounmap(priv->regs_win);
> > - kfree(ops);
> > + iounmap(priv->regs_win);
> > +err_priv:
> > kfree(priv);
> > +err_port:
> > kfree(port);
> > +err_ops:
> > + kfree(ops);
> > return rc;
>
> There seems to be a goto-off-by-one error here.
>
> If xxxx = kxalloc() fails, you goto err_xxxx, and do a kfree(xxxx) where xxxx
> is
> already proven to be NULL.
>
> Is there a reason for this that eludes me?
No, I messed up... I will fix it.
julia
> I'd expect that last hunk to look something like
>
> @@ -1169,13 +1177,15 @@ int fsl_rio_setup(struct of_device *dev)
>
> return 0;
> err:
> - if (priv)
> - iounmap(priv->regs_win);
> - kfree(ops);
> + iounmap(priv->regs_win);
> kfree(priv);
> +err_priv:
> kfree(port);
> +err_port:
> + kfree(ops);
> +err_ops:
> return rc;
> }
>
>
> Daniel K.
> --
> To unsubscribe from this list: send the line "unsubscribe kernel-janitors" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
--
To unsubscribe from this list: send the line "unsubscribe kernel-janitors" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
