Messages in this thread | | | Date | Thu, 19 Feb 2009 20:41:58 -0500 (EST) | From | Nicolas Pitre <> | Subject | Re: [PATCH v2] SDIO driver for Marvell SoCs |
| |
On Thu, 19 Feb 2009, Andrew Morton wrote:
> --- a/drivers/mmc/host/mvsdio.c~sdio-driver-for-marvell-socs-update > +++ a/drivers/mmc/host/mvsdio.c > @@ -783,9 +783,9 @@ out: > gpio_free(host->gpio_write_protect); > if (host->base) > iounmap(host->base); > - if (host->res) > - release_resource(host->res); > } > + if (r) > + release_resource(host->res); > if (mmc) > mmc_free_host(mmc); > > _ > > Nope, this will deref a null pointer on that error path and wouldn't > release the right thing anyway.
Obviously. I'm just too stupid sometimes. A s/host->res/r/ in both places would be better.
> Something like this... > > --- a/drivers/mmc/host/mvsdio.c~sdio-driver-for-marvell-socs-update > +++ a/drivers/mmc/host/mvsdio.c > @@ -688,6 +688,7 @@ static int __init mvsd_probe(struct plat > > mmc = mmc_alloc_host(sizeof(struct mvsd_host), &pdev->dev); > if (!mmc) { > + release_resource(r); > ret = -ENOMEM; > goto out; > }
That's fine too.
Nicolas
| |