[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <5F69C18A.2070800@huawei.com>
Date: Tue, 22 Sep 2020 17:19:06 +0800
From: Jing Xiangfeng <jingxiangfeng@...wei.com>
To: Dan Carpenter <dan.carpenter@...cle.com>
CC: <mporter@...nel.crashing.org>, <alex.bou9@...il.com>,
<akpm@...ux-foundation.org>, <gustavoars@...nel.org>,
<jhubbard@...dia.com>, <keescook@...omium.org>,
<madhuparnabhowmik10@...il.com>, <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] rapidio: fix the missed put_device() for
rio_mport_add_riodev
On 2020/9/22 16:04, Dan Carpenter wrote:
> On Tue, Sep 22, 2020 at 03:25:25PM +0800, Jing Xiangfeng wrote:
>> rio_mport_add_riodev() misses to call put_device() when the device
>> already exists. Add the missed function call to fix it.
>>
> Looks good.
>
> Reviewed-by: Dan Carpenter <dan.carpenter@...cle.com>
>
> I notice that rio_mport_del_riodev() has a related bug.
>
> 1802 err = rio_add_device(rdev);
> 1803 if (err)
> 1804 goto cleanup;
> 1805 rio_dev_get(rdev);
> ^^^^^^^^^^^^^^^^^
> This calls get_device(&rdev->dev);
>
> 1806
> 1807 return 0;
> 1808 cleanup:
> 1809 kfree(rdev);
> 1810 return err;
> 1811 }
> 1812
> 1813 static int rio_mport_del_riodev(struct mport_cdev_priv *priv, void __user *arg)
> 1814 {
> 1815 struct rio_rdev_info dev_info;
> 1816 struct rio_dev *rdev = NULL;
> 1817 struct device *dev;
> 1818 struct rio_mport *mport;
> 1819 struct rio_net *net;
> 1820
> 1821 if (copy_from_user(&dev_info, arg, sizeof(dev_info)))
> 1822 return -EFAULT;
> 1823 dev_info.name[sizeof(dev_info.name) - 1] = '\0';
> 1824
> 1825 mport = priv->md->mport;
> 1826
> 1827 /* If device name is specified, removal by name has priority */
> 1828 if (strlen(dev_info.name)) {
> 1829 dev = bus_find_device_by_name(&rio_bus_type, NULL,
> 1830 dev_info.name);
> 1831 if (dev)
> 1832 rdev = to_rio_dev(dev);
>
> This path takes a second get_device(&rdev->dev);
>
> 1833 } else {
> 1834 do {
> 1835 rdev = rio_get_comptag(dev_info.comptag, rdev);
> 1836 if (rdev && rdev->dev.parent == &mport->net->dev &&
> 1837 rdev->destid == dev_info.destid &&
> 1838 rdev->hopcount == dev_info.hopcount)
> 1839 break;
>
> This path does not call get_device().
Add calling rio_dev_get() in this path? like the following changes:
static int rio_mport_del_riodev(struct mport_cdev_priv *priv, void
__user *arg)
rdev = rio_get_comptag(dev_info.comptag, rdev);
if (rdev && rdev->dev.parent == &mport->net->dev &&
rdev->destid == dev_info.destid &&
- rdev->hopcount == dev_info.hopcount)
+ rdev->hopcount == dev_info.hopcount) {
+ rio_dev_get(rdev);
break;
+ }
} while (rdev);
}
>
> 1840 } while (rdev);
> 1841 }
> 1842
> 1843 if (!rdev) {
> 1844 rmcd_debug(RDEV,
> 1845 "device name:%s ct:0x%x did:0x%x hc:0x%x not found",
> 1846 dev_info.name, dev_info.comptag, dev_info.destid,
> 1847 dev_info.hopcount);
> 1848 return -ENODEV;
> 1849 }
> 1850
> 1851 net = rdev->net;
> 1852 rio_dev_put(rdev);
>
> This drops a reference.
>
> 1853 rio_del_device(rdev, RIO_DEVICE_SHUTDOWN);
>
> This drops a second reference. So presumably deleting by component tag
> will lead to a use after free.
Indeed, it is a mismatched reference.
>
> 1854
> 1855 if (list_empty(&net->devices)) {
> 1856 rio_free_net(net);
> 1857 mport->net = NULL;
> 1858 }
> 1859
> 1860 return 0;
> 1861 }
>
> regards,
> dan carpenter
> .
>
Powered by blists - more mailing lists