[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <35cab334-0856-44e1-b18e-22668011b429@gmail.com>
Date: Fri, 14 Dec 2018 14:47:50 -0800
From: Frank Rowand <frowand.list@...il.com>
To: Rob Herring <robh+dt@...nel.org>
Cc: mwb@...ux.vnet.ibm.com,
linuxppc-dev <linuxppc-dev@...ts.ozlabs.org>,
Michael Ellerman <mpe@...erman.id.au>,
Tyrel Datwyler <tyreld@...ux.vnet.ibm.com>,
tlfalcon@...ux.vnet.ibm.com, minkim@...ibm.com,
devicetree@...r.kernel.org,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
Frank Rowand <frowand.list@...il.com>
Subject: Re: [PATCH 1/2] of: of_node_get()/of_node_put() nodes held in phandle
cache
On 12/14/18 9:15 AM, Rob Herring wrote:
> On Fri, Dec 14, 2018 at 12:43 AM <frowand.list@...il.com> wrote:
>>
>> From: Frank Rowand <frank.rowand@...y.com>
>>
>> The phandle cache contains struct device_node pointers. The refcount
>> of the pointers was not incremented while in the cache, allowing use
>> after free error after kfree() of the node. Add the proper increment
>> and decrement of the use count.
>
> Since we pre-populate the cache at boot, all the nodes will have a ref
> count and will never be freed unless we happen to repopulate the whole
> cache. That doesn't seem ideal. The node pointer is not "in use" just
> because it is in the cache.
>
> Rob
>
This patch also adds of_node_put() so that the refcount will go to zero
when the node is removed as part of an overlay remove, if the node was
added by an overlay.
Patch 2/2 adds the free cache entry call to __of_detach_node(), so the
refcount will go to zero when the node is removed for dynamic use cases
other than overlays. (For overlays, all nodes are instead removed from
the cache before __of_detach_node() is called.)
-Frank
Powered by blists - more mailing lists