Re: [PATCH] kernel/resource: Invalid memory access in __release_resource

From: Bjorn Helgaas
Date: Mon Apr 20 2015 - 16:36:56 EST


On Mon, Apr 20, 2015 at 10:24:25PM +0200, Ricardo Ribalda Delgado wrote:
> Hi Bjorn!
>
> Thanks for your promtly response.
>
> On Mon, Apr 20, 2015 at 9:28 PM, Bjorn Helgaas <bhelgaas@xxxxxxxxxx> wrote:
> > [+cc Grant (author of ac80a51e2ce5)]
> >
> > Hi Ricardo,
> >
> > On Mon, Apr 20, 2015 at 06:22:52PM +0200, Ricardo Ribalda Delgado wrote:
> >>
> >> If of_platform_depopulate is called later, resource->parent is
> >> accessed (Offset 0x30 of address 0), causing a kernel error.
> >
> > Interesting; how'd you find this? It looks like the
> > of_platform_depopulate() code has been this way for a long time, so we
> > must be doing something new that makes us trip over this now. More
> > analysis below...
>
> I have an out of tree driver that dynamically adds devices to the device tree.
>
> It was developed before the dynamic_of and dt_overlays existed. Now I
> am porting my code to the new interfaces available. I am trying to do
> it small steps.
>
> First step was being able to depopulate a previously loaded device
> tree. Old, code was calling of_platform_populate, so calling
> of_platform_depopulate looked like the right choice. Unfortunately
> everything crashed, and it turned out that this was the issue.
>
> On my defense I would say, that the plan is to make this driver
> public, once the hardware is stabilized and sold to the public.

No need to defend yourself; to me this looks like a bug in the
of_platform code, so it's a good thing you tripped over it :)

The obvious bug is the NULL pointer dereference. The not-quite-so-
obvious bug is that I think the lack of insert_resource() means the
resource tree (/proc/iomem, /proc/ioports) is missing some useful
information.

> > From reading drivers/base/platform.c, it looks like the intent is
> > that platform device users would use these interfaces:
>
> I can take a look to modify OF to use insert_resource(), but I still
> think that no matter what, we should add this extra check, like the
> propossed patch or maybe with a BUG_ON()....

I think it would be nicer to make OF use platform_device_add_resources()
and platform_device_add() because then there's less duplication of code.
But Grant might have had a reason for avoiding that.

Bjorn
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/