[PATCH] mtd: spi-nor: intel-spi: Do not try to make the SPI flash chip writable

[Daniel Gutson]

On Sun, Aug 16, 2020 at 5:42 AM Arnd Bergmann <arnd at arndb.de> wrote:
>
> On Thu, Aug 13, 2020 at 11:40 PM Daniel Gutson <daniel at eclypsium.com> wrote:
> >
> > On Thu, Aug 13, 2020 at 12:41 PM Arnd Bergmann <arnd at arndb.de> wrote:
> > >
> > > On Tue, Aug 4, 2020 at 11:26 PM Daniel Gutson <daniel at eclypsium.com> wrote:
> > > > On Tue, Aug 4, 2020 at 5:46 PM Arnd Bergmann <arnd at arndb.de> wrote:
> > > > > > But wait, Mika, the author of the file, asked earlier not to remove
> > > > > > the module parameter of intel-spi, and just remove the unconditional
> > > > > > attempt to turn the chip writable in intle-spi-pci.
> > > > >
> > > > > Yes, and I think that is fine (aside from the inconsistency with bay trail
> > > > > that you have not commented on),
> > > >
> > > > There are two inconsistencies before any of my patches:
> > > > 1) in intel-spi.c: uses the module parameter only for bay trail.
> > > > 2) intel-spi.c uses a module parameter whereas intel-spi-pci doesn't
> > >
> > > Neither of these matches what I see in the source code. Please
> > > check again.
> > >
> > > Once more: intel-spi.c has a module parameter that controls writing
> > > to the device regardless of the back-end (platform or pci), purely
> > > in software.
> >
> > If I understand you correctly, this is not what I see:
> > If the deviceID is listed in intel-spi-pci.c
> > (https://elixir.bootlin.com/linux/latest/source/drivers/mtd/spi-nor/controllers/intel-spi-pci.c#L66)
> > then intel_spi_pci_probe will be called, where it unconditionally will
> > try to make the chip writable
> > https://elixir.bootlin.com/linux/latest/source/drivers/mtd/spi-nor/controllers/intel-spi-pci.c#L44
> > These devices correspond to the BXT and CNL devices (lines 19 and 23 resp.).
> >
> > Lines later (53), it will call intel-spi.c 's intel_spi_probe
> > function, which ends up calling intel_spi_init,
> > which checks for the type
> > (https://elixir.bootlin.com/linux/latest/source/drivers/mtd/spi-nor/controllers/intel-spi.c#L313)
> > It is in this switch where the module parameter is checked, but only
> > in the BYT case; however,
> > flow coming from intel-spi-pci is BXT and CNL as mentioned before,
> > landing in their case labels (lines 343 and 351 respectively)
> > where the module parameter is not checked.
> >
> > Therefore, for BXT and CNL probed in intel-spi-pci, the chip is turned
> > writable and later the module parameter is not honored.
>
> The module parameter is definitely honored on all hardware, but the driver
> only cares about the functionality it provides through the mtd interface:
>
> https://elixir.bootlin.com/linux/latest/source/drivers/mtd/spi-nor/controllers/intel-spi.c#L941

That is a logical constraint which doesn't impact in the hardware, which already
was changed before in
https://elixir.bootlin.com/linux/latest/source/drivers/mtd/spi-nor/controllers/intel-spi.c#L924

>
> If you care about other (malicious) code writing to the driver, please explain
> what the specific attack scenario is that you are worried about, and
> why you think
> this is not sufficient. What code would be able to write to the device
> if not the
> device driver itself?

Maybe Mika can answer this better, but what I'm trying to do is to
limit the possibility of
damage, as explained in the Kconfig:
"Intel PCH/PCU SPI flash PCI driver (DANGEROUS)"
"Say N here unless you know what you are doing. Overwriting the
  SPI flash may render the system unbootable."


>
>     Arnd



-- 
Daniel Gutson
Argentina Site Director
Enginieering Director
Eclypsium

Below The Surface: Get the latest threat research and insights on
firmware and supply chain threats from the research team at Eclypsium.
https://eclypsium.com/research/#threatreport