[PATCH 1/3] mtd: spi-nor: intel-spi: Disable write protection only if asked

Mauro Lima mauro.lima at eclypsium.com
Wed Oct 13 11:22:40 PDT 2021


Hi!
On Wed, Oct 13, 2021 at 6:04 AM Mika Westerberg
<mika.westerberg at linux.intel.com> wrote:
>
> Hi,
>
> On Tue, Oct 12, 2021 at 03:49:22PM -0300, Mauro Lima wrote:
> > Hi Mika
> >
> > On Mon, Oct 4, 2021 at 2:18 AM Mika Westerberg
> > <mika.westerberg at linux.intel.com> wrote:
> > >
> > > Hi,
> > >
> > > On Fri, Oct 01, 2021 at 05:23:23PM -0300, Mauro Lima wrote:
> > > > Question for maintainers: With these changes is it safe to remove the
> > > > *(DANGEROUS)* tag from menuconfig?
> > >
> > > I don't think we want to remove that. This driver is not for regular
> > > users, at least in its current form.
> > Do we know why this is still dangerous for the user?
>
> There was a bug in the driver in the past (that was already fixed but it
> did not yet reach the stable trees or something like that). At this
> unfortunate time there was no DANGEROUS in the name so Ubuntu kernel
> went and enabled it. Combined with the bug certain Lenovo laptops BIOS
> turned into read-only which prevented booting from non-default devices.
>
> This happened even when the driver did not do any "write" or "erase"
> operations, just clearing the status register or so.
>
> We don't want that to happen again.
>
> > My plan is to make a sys/class driver to query write protection status
> > of the SPI, this will be
> > used by fwupd to gather information about vendors, also should be easy
> > for the user to use
> > 'cat' and see the information from userspace. For this to be possible
> > we need kernel engineers
> > to compile the kernel with this driver enabled, but the (DANGEROUS)
> > tag is a no go for most
> > of them.
> > Is there anything I can do to make this possible?
>
> IMHO we can make certain parts of the driver, that are known not to
> cause any issues available without the DANGEROUS. I mean the controller
> exposes some information that I think you are intersted in and that does
> not cause anything to be sent to the flash chip itself.
This will work for me.
Thanks!



More information about the linux-mtd mailing list