Re: [PATCH v2 2/2] rust: miscdevice: access the `struct miscdevice` from fops->open()

[Danilo Krummrich <dakr@kernel.org>]

On Mon, Dec 09, 2024 at 02:36:31PM +0100, Alice Ryhl wrote:
> On Mon, Dec 9, 2024 at 2:13 PM Greg Kroah-Hartman
> <gregkh@linuxfoundation.org> wrote:
> >
> > On Mon, Dec 09, 2024 at 01:53:42PM +0100, Alice Ryhl wrote:
> > > On Mon, Dec 9, 2024 at 1:08 PM Greg Kroah-Hartman
> > > <gregkh@linuxfoundation.org> wrote:
> > > >
> > > > On Mon, Dec 09, 2024 at 01:00:05PM +0100, Alice Ryhl wrote:
> > > > > On Mon, Dec 9, 2024 at 12:53 PM Greg Kroah-Hartman
> > > > > <gregkh@linuxfoundation.org> wrote:
> > > > > >
> > > > > > On Mon, Dec 09, 2024 at 12:38:32PM +0100, Alice Ryhl wrote:
> > > > > > > On Mon, Dec 9, 2024 at 12:10 PM Greg Kroah-Hartman
> > > > > > > <gregkh@linuxfoundation.org> wrote:
> > > > > > > >
> > > > > > > > On Mon, Dec 09, 2024 at 11:50:57AM +0100, Alice Ryhl wrote:
> > > > > > > > > On Mon, Dec 9, 2024 at 9:48 AM Greg Kroah-Hartman
> > > > > > > > > <gregkh@linuxfoundation.org> wrote:
> > > > > > > > > >
> > > > > > > > > > On Mon, Dec 09, 2024 at 07:27:47AM +0000, Alice Ryhl wrote:
> > > > > > > > > > > Providing access to the underlying `struct miscdevice` is useful for
> > > > > > > > > > > various reasons. For example, this allows you access the miscdevice's
> > > > > > > > > > > internal `struct device` for use with the `dev_*` printing macros.
> > > > > > > > > > >
> > > > > > > > > > > Note that since the underlying `struct miscdevice` could get freed at
> > > > > > > > > > > any point after the fops->open() call, only the open call is given
> > > > > > > > > > > access to it. To print from other calls, they should take a refcount on
> > > > > > > > > > > the device to keep it alive.
> > > > > > > > > >
> > > > > > > > > > The lifespan of the miscdevice is at least from open until close, so
> > > > > > > > > > it's safe for at least then (i.e. read/write/ioctl/etc.)
> > > > > > > > >
> > > > > > > > > How is that enforced? What happens if I call misc_deregister while
> > > > > > > > > there are open fds?
> > > > > > > >
> > > > > > > > You shouldn't be able to do that as the code that would be calling
> > > > > > > > misc_deregister() (i.e. in a module unload path) would not work because
> > > > > > > > the module reference count is incremented at this point in time due to
> > > > > > > > the file operation module reference.
> > > > > > >
> > > > > > > Oh .. so misc_deregister must only be called when the module is being unloaded?
> > > > > >
> > > > > > Traditionally yes, that's when it is called.  Do you see it happening in
> > > > > > any other place in the kernel today?
> > > > >
> > > > > I had not looked, but I know that Binder allows dynamically creating
> > > > > and removing its devices at runtime. It happens to be the case that
> > > > > this is only supported when binderfs is used, which is when it doesn't
> > > > > use miscdevice, so technically Binder does not call misc_deregister()
> > > > > outside of module unload, but following its example it's not hard to
> > > > > imagine that such removals could happen.
> > > >
> > > > That's why those are files and not misc devices :)
> > >
> > > I grepped for misc_deregister and the first driver I looked at is
> > > drivers/misc/bcm-vk which seems to allow dynamic deregistration if the
> > > pci device is removed.
> >
> > Ah, yeah, that's going to get messy and will be a problem if someone has
> > the file open then.
> >
> > > Another tricky path is error cleanup in its probe function.
> > > Technically, if probe fails after registering the misc device, there's
> > > a brief moment where you could open the miscdevice before it gets
> > > removed in the cleanup path, which seems to me that it could lead to
> > > UAF?
> > >
> > > Or is there something I'm missing?
> >
> > Nope, that too is a window of a problem, luckily you "should" only
> > register the misc device after you know the device is safe to use as
> > once it is registered, it could be used so it "should" be the last thing
> > you do in probe.
> >
> > So yes, you are right, and we do know about these issues (again see the
> > talk I mentioned and some previous ones for many years at plumbers
> > conferences by different people.)  It's just up to someone to do the
> > work to fix them.
> >
> > If you think we can prevent the race in the rust side, wonderful, I'm
> > all for that being a valid fix.
> 
> The current patch prevents the race by only allowing access to the
> `struct miscdevice` in fops->open(). That's safe since
> `file->f_op->open` runs with `misc_mtx` held. Do we really need the
> miscdevice to stay alive for longer? You can already take a refcount
> on `this_device` if you want to keep the device alive for longer for
> dev_* printing purposes, but it seems like that is the only field you
> really need from the `struct miscdevice` past fops->open()?

Good point, I also can't really see anything within struct miscdevice that a
driver could need other than `this_device`.

How would you provide the `device::Device` within the `MiscDevice` trait
functions?

If we don't guarantee that the `struct miscdevice` is still alive past open() we
need to take a reference on `this_device` in open().

I guess the idea would be to let `MiscDeviceRegistration` provide a function to
obtain an `ARef<device::Device>`?

> 
> Alice
>