Re: [PATCH v3 0/2] driver core: add TAINT_FORCED_BIND for when userspace manually messes with devices and drivers

Am 05.09.26 um 13:27 schrieb Greg Kroah-Hartman:

On Fri, Sep 04, 2026 at 09:46:59PM +0200, Armin Wolf wrote:

quoted

Am 04.09.26 um 07:10 schrieb Greg Kroah-Hartman:

quoted

The ability to add and remove devices from a driver through the sysfs
"bind" and "unbind" files was created all those decades ago as a way
that kernel developers can iterate faster, and provide a debugging way
for users to attempt to add a new device to a driver without having to
rebuild their kernel.

This api over the years has been abused and recently come under a major
fuzzing "attack" through tools like syzbot which decided that it would
attempt to just randomly bind any driver to any type of device, causing
loads of unneeded errors and pointless kernel patches to be generated by
unsuspecting new developers.

Handle all of this by adding a new taint flag, TAINT_FORCED_BIND, which
will be set on the driver if the bind/unbind sysfs files are ever
written to.  This lets kernel developers "know" that a user is
attempting to do something that is not normal, and as such, if the
kernel breaks they get to keep the shiny pieces laying around on the
floor.

The flag is 'Y' which was unused, and can remembered as the user is
"yeeting" the device being operated on here (thrown with force without
regard for the thing being thrown).

Note, the taint flag gets set _BEFORE_ the bind/unbind callback happens,
as many times crashes/oops/warnings/failures happen within the callback,
and the taint flag needs to be there to show what was being attempted.
If it were to be set after the callback happens, the oops report would
not properly reflect what foolishness was being attempted.

Fuzzing tools like syzbot, that doesn't have hand-crafted rules to keep
the tool from hitting bind/unbind, should be run with panic_on_taint
enabled so that they fall over and don't continue on, thinking that they
actually found a real issue.

Userspace operations that rely on the bind/unbind files to work around
the lack of will to upgrade a kernel image to a newer version with
proper support for new devices, or the lack of will to submit valid
device ids to driver authors, will still work properly, but now the
kernel will be flagged in a way that will show that perhaps those users
should reconsider their behavior and work to have the drivers properly
support these devices in a "native" manner.

What about users that use driver_override together with vfio-pci? AFAIK
they still need to write to bind/unbind.

As I state above, that is an abuse of this api.

quoted

Also i always liked the fact that the bind/unbind attributes make it very
easy in creating generic drivers. In fact, i am intending to use this feature
in order to add netlink support to the ACPI-WMI subsystem in the future.
I am planning to realize this thru a generic WMI driver that has to be manually
bound to a given WMI device to make it available over netlink, so having
users be able to write bind/unbind without potentially triggering a kernel panic
would be very nice.

It's a taint flag being set, not a panic, showing developers that "hey,
the user just did something odd here, ask them about it when handling
this bug report."

quoted

Additionally, writing into the bind attribute still checks if the driver and the
device match:

/*
  * Manually attach a device to a driver.
  * Note: the driver must want to bind to the device,
  * it is not possible to override the driver's id table.
  */
static ssize_t bind_store(struct device_driver *drv, const char *buf, size_t count)

Because of this i would suggest that the taint (if at all) be added to the
driver_override attribute instead, because only with this attribute users can bind
random drivers to truly random devices.

I had thought of that, but the force removal of drivers from devices
wouldn't show up this way, which is one of the things that we do want to
warn developers about.

IMHO drivers that do not support this should rather just set suppress_bind_attrs.
And drivers for busses that do support hotplugging must support surprise removal
anyway.

Would it be possible to only taint the kernel if the underlying bus does not
claim to be hotplugable?

And, if we just did the driver_override taint, it would allow those
drivers who claim to be "generic" like you are referring to, to keep
"getting away with that" when it's obviously not true that they can
handle any device for that bus, otherwise it would have automatically
bound to it already (and no manual intervention would be needed.)

By keeping the manual configuration of bind, you are forcing userspace
to manually do work, while still a possible thing to do, now at least a
developer who gets a bug report will _know_ it happened, which is why
the taint flag is there.

Does that make a bit more sense?

How are users supposed to select between multiple drivers for the same device
if all those drivers claim to support said device? Assuming i have a generic driver
and a more specialized driver that both claim support for the same device. How
can i control which one is ultimately bound to the device without using bind/unbind?
How to utilize driver_override when another driver is already bound to the device
without using the unbind attribute?

Also drivers that claim to be generic while obviously not being able to handle all
devices are themselves broken, not the surrounding mechanism that relies on the data
being supplied by the broken drivers.

Maybe we should instead fix those drivers to stop claiming to be generic? Do you know
the names of some drivers that are affected by this?

Thanks,
Armin Wolf

thanks,

greg k-h