Re: [RFC PATCH] driver core: Don't link the device to the bus until we're ready to probe

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Fri, Mar 20, 2026 at 08:06:58PM -0700, Douglas Anderson wrote:
> The moment we link a "struct device" into the list of devices for the
> bus, it's possible probe can happen. This is because another thread
> can load the driver at any time and that can cause the device to
> probe. This has been seen in practice with a stack crawl that looks
> like this [1]:
> 
>   really_probe()
>   __driver_probe_device()
>   driver_probe_device()
>   __driver_attach()
>   bus_for_each_dev()
>   driver_attach()
>   bus_add_driver()
>   driver_register()
>   __platform_driver_register()
>   init_module() [some module]
>   do_one_initcall()
>   do_init_module()
>   load_module()
>   __arm64_sys_finit_module()
>   invoke_syscall()

Are you sure this isn't just a platform bus issue?  A bus should NOT be
allowing a driver to be added at the same time a device is being added
for that bus, ideally there should be a bus-specific lock somewhere for
this.

When a device is added to the bus, yes, a probe can happen, and is
expected to happen, for that device, so this feels odd.

that being said, your patch does seem sane, and I don't see anything
obviously wrong with it.  But it feels odd that this is just now showing
up for something that has been this way for a few decades...

thanks,

greg k-h