Re: [PATCH v1] driver core: Take parent lock in async device attach

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

On Tue, Oct 06, 2026 at 01:23:35PM +0000, Mario Peter wrote:
> On buses with need_parent_lock set (only USB), probe() must run with the
> parent device locked. A synchronous attach after device_add() gets this
> from the caller, e.g. usb_set_configuration() holds the udev lock while
> adding interfaces. __device_attach_async_helper() runs the probe from an
> async worker instead, where that lock isn't held, and only takes
> device_lock(dev).
> 
> With async probing enabled for USB drivers (driver_async_probe=*,
> module.async_probe=1), hub_probe() of a multi-TT hub then races with
> usb_set_configuration() and both create the interface's endpoint
> devices. On an i.MX8MM board this hit 11 of 100 boots:
> 
>   sysfs: cannot create duplicate filename '.../1-1/1-1:1.0/ep_81'
>   Call trace:
>    sysfs_warn_dup
>    usb_create_ep_devs
>    create_intf_ep_devs
>    usb_set_interface
>    hub_probe
>    usb_probe_interface
>    really_probe
>    __device_attach_async_helper
>    async_run_entry_fn
> 
> Take the parent lock there as well, like __driver_attach_async_helper()
> does, and move __device_driver_lock/unlock() up for that. With this the
> warning was gone in 300 boots.
> 
> Fixes: 765230b5f084 ("driver-core: add asynchronous probing support for drivers")
> Assisted-by: LLM
> Signed-off-by: Mario Peter <mario.peter@leica-geosystems.com>
> ---
> Seen and tested on 6.16.y, which has the same code here. On mainline
> only build-tested.

Please verify this on the latest tree, 6.16.y is _VERY_ old and
obsolete, lots has changed in the year it was released.

> Not covered: deferred_probe_work_func() also re-probes via
> __device_attach() without the parent lock.
> 
> Analysis and patch done with the help of Claude Code.

This feels really wrong, what bus is the host controller on for these?

thanks,

greg k-h