Re: Possible case of Race in kobject_get_path()

[Greg KH <gregkh@linuxfoundation.org>]

On Wed, Mar 30, 2022 at 03:41:04PM +0530, Mukesh Ojha wrote:
> Hi All,
> 
> We are facing one issue where one driver (p1) is trying to register its
> device from driver probe
> and from another path (p2) dev_set_name(new name) done from driver probe of
> the added device whose
> new name length can be more than earlier done in (p1 path) which could
> result in redzone overwritten issue.

I do not understand, what specific driver is doing this so that we can
see an example of this problem?

> Can we get your suggestion here ? is this case of a race here ?
> 
>             p1                                                   p2
> 
>  device_register()
>     kobject_get_path()
>             =>  get_kobj_path_length
>             (length is calculated from this path)
> 
>                                                  dev_set_name()
> 
>             => fill_kobj_path
>               (path is copied here)
> 

I can not understand this example, any specific code you can point me
at?

thanks,

greg k-h