RE: [PATCH v2 1/4] cdx: Introduce lock to protect controller ops and controller list

[Gangurde, Abhijit]

> From: Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx>
> Sent: Monday, July 31, 2023 5:55 PM
> To: Gangurde, Abhijit <abhijit.gangurde@xxxxxxx>
> Cc: masahiroy@xxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; Simek, Michal
> <michal.simek@xxxxxxx>; git (AMD-Xilinx) <git@xxxxxxx>; Agarwal, Nikhil
> <nikhil.agarwal@xxxxxxx>; Gupta, Nipun <Nipun.Gupta@xxxxxxx>
> Subject: Re: [PATCH v2 1/4] cdx: Introduce lock to protect controller ops and
> controller list
> 
> On Mon, Jul 31, 2023 at 05:38:10PM +0530, Abhijit Gangurde wrote:
> > Add a mutex lock to prevent race between controller ops initiated by
> > the bus subsystem and the controller registration/unregistration.
> >
> > Signed-off-by: Abhijit Gangurde <abhijit.gangurde@xxxxxxx>
> > ---
> >  drivers/cdx/cdx.c | 14 ++++++++++++++
> >  1 file changed, 14 insertions(+)
> >
> > diff --git a/drivers/cdx/cdx.c b/drivers/cdx/cdx.c
> > index d2cad4c670a0..66797c8fe400 100644
> > --- a/drivers/cdx/cdx.c
> > +++ b/drivers/cdx/cdx.c
> > @@ -72,6 +72,8 @@
> >
> >  /* CDX controllers registered with the CDX bus */
> >  static DEFINE_XARRAY_ALLOC(cdx_controllers);
> > +/* Lock to protect controller ops and controller list */
> > +static DEFINE_MUTEX(cdx_controller_lock);
> 
> Wait, why do you have a local list and not just rely on the list the
> driver core has for you already?  Isn't this a duplicate list where you
> have objects on two different lists with a lifespan controlled only by
> one of them?

cdx_controllers list is holding just the controllers registered on the cdx bus system.
CDX devices are still maintained by driver core list. Controller list is used by rescan
which triggers rescan on all the controllers.

Thanks,
Abhijit