Re: [PATCH v4 1/2] tty: serial: dz: convert atomic_* to refcount_* APIs for map_guard

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

On Tue, Jan 10, 2023 at 01:17:54PM +0530, Deepak R Varma wrote:
> On Tue, Jan 10, 2023 at 07:27:44AM +0000, Reshetova, Elena wrote:
> >  
> > > On Tue, Jan 03, 2023 at 09:59:52AM +0100, Jiri Slaby wrote:
> > > > On 26. 12. 22, 7:21, Deepak R Varma wrote:
> > > > > The refcount_* APIs are designed to address known issues with the
> > > > > atomic_t APIs for reference counting. They provide following distinct
> > > > > advantages
> > > > >     - protect the reference counters from overflow/underflow
> > > > >     - avoid use-after-free errors
> > > > >     - provide improved memory ordering guarantee schemes
> > > > >     - neater and safer.
> > > >
> > > > Really? (see below)
> > > >
> > > > > --- a/drivers/tty/serial/dz.c
> > > > > +++ b/drivers/tty/serial/dz.c
> > > > ...
> > > > > @@ -687,23 +686,19 @@ static int dz_map_port(struct uart_port *uport)
> > > > >   static int dz_request_port(struct uart_port *uport)
> > > > >   {
> > > > >   	struct dz_mux *mux = to_dport(uport)->mux;
> > > > > -	int map_guard;
> > > > >   	int ret;
> > > > >
> > > > > -	map_guard = atomic_add_return(1, &mux->map_guard);
> > > > > -	if (map_guard == 1) {
> > > > > -		if (!request_mem_region(uport->mapbase, dec_kn_slot_size,
> > > > > -					"dz")) {
> > > > > -			atomic_add(-1, &mux->map_guard);
> > > > > -			printk(KERN_ERR
> > > > > -			       "dz: Unable to reserve MMIO resource\n");
> > > > > +	refcount_inc(&mux->map_guard);
> > > > > +	if (refcount_read(&mux->map_guard) == 1) {
> > > >
> > > > This is now racy, right?
> > > 
> > > Hello Jiri,
> > > I found this [1] commit which introduced similar transformation in a
> > > neighbouring driver. Can you please comment how is this different from the
> > > current patch proposal?
> > > 
> > > [1] commit ID: 22a33651a56f ("convert sbd_duart.map_guard from atomic_t to
> > > refcount_t")
> > > 
> > > On a side note, I have not been able to find an exact 1:1 map to the
> > > atomic_add_result API. I am wondering should we have one?
> > 
> 
> Hello Elena,
> 
> > In past we have decided not to provide this API for refcount_t
> > because for truly correctly behaving reference counters it should not be needed
> > (vs atomics that cover a broader range of use cases). 
> 
> So, there is no FAA refcount wrapper? I think this is a pretty common need.
> Please correct me if I am wrong.
> 
> > Can you use !refcount_inc_not_zero in the above case?
> 
> I actually did try that but was not sure if truly addresses the objection.
> Please attached and let me know if you have a feedback on the alternate
> approach.
> 
> Thank you,
> ./drv
> 
> 
> > 
> > Best Regards,
> > Elena.

> ############## ORIGINAL CODE ##################################
> -       map_guard = atomic_add_return(1, &mux->map_guard);
> -       if (map_guard == 1) {
> -               if (!request_mem_region(uport->mapbase, dec_kn_slot_size,
> -                                       "dz")) {
> -                       atomic_add(-1, &mux->map_guard);
> -                       printk(KERN_ERR
> -                              "dz: Unable to reserve MMIO resource\n");
>                         return -EBUSY;
>                 }
>         }
> 
> ############## INITIAL APPROACH ##################################
> +       refcount_inc(&mux->map_guard);
> +       if (refcount_read(&mux->map_guard) == 1) {
> +               if (!request_mem_region(uport->mapbase, dec_kn_slot_size, "dz")) {
> +                       refcount_dec(&mux->map_guard);
> +                       printk(KERN_ERR "dz: Unable to reserve MMIO resource\n");
>                         return -EBUSY;
>                 }
>         }
> 
> ############## ALTERNATE APPROACH ##################################
> 
> +       if (!refcount_inc_not_zero(&mux->map_guard)) {
> +               refcount_inc(&mux->map_guard);
> +               if (!request_mem_region(uport->mapbase, dec_kn_slot_size, "dz")) {
> +                       refcount_dec(&mux->map_guard);
> +                       printk(KERN_ERR "dz: Unable to reserve MMIO resource\n");
>                         return -EBUSY;
>                 }
>         }
> 

This feels odd to me, why not just use a normal lock instead?

thanks,

greg k-h