Re: [PATCH v2 05/15] spi: Remove code duplication in spi_add_device_locked()

[Sebastian Reichel]

Hi,

On Tue, Jul 11, 2023 at 02:06:20PM +0300, Andy Shevchenko wrote:
> On Mon, Jul 10, 2023 at 06:16:22PM +0100, Mark Brown wrote:
> > On Mon, Jul 10, 2023 at 06:49:22PM +0300, Andy Shevchenko wrote:
> > > Seems by unknown reason, probably some kind of mis-rebase,
> > > the commit 0c79378c0199 ("spi: add ancillary device support")
> > > adds a dozen of duplicating lines of code. Drop them.
> > > 
> > > Signed-off-by: Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx>
> > > ---
> > >  drivers/spi/spi.c | 11 -----------
> > >  1 file changed, 11 deletions(-)
> > > 
> > > diff --git a/drivers/spi/spi.c b/drivers/spi/spi.c
> > > index c99ee4164f11..46cbda383228 100644
> > > --- a/drivers/spi/spi.c
> > > +++ b/drivers/spi/spi.c
> > > @@ -712,17 +712,6 @@ EXPORT_SYMBOL_GPL(spi_add_device);
> > >  static int spi_add_device_locked(struct spi_device *spi)
> > >  {
> > >  	struct spi_controller *ctlr = spi->controller;
> > > -	struct device *dev = ctlr->dev.parent;
> > > -
> > > -	/* Chipselects are numbered 0..max; validate. */
> > > -	if (spi_get_chipselect(spi, 0) >= ctlr->num_chipselect) {
> > > -		dev_err(dev, "cs%d >= max %d\n", spi_get_chipselect(spi, 0),
> > > -			ctlr->num_chipselect);
> > > -		return -EINVAL;
> > > -	}
> > > -
> > > -	/* Set the bus ID string */
> > > -	spi_dev_set_name(spi);
> > 
> > I see that this is duplicating spi_add_device() (and we really could do
> > better with code sharing there I think) but I can't immediately see
> > where the duplication that's intended to be elimiated is here - where
> > else in the one call path that spi_add_device_locked() has would we do
> > the above?  Based on the changelog I was expecting to see some
> > duplicated code in the function itself.
> 
> Oh, by some reason Sebastian wasn't in this rather long Cc list.
> Added him.
> 
> Reading again I don't see any useful explanation why that piece of code has to
> be duplicated among these two functions. It's 100% a copy.
> 
> Sebastian, can you shed some light here?

The patch in this thread is obviously wrong. It results in the
checks never beeing called for spi_add_device_locked(). The copy is
in spi_add_device() and those two are not calling into each other.

But it should be fine to move the code to the start of
__spi_add_device(), which allows removing the duplication. In that
case the code will be run with the add_lock held, which is probably
what I was worried about two years ago. Looking at it again, the
lock is held anyways in case of spi_add_device_locked().

Greetings,

-- Sebastian

Attachment: signature.asc
Description: PGP signature