Re: [PATCH] spi: spi-fsl-lpspi: fix watermark truncation caused by type cast

[Andy Shevchenko <andriy.shevchenko@intel.com>]

On Wed, Nov 26, 2025 at 02:10:25AM +0000, Carlos Song wrote:
> > -----Original Message-----
> > From: Andy Shevchenko <andriy.shevchenko@intel.com>
> > Sent: Tuesday, November 25, 2025 11:11 PM
> > On Sat, Nov 22, 2025 at 10:57:16AM +0000, david laight wrote:
> > > On Fri, 21 Nov 2025 11:19:34 -0500
> > > Frank Li <Frank.li@nxp.com> wrote:
> > > > On Mon, Nov 17, 2025 at 11:03:55AM +0800, carlos.song@nxp.com wrote:

...

> > > > > + /*
> > > > > +  * t->len is 'unsigned' and txfifosize and watermrk is 'u8',
> > > > > + force
> > > > > +  * type cast is inevitable. When len > 255, len will be
> > > > > + truncated in min_t(),
> > > > > +  * it caused wrong watermark set. 'unsigned int' is as the
> > > > > + designated type
> > > > > +  * for min_t() to avoid truncation.
> > > > > +  */
> > > > > + fsl_lpspi->watermark = min_t(unsigned int,
> > > > >                                fsl_lpspi->txfifosize,
> > > > >                                t->len);
> > > >
> > > > There are thread discussion about min() and min_t()
> > > >
> > > > https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Flo
> > > > re.kernel.org%2Fall%2F20251119224140.8616-1-david.laight.linux%40gma
> > > >
> > il.com%2F&data=05%7C02%7Ccarlos.song%40nxp.com%7C24c955c5ab414a26
> > 730
> > > >
> > a08de2c34dda3%7C686ea1d3bc2b4c6fa92cd99c5c301635%7C0%7C0%7C6389
> > 96802
> > > >
> > 735067934%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiO
> > iIwLj
> > > >
> > AuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%
> > 7C
> > > > %7C&sdata=NEBirtRBR5RXz9cUXuqWVtHT6b3memVhsRe5mA7AIAQ%3D&
> > reserved=0
> > >
> > > The big comment even carefully explains that the two types are unsigned
> > ones.
> > > So a simple min() is absolutely fine (and the comment can go away).
> > >
> > > The old typecheck in min was just so stupid.
> > > In this case the 'u8' variable is promoted to 'int' (they always are)
> > > and then converted to 'unsigned int' to match the other type.
> > > Even though there is an implicit 'int' => 'unsigned int' cast it is
> > > impossible for a negative value to become a large positive on (which
> > > is the only justification for the type check).
> > >
> > > I'd check the file for other uses on min_t() as well.
> > 
> > Just came to this thread to echoing what David said. +1 to the above, please
> > convert to simple min(). The use cases for min_t() and max_t() should be rare
> > really.
> 
> Thank you for all ack about this patch.
> From my points, min() ot min_t(unsigned, x, x) both are ok.
> This patch has been picked, do I need to do a new patch to use min() instead this patch?

It's already applied, so send a followup that moves from min_t() to min().

-- 
With Best Regards,
Andy Shevchenko