Re: [PATCH v3 3/3] serial: core: Update uart_poll_timeout() function to return unsigned long

From: Ilpo Järvinen
Date: Fri Oct 27 2023 - 04:20:28 EST


On Fri, 27 Oct 2023, Greg Kroah-Hartman wrote:

> On Thu, Oct 26, 2023 at 07:26:28PM +0530, Vamshi Gajjela wrote:
> > From: VAMSHI GAJJELA <vamshigajjela@xxxxxxxxxx>
>
> Please use lower case letters like I think you mean to?
>
> Also, where are patches 1/3 and 2/3 of this series? I can't do anything
> without them as well.
>
> > The function uart_fifo_timeout() returns an unsigned long value, which
> > is the number of jiffies. Therefore, change the variable timeout in the
> > function uart_poll_timeout() from int to unsigned long.
> > Change the return type of the function uart_poll_timeout() from int to
> > unsigned long to be consistent with the type of timeout values.
> >
> > Signed-off-by: VAMSHI GAJJELA <vamshigajjela@xxxxxxxxxx>
> > ---
> > v3:
> > - updated description
> > v2:
> > - unsigned long instead of unsigned int
> > - added () after function name in short log
> > - updated description
> >
> > include/linux/serial_core.h | 4 ++--
> > 1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/include/linux/serial_core.h b/include/linux/serial_core.h
> > index bb6f073bc159..6916a1d7e477 100644
> > --- a/include/linux/serial_core.h
> > +++ b/include/linux/serial_core.h
> > @@ -773,9 +773,9 @@ static inline unsigned long uart_fifo_timeout(struct uart_port *port)
> > }
> >
> > /* Base timer interval for polling */
> > -static inline int uart_poll_timeout(struct uart_port *port)
> > +static inline unsigned long uart_poll_timeout(struct uart_port *port)
> > {
> > - int timeout = uart_fifo_timeout(port);
> > + unsigned long timeout = uart_fifo_timeout(port);
> >
> > return timeout > 6 ? (timeout / 2 - 2) : 1;
>
> So we are now doing 64bit math? Did that just make things slower?

That divide with a constant 2 though so I'd expect compiler to turn it
into a shift.

> What bug is this actually fixing? How have you tested this to verify it
> works?

AFAIK this doesn't fix anything because emptying when measured in jiffies
isn't that big number. It's just about making the types more consistent.

--
i.