Re: Spurious timeouts in mvmdio

[Sebastian Hesselbarth]

On 12/03/2013 09:57 PM, Leigh Brown wrote:
> Hi Russell and Nicolas,
>
> Apologies for taking so long to respond to this thread.
>
> On 2013-12-03 12:40, Russell King - ARM Linux wrote:
> > On Tue, Dec 03, 2013 at 07:23:46AM -0500, Jason Cooper wrote:
> > > On Mon, Dec 02, 2013 at 04:15:54PM +0100, Nicolas Schichan wrote:
> [...]
> > > 11:31 < shesselba> kos_tom: it is already using usecs_to_jiffies()
> > > 11:31 < shesselba> the thing is: 1ms is less than a jiffy
> >
> > Yes, and the kernels time conversion functions aren't stupid.  Let's
> > look at this function's implementation:
> >
> > unsigned long usecs_to_jiffies(const unsigned int u)
> > {
> >         if (u > jiffies_to_usecs(MAX_JIFFY_OFFSET))
> >                 return MAX_JIFFY_OFFSET;
> > #if HZ <= USEC_PER_SEC && !(USEC_PER_SEC % HZ)
> >         return (u + (USEC_PER_SEC / HZ) - 1) / (USEC_PER_SEC / HZ);
> > #elif HZ > USEC_PER_SEC && !(HZ % USEC_PER_SEC)
> >         return u * (HZ / USEC_PER_SEC);
> > #else
> >         return (USEC_TO_HZ_MUL32 * u + USEC_TO_HZ_ADJ32)
> >                 >> USEC_TO_HZ_SHR32;
> > #endif
> > }
> >
> > Now, assuming HZ=100 and USEC_PER_SEC=1000000, we will use:
> >
> >     return (u + (USEC_PER_SEC / HZ) - 1) / (USEC_PER_SEC / HZ);
> >
> > If you ask for 1us, this comes out as:
> >
> >     return (1 + (1000000 / 100) - 1) / (1000000 / 100);
> >
> > which is one jiffy.  So, for a requested 1us period, you're given a
> > 1 jiffy interval, or 10ms.  For other (sensible) values:
> >
> >         return (USEC_TO_HZ_MUL32 * u + USEC_TO_HZ_ADJ32)
> >                 >> USEC_TO_HZ_SHR32;
> >
> > gets used, which has a similar behaviour.
> >
> > Now, depending on how you use this one jiffy interval, the thing to
> > realise
> > is that with this kind of loop:
> >
> >     timeout = jiffies + usecs_to_jiffies(1);
> >     do {
> >         something;
> >     } while (time_is_before_jiffies(timeout));
> >
> > what this equates to is:
> >
> >     } while (jiffies - timeout < 0);
>
> You've got that last line the wrong way round, but I understand what you
> are
> getting at.
>
> > What this means is that the loop breaks at jiffies = timeout, so it can
> > indeed timeout before one tick - within 0 to 10ms for HZ=100.  The
> > problem
> > is not the usecs_to_jiffies(), it's with the implementation.
>
> > If you use time_is_before_eq_jiffies() instead, it will also loop if
> > jiffies == timeout, which will give you the additional safety margin -
> > meaning it will timeout after 10 to 20ms instead.
>
> The code in question uses the logic in reverse so it *exits* if
> time_is_before_jiffies(end) is true.  In other words it exits if
> "jiffies" is
> greater than "end".  Since, as you say, usecs_to_jiffies(somevalue) will
> be a
> minimum of 1, that means it will always loop at least twice.  So, I
> think it's
> doing what you suggest, but in a different way, when polling.
>
> > You may wish to consider coding this differently as well - if you have
> > the error interrupt, there's no need for this loop.  You only need the
> > loop if you're using usleep_range().  Note the return value of
> > wait_event_timeout() will tell you positively and correctly if the waited
> > condition succeeded or you timed out.
>
> Although the the wait_event_timeout is in the loop, it always exits due to
> setting timedout to 1.  This was done to make the code smaller but I was
> probably trying to be cleverer than I should have been.
>
> So, given the above I believe the polling code is correct.  My mistake
> was to
> assume that wait_event_timeout() guarantees a minimum of 1 jiffie wait.
>
> Nicolas' patch should fix the issue, but I prefer the following as it is
> more
> correct, as it only adjusts the timeout when calling
> wait_event_timeout().  As
> I said above,I believe the polling code is correct.
>
> diff --git a/drivers/net/ethernet/marvell/mvmdio.c
> b/drivers/net/ethernet/marvell/mvmdio.c
> index 7354960..b187c08 100644
> --- a/drivers/net/ethernet/marvell/mvmdio.c
> +++ b/drivers/net/ethernet/marvell/mvmdio.c
> @@ -92,6 +92,14 @@ static int orion_mdio_wait_ready(struct mii_bus *bus)
>               if (time_is_before_jiffies(end))
>                   ++timedout;
>               } else {
> +            /*
> +             * wait_event_timeout does not guarantee a delay of at
> +             * least one whole jiffie, so timeout must be no less
> +             * than two.
> +             */
> +            if (timeout < 2)
> +                timeout = 2;

If you always want to wait at least two jiffies, why not just increase
TIMEOUT makro to 20ms instead of messing here with it again?
As said on IRC log above, originally timeout was 100ms.

Sebastian

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/