Re: [BUG] 2.6.2 crazy mouse under heavy load

[Vojtech Pavlik]

On Thu, Feb 05, 2004 at 01:59:12PM -0700, Bob Gill wrote:
> The exception seems to be coming from
> linux-2.6.2/drivers/input/mouse/psmouse-base.c, specifically from
> 
> if (psmouse->state == PSMOUSE_ACTIVATED &&
>             psmouse->pktcnt && time_after(jiffies, psmouse->last +
> HZ/2)) {
>                 printk(KERN_WARNING "psmouse.c: %s at %s lost
> synchronization, throwing %d bytes away.\n",
>                        psmouse->name, psmouse->phys, psmouse->pktcnt);
>                 psmouse->pktcnt = 0;
>         }
> 
> 
> where (for me) HZ is 1804768000, and therefore HZ/2 is 902384000,  

This looks very wrong. HZ should be 1000 on a normal machine.
Maybe you were looking at jiffies?

> psmouse->pktcnt is 3, and (I assume) PSMOUSE_ACTIVATED is non-0 after
> boot.  I assume that pktcnt is fed by interrupt, and the problem then is
> that psmouse->last + HZ/2 blows past the jiffies value, causing the
> warning message to be issued.  When mouse service finally comes back,
> pktcnt is non-zero (and possibly whatever the maximum is that it will
> hold), and when it flushes, the mouse pointer goes nuts for a second. 
> The real problem then, is why does the sum of ps->last + HZ/2 grow to
> beyond the size of jiffies (or what is delaying the mouse service)?

Exactly. The if statement means "If no data arrived for more than half a
second, and we're in the middle of a packet, something is wrong - we
better should start from scratch."

The goal of that statement is to resynchronize the mouse data stream if
a byte is lost.

-----

There are two scenarios of this message appearing (assuming 4-byte
packets):

#1 Messages always appearing in pairs, and the sum being 4. This is a
delayed or lost interrupt. Since the driver polls the controller in
addition to servicing interrupts, if the controller has the byte,
regardless of whether an interrupt arrived, it'll be processed.

#2 Messages appearing single, with numbers from 1 to 3. This is a lost
byte. The controller either didn't receive the byte at all or something
very fishy happened.

> This is just a rough guesstimate of what is going on, but seems to fit
> the facts.   Cheers!
 

-- 
Vojtech Pavlik
SuSE Labs, SuSE CR
-
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/