Re: [2.6.16 PATCH] Connector: Filesystem Events Connector

[Evgeniy Polyakov <johnpol@2ka.mipt.ru>]

On Fri, Mar 24, 2006 at 01:42:17AM -0800, Matt Helsley (matthltc@us.ibm.com) wrote:
> On Fri, 2006-03-24 at 11:11 +0300, Evgeniy Polyakov wrote:
> > On Thu, Mar 23, 2006 at 11:35:50PM -0800, Matt Helsley (matthltc@us.ibm.com) wrote:
> > > I would argue preemption should be disabled around the if-block at the
> > > very least. Suppose your rate limit is 10k calls/sec and you have 4
> > > procs. Each proc has a sequence of three instructions:
> > > 
> > > load fsevent_sum into register rx (rx <= 1000)
> > > rx++ (rx <= 1001)
> > > store contents of register rx in fsevent_sum (fsevent_sum <= 1001)
> > > 
> > > 
> > > Now consider the following sequence of steps:
> > > 
> > > load fsevent_sum into rx (rx <= 1000)
> > > <preempted>
> > > <3 other processors each manage to increment the sum by 3333 bringing us
> > > to 9999>
> > > <resumed>
> > > rx++ (rx <= 1001)
> > > store contents of rx in fsevent_sum (fsevent_sum <= 1001)
> > > 
> > > So every processor now thinks it won't exceed the rate limit by
> > > generating more events when in fact we've just exceeded the limit. So,
> > > unless my example is flawed, I think you need to disable preemption
> > > here.
> > 
> > Doesn't it just exceed the limit by one event per cpu?
> 
> The example exceeds it by one at the time of the final store. Thanks to
> the fact that the value is then 1001 it may shortly be exceeded by much
> more than 1.

+
+       if (jiffies - last <= fsevent_ratelimit) {
+               if (fsevent_sum > fsevent_burst_limit)
+                       return -2;
+               fsevent_sum++;

Only process (and not process' syscall) can preempt us here, 
so fsevent_sum can only exceed fsevent_burst_limit by one per process
(process can not preempt itself, so when it has finished syscall which
ends up in event generation, fsevent_sum will be increased).

+       } else {
+               last = jiffies;
+               fsevent_sum = 0;
+       }

Actually, since jiffies and atomic operations are already used, I do not
think addition of new atomic_inc_return or something similar will 
even somehow change the picture.


> Cheers,
> 	-Matt Helsley

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