Re: [PATCH v2] vmpressure: implement strict mode

[Luiz Capitulino]

On Fri, 28 Jun 2013 11:45:47 -0700
Anton Vorontsov <anton@xxxxxxxxxx> wrote:

> On Fri, Jun 28, 2013 at 02:25:58PM -0400, Luiz Capitulino wrote:
> > > > > That's how it's expected to work, because on strict mode you're notified
> > > > > for the level you registered for. So apps registering for critical, will
> > > > > still be notified on critical just like before.
> > > > 
> > > > Suppose you introduce a new level, and the system hits this level. Before,
> > > > the app would receive at least some notification for the given memory load
> > > > (i.e. one of the old levels), with the new level introduced in the kernel,
> > > > the app will receive no events at all.
> > 
> > That's not true. If an app registered for critical it will still get
> > critical notification when the system is at the critical level. Just as it
> > always did. No new events will change this.
> > 
> > With today's semantics though, new events will change when current events
> > are triggered. So each new extension will cause applications to have
> > different behaviors, in different kernel versions. This looks quite
> > undesirable to me.
> 
> I'll try to explain it again.
> 
> Old behaviour:
> 
> low -> event
>   x <- but the system is at this unnamed level, between low and med
> med
> crit
> 
> 
> We add a level:
> 
> low
> low-med <- system at this state, we send an event, but the old app does
>            not know about it, so it won't receive *any* notifications. (In
> 	   older kernels it would receive low level notification
> med
> crit
> 
> You really don't see a problem here?

I do get what you're saying. We disagree it's a problem. In my mind the
best API is to get what you registered for. Nothing more, nothing less.

Now, there might be ways around it (being it a problem or not). I was
also considering this:

> 3. Never change the levels (how can we know?)

If we fail at determining levels (I honestly think current levels are
all we need), we can add a new interface later.

Also, what I said in the last email should work, which is to make
memory.pressure_level return supported levels, so an application can
register for all available levels. This way it will never miss a level.

I also think this matches having the mechanism in the kernel and
policy in user-space.
--
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/