Re: [RFC][PATCH 3/8] mem_notify v5: introduce /dev/mem_notify new device (the core of this patch series)

[KOSAKI Motohiro]

Hi Daniel

> > +#define PROC_WAKEUP_GUARD  (10*HZ)
> [...]
> > +       timeout = info->last_proc_notify + PROC_WAKEUP_GUARD;
> 
> If only one or a few processes are using the system I think 10 seconds
> is a little long time to wait before they get the notification again.
> Can we decrease this value? Or make it configurable under /proc? Or
> make it lower with fewer users? Something like:

Oh, that is very interesting issue.
tank you good point out.

after deep thinking, I understand my current implementation is fully stupid.
current, worst case is below.

  1. low end
     - many process of used only a bit memory(sh, cp etc..) exist.
     - 1 memory eater process exist(may be, it is fat browser)
       and it watching /dev/mem_notify.

  2. high end
     - many process of used only a bit memory(sh, cp etc..) exist.
     - 1 memory eater process exist(may be, it is DB)
       and it watching /dev/mem_notify.

the point is "only 1 process watch /dev/mem_notify", but not a few processor.
I fix it with pleasure. 


> timeout = info->last_proc_notify + min(mem_notify_users, PROC_WAKEUP_GUARD);

I like this formula.
the rest problem is decide to default value when only 1 process watch /dev/mem_notify.

What do you think it?
and if my low end worst case situation doesn't match yours, 
Could you please explain your usage more?


BTW: 
end up, We will add /proc configuration the future.
but I think it is too early.
sometimes configrable parameter prevent the discussion of nicer default value.
Instead, I hope the default value changed by adjust your usage.


- kosaki


--
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/