Re: [RFC][PATCH 2/9] deadlock prevention core

[Neil Brown <neilb@suse.de>]

On Thursday August 17, phillips@google.com wrote:
> Andrew Morton wrote:
> > Daniel Phillips <phillips@google.com> wrote:
> >>What happened to the case where we just fill memory full of dirty file
> >>pages backed by a remote disk?
> > 
> > Processes which are dirtying those pages throttle at
> > /proc/sys/vm/dirty_ratio% of memory dirty.  So it is not possible to "fill"
> > memory with dirty pages.  If the amount of physical memory which is dirty
> > exceeds 40%: bug.
> 
> Hi Andrew,
> 
> So we make 400 MB of a 1 GB system unavailable for write caching just to
> get around the network receive starvation issue?

No.  We make it unavailable for write caching so it is available for
other important things like cached clean pages and executables etc.
You have to start throttling some where or you get very bad
behaviour.  40% seems a good number, but it is tunable.
The fact that it helps with receive starvation is just s bonus.

> 
> What happens if some in kernel user grabs 68% of kernel memory to do some
> very important thing, does this starvation avoidance scheme still work?

That is a very good question.  Is memlocked memory throttled against
dirty pages, and does it decrease the space available to the 40%
calculation?  I don't know.  I guess I could look at the code...

get_dirty_limits in page_writeback caps 'dirty_ratio' (40 by default,
hence the 40%) at unmapped_ratio/2.
So yes.  If 68% is mapped and locked (I assume that it the situation
you are referring to) that leaves 32% unlocked, so the 40% above is
reduced to 16% and you should still have 160Meg of breathing space.

NeilBrown
-
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/