Re: [PATCHv2 2/5] vmscan: Kill hibernation specific reclaim logic and unify it

[Rafael J. Wysocki]

On Tuesday 03 November 2009, KOSAKI Motohiro wrote:
> > On Monday 02 November 2009, KOSAKI Motohiro wrote:
> > > > > Then, This patch changed shrink_all_memory() to only the wrapper function of 
> > > > > do_try_to_free_pages(). it bring good reviewability and debuggability, and solve 
> > > > > above problems.
> > > > > 
> > > > > side note: Reclaim logic unificication makes two good side effect.
> > > > >  - Fix recursive reclaim bug on shrink_all_memory().
> > > > >    it did forgot to use PF_MEMALLOC. it mean the system be able to stuck into deadlock.
> > > > >  - Now, shrink_all_memory() got lockdep awareness. it bring good debuggability.
> > > > 
> > > > As I said previously, I don't really see a reason to keep shrink_all_memory().
> > > > 
> > > > Do you think that removing it will result in performance degradation?
> > > 
> > > Hmm...
> > > Probably, I misunderstood your mention. I thought you suggested to kill
> > > all hibernation specific reclaim code. I did. It's no performance degression.
> > > (At least, I didn't observe)
> > > 
> > > But, if you hope to kill shrink_all_memory() function itsef, the short answer is,
> > > it's impossible.
> > > 
> > > Current VM reclaim code need some preparetion to caller, and there are existing in
> > > both alloc_pages_slowpath() and try_to_free_pages(). We can't omit its preparation.
> > 
> > Well, my grepping for 'shrink_all_memory' throughout the entire kernel source
> > code seems to indicate that hibernate_preallocate_memory() is the only current
> > user of it.  I may be wrong, but I doubt it, unless some new users have been
> > added since 2.6.31.
> > 
> > In case I'm not wrong, it should be safe to drop it from
> > hibernate_preallocate_memory(), because it's there for performance reasons
> > only.  Now, since hibernate_preallocate_memory() appears to be the only user of
> > it, it should be safe to drop it entirely.
> 
> Hmmm...
> I've try the dropping shrink_all_memory() today. but I've got bad result.
> 
> In 3 times test, result were
> 
>  2 times: kernel hang-up ;)
>  1 time:   success, but make slower than with shrink_all_memory() about 100x times.
> 
> 
> Did you try to drop it yourself on your machine? Is this success?

Generally, yes, but the performance was hit really badly.

So, the conclusion is that we need shrink_all_memory() for things to work,
which is kind of interesting.

In that case, please feel free to add Acked-by: Rafael J. Wysocki <rjw@xxxxxxx>
to the patch.

Thanks,
Rafael
--
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/