Re: [RFC][PATCHv2 8/8] zsmalloc: register a shrinker to trigger auto-compaction

[Minchan Kim]

On Thu, Jun 18, 2015 at 12:01:36PM +0900, Sergey Senozhatsky wrote:
> On (06/18/15 11:41), Sergey Senozhatsky wrote:
> [..]
> > > My concern is not a compacion overhead but higher memory footprint
> > > consumed by zram in reserved memory.
> > > It might hang system if zram used up reserved memory of system with
> > > ALLOC_NO_WATERMARKS. With auto-compaction, userspace has a higher chance
> > > to use more memory with uncompressible pages or file-backed pages
> > > so zram-swap can use more reserved memory. We need to evaluate it, I think.
> > > 
> 
> a couple of _not really related_ ideas that I want to voice.
> 
> (a) I'm thinking of extending zramX/compact attr. right now it's WO,
>   and I think it makes sense to make it RW:
>     ->write will trigger compaction
>     ->read will return estimated number of bytes
>   "zs_can_compact() * pages per zspage * page_size" that can be freed.
>   so user-space will have at least minimal idea whether compaction is
>   reasonable. but sure, this is racy and in general case things may
>   change between `cat compact` and `echo 1 > compact`.

It's a good idea. with that, memory manager on platform could be smart.

if memory pressure == soft and zram.can_compact > 20M
        do zram.compact
if memory pressure == hard and zram.can_compact > 5M
        do zram.compact

With this, userspace have more flexibility. :)

However, FYI, I want to make auto-compact default in future
so let's see how auto-compact is going.

> 
> 
> (b) adding a knob (yeah, like we don't have enough knobs already :-))
> that will allow 'enable/disable auto compaction'.

I agree.

> 
> 	-ss
--
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/