Re: [PATCH] mm: extend zero pages to same element pages for zram

[Minchan Kim <minchan@kernel.org>]

On Fri, Jan 13, 2017 at 04:02:45PM +0900, Sergey Senozhatsky wrote:
> On (01/13/17 15:47), Minchan Kim wrote:
> [..]
> > > > Could you elaborate a bit? Do you mean this?
> > > > 
> > > >         ret = scnprintf(buf, PAGE_SIZE,
> > > >                         "%8llu %8llu %8llu %8lu %8ld %8llu %8lu\n",
> > > >                         orig_size << PAGE_SHIFT,
> > > >                         (u64)atomic64_read(&zram->stats.compr_data_size),
> > > >                         mem_used << PAGE_SHIFT,
> > > >                         zram->limit_pages << PAGE_SHIFT,
> > > >                         max_used << PAGE_SHIFT,
> > > >                         // (u64)atomic64_read(&zram->stats.zero_pages),
> > > >                         (u64)atomic64_read(&zram->stats.same_pages),
> > > >                         pool_stats.pages_compacted);
> > > 
> > > yes, correct.
> > > 
> > > do we need to export it as two different stats (zero_pages and
> > > same_pages), if those are basically same thing internally?
> > 
> > So, let summary up.
> > 
> > 1. replace zero_page stat into same page stat in mm_stat
> > 2. s/zero_pages/same_pages/Documentation/blockdev/zram.txt
> > 3. No need to warn to "cat /sys/block/zram0/mm_stat" user to see zero_pages
> >    about semantic change
> 
> 1) account zero_page and same_pages in one attr.
> 
> 	this already is in the patch.
> 
> 2) do not rename zero_pages attr.
> 
> 	we can't do this so fast, I think.
> 
> 
> > 3. No need to warn to "cat /sys/block/zram0/mm_stat" user to see zero_pages
> >    about semantic change
> 
> yes. we just _may_ have more pages (depending on data pattern) which we treat
> as "zero" pages internally. this results in lower memory consumption. I don't
> think warn users about this change is necessary; they won't be able to do
> anything about it anyway. zero_pages stat is pretty much just a fun number to
> know. isn't it?
> 

I just wanted to confirm your thought.
I hope anyone doesn't is sensitive on that number. ;)
Let's replace zero_page stat field in mm_stat.

Thanks.