linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Michal Hocko <mhocko@kernel.org>
To: Pankaj Suryawanshi <pankajssuryawanshi@gmail.com>
Cc: linux-kernel@vger.kernel.org, linux-mm@kvack.org,
	Vlastimil Babka <vbabka@suse.cz>,
	pankaj.suryawanshi@einfochips.com
Subject: Re: PageBlocks and Migrate Types
Date: Tue, 27 Aug 2019 07:43:37 +0200	[thread overview]
Message-ID: <20190827054337.GK7538@dhcp22.suse.cz> (raw)
In-Reply-To: <CACDBo555_pxZjixThUZcqnADVVcmH1Qtfrr5H-2AR12L0=Rx3A@mail.gmail.com>

On Mon 26-08-19 22:35:08, Pankaj Suryawanshi wrote:
> On Mon, Aug 26, 2019 at 12:34 PM Michal Hocko <mhocko@kernel.org> wrote:
> >
> > On Thu 22-08-19 23:54:19, Pankaj Suryawanshi wrote:
> > > On Thu, Aug 22, 2019 at 6:22 PM Michal Hocko <mhocko@kernel.org> wrote:
> > > >
> > > > On Wed 21-08-19 22:23:44, Pankaj Suryawanshi wrote:
> > > > > Hello,
> > > > >
> > > > > 1. What are Pageblocks and migrate types(MIGRATE_CMA) in Linux
> memory ?
> > > >
> > > > Pageblocks are a simple grouping of physically contiguous pages with
> > > > common set of flags. I haven't checked closely recently so I might
> > > > misremember but my recollection is that only the migrate type is
> stored
> > > > there. Normally we would store that information into page flags but
> > > > there is not enough room there.
> > > >
> > > > MIGRATE_CMA represent pages allocated for the CMA allocator. There are
> > > > other migrate types denoting unmovable/movable allocations or pages
> that
> > > > are isolated from the page allocator.
> > > >
> > > > Very broadly speaking, the migrate type groups pages with similar
> > > > movability properties to reduce fragmentation that compaction cannot
> > > > do anything about because there are objects of different properti
> > > > around. Please note that pageblock might contain objects of a
> different
> > > > migrate type in some cases (e.g. low on memory).
> > > >
> > > > Have a look at gfpflags_to_migratetype and how the gfp mask is
> converted
> > > > to a migratetype for the allocation. Also follow different
> MIGRATE_$TYPE
> > > > to see how it is used in the code.
> > > >
> > > > > How many movable/unmovable pages are defined by default?
> > > >
> > > > There is nothing like that. It depends on how many objects of a
> specific
> > > > type are allocated.
> > >
> > >
> > > It means that it started creating pageblocks after allocation of
> > > different objects, but from which block it allocate initially when
> > > there is nothing like pageblocks ? (when memory subsystem up)
> >
> > Pageblocks are just a way to group physically contiguous pages. They
> > just exist along with the physically contiguous memory. The migrate type
> > for most of the memory is set to MIGRATE_MOVABLE. Portion of the memory
> > might be reserved by CMA then that memory has MIGRATE_CMA. Following
> > set_pageblock_migratetype call paths will give you a good picture.
> 
> it means if i have 4096 continuous pages = 1 pageblock
> then all the 4096 pages of same type. but if any one page is different than
> block type then ? it changed the block type or something else ?

That really depends on the specific migrate type. CMA, ISOLATE migrate
types are all or nothing IIRC. I would have to check the code to tell
exactly when MOVABLE/UNMOVABLE pageblocks transitions are done.
steal_suitable_fallback sounds like a good start to look at.
-- 
Michal Hocko
SUSE Labs

      parent reply	other threads:[~2019-08-27  5:43 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <CACDBo57u+sgordDvFpTzJ=U4mT8uVz7ZovJ3qSZQCrhdYQTw0A@mail.gmail.com>
2019-08-22 12:52 ` PageBlocks and Migrate Types Michal Hocko
2019-08-22 18:24   ` Pankaj Suryawanshi
2019-08-26  7:04     ` Michal Hocko
     [not found]       ` <CACDBo555_pxZjixThUZcqnADVVcmH1Qtfrr5H-2AR12L0=Rx3A@mail.gmail.com>
2019-08-27  5:43         ` Michal Hocko [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20190827054337.GK7538@dhcp22.suse.cz \
    --to=mhocko@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=pankaj.suryawanshi@einfochips.com \
    --cc=pankajssuryawanshi@gmail.com \
    --cc=vbabka@suse.cz \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).