All of lore.kernel.org
 help / color / mirror / Atom feed
From: Mel Gorman <mel@csn.ul.ie>
To: Christoph Lameter <cl@linux-foundation.org>
Cc: Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>,
	linux-mm@kvack.org, Nick Piggin <npiggin@suse.de>,
	Chris Mason <chris.mason@oracle.com>,
	Jens Axboe <jens.axboe@oracle.com>,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH 1/3] page-allocator: Under memory pressure, wait on pressure to relieve instead of congestion
Date: Tue, 9 Mar 2010 17:01:23 +0000	[thread overview]
Message-ID: <20100309170123.GG4883@csn.ul.ie> (raw)
In-Reply-To: <alpine.DEB.2.00.1003091005310.28897@router.home>

On Tue, Mar 09, 2010 at 10:09:11AM -0600, Christoph Lameter wrote:
> On Tue, 9 Mar 2010, Christian Ehrhardt wrote:
> 
> > > What happens if memory becomes available in another zone? Lets say we are
> > > waiting on HIGHMEM and memory in ZONE_NORMAL becomes available?
> >
> > Do you mean the same as Nick asked or another aspect of it?
> > citation:
> > "I mean the other way around. If that zone's watermarks are not met, then why
> > shouldn't it be woken up by other zones reaching their watermarks."
> 
> Just saw that exchange. Yes it is similar. Mel only thought about NUMA
> but the situation can also occur in !NUMA because multiple zones do not
> require NUMA.
> 

True, although rare. Elsewhere I suggested that the wait could be on a
per-node basis instead of per-zone. My main concern there would be
adding a new hot cache line in the page free path or an unfortunate mix
of zone and node logic. I'm not fully convinced it's worth it but will
check it out.

> If a process goes to sleep on an allocation that has a preferred zone of
> HIGHMEM then other processors may free up memory in ZONE_DMA and
> ZONE_NORMAL and therefore memory may become available but the process will
> continue to sleep.
> 

Until it's timeout at least. It's still better than the current
situation of sleeping on congestion.

The ideal would be waiting on a per-node basis. I'm just not liking having
to look up the node structure when freeing a patch of pages and making a
cache line in there unnecessarily hot.

> The wait structure needs to be placed in the pgdat structure to make it
> node specific.
> 
> But then an overallocated node may stall processes. If that node is full
> of unreclaimable memory then the process may never wake up?
> 

Processes wake after a timeout.

-- 
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab

WARNING: multiple messages have this Message-ID (diff)
From: Mel Gorman <mel@csn.ul.ie>
To: Christoph Lameter <cl@linux-foundation.org>
Cc: Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>,
	linux-mm@kvack.org, Nick Piggin <npiggin@suse.de>,
	Chris Mason <chris.mason@oracle.com>,
	Jens Axboe <jens.axboe@oracle.com>,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH 1/3] page-allocator: Under memory pressure, wait on pressure to relieve instead of congestion
Date: Tue, 9 Mar 2010 17:01:23 +0000	[thread overview]
Message-ID: <20100309170123.GG4883@csn.ul.ie> (raw)
In-Reply-To: <alpine.DEB.2.00.1003091005310.28897@router.home>

On Tue, Mar 09, 2010 at 10:09:11AM -0600, Christoph Lameter wrote:
> On Tue, 9 Mar 2010, Christian Ehrhardt wrote:
> 
> > > What happens if memory becomes available in another zone? Lets say we are
> > > waiting on HIGHMEM and memory in ZONE_NORMAL becomes available?
> >
> > Do you mean the same as Nick asked or another aspect of it?
> > citation:
> > "I mean the other way around. If that zone's watermarks are not met, then why
> > shouldn't it be woken up by other zones reaching their watermarks."
> 
> Just saw that exchange. Yes it is similar. Mel only thought about NUMA
> but the situation can also occur in !NUMA because multiple zones do not
> require NUMA.
> 

True, although rare. Elsewhere I suggested that the wait could be on a
per-node basis instead of per-zone. My main concern there would be
adding a new hot cache line in the page free path or an unfortunate mix
of zone and node logic. I'm not fully convinced it's worth it but will
check it out.

> If a process goes to sleep on an allocation that has a preferred zone of
> HIGHMEM then other processors may free up memory in ZONE_DMA and
> ZONE_NORMAL and therefore memory may become available but the process will
> continue to sleep.
> 

Until it's timeout at least. It's still better than the current
situation of sleeping on congestion.

The ideal would be waiting on a per-node basis. I'm just not liking having
to look up the node structure when freeing a patch of pages and making a
cache line in there unnecessarily hot.

> The wait structure needs to be placed in the pgdat structure to make it
> node specific.
> 
> But then an overallocated node may stall processes. If that node is full
> of unreclaimable memory then the process may never wake up?
> 

Processes wake after a timeout.

-- 
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

  reply	other threads:[~2010-03-09 17:01 UTC|newest]

Thread overview: 136+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-03-08 11:48 [RFC PATCH 0/3] Avoid the use of congestion_wait under zone pressure Mel Gorman
2010-03-08 11:48 ` Mel Gorman
2010-03-08 11:48 ` [PATCH 1/3] page-allocator: Under memory pressure, wait on pressure to relieve instead of congestion Mel Gorman
2010-03-08 11:48   ` Mel Gorman
2010-03-09 13:35   ` Nick Piggin
2010-03-09 13:35     ` Nick Piggin
2010-03-09 14:17     ` Mel Gorman
2010-03-09 14:17       ` Mel Gorman
2010-03-09 15:03       ` Nick Piggin
2010-03-09 15:03         ` Nick Piggin
2010-03-09 15:42         ` Christian Ehrhardt
2010-03-09 15:42           ` Christian Ehrhardt
2010-03-09 18:22           ` Mel Gorman
2010-03-09 18:22             ` Mel Gorman
2010-03-10  2:38             ` Nick Piggin
2010-03-10  2:38               ` Nick Piggin
2010-03-09 17:35         ` Mel Gorman
2010-03-09 17:35           ` Mel Gorman
2010-03-10  2:35           ` Nick Piggin
2010-03-10  2:35             ` Nick Piggin
2010-03-09 15:50   ` Christoph Lameter
2010-03-09 15:50     ` Christoph Lameter
2010-03-09 15:56     ` Christian Ehrhardt
2010-03-09 15:56       ` Christian Ehrhardt
2010-03-09 16:09       ` Christoph Lameter
2010-03-09 16:09         ` Christoph Lameter
2010-03-09 17:01         ` Mel Gorman [this message]
2010-03-09 17:01           ` Mel Gorman
2010-03-09 17:11           ` Christoph Lameter
2010-03-09 17:11             ` Christoph Lameter
2010-03-09 17:30             ` Mel Gorman
2010-03-09 17:30               ` Mel Gorman
2010-03-08 11:48 ` [PATCH 2/3] page-allocator: Check zone pressure when batch of pages are freed Mel Gorman
2010-03-08 11:48   ` Mel Gorman
2010-03-09  9:53   ` Nick Piggin
2010-03-09  9:53     ` Nick Piggin
2010-03-09 10:08     ` Mel Gorman
2010-03-09 10:08       ` Mel Gorman
2010-03-09 10:23       ` Nick Piggin
2010-03-09 10:23         ` Nick Piggin
2010-03-09 10:36         ` Mel Gorman
2010-03-09 10:36           ` Mel Gorman
2010-03-09 11:11           ` Nick Piggin
2010-03-09 11:11             ` Nick Piggin
2010-03-09 11:29             ` Mel Gorman
2010-03-09 11:29               ` Mel Gorman
2010-03-08 11:48 ` [PATCH 3/3] vmscan: Put kswapd to sleep on its own waitqueue, not congestion Mel Gorman
2010-03-08 11:48   ` Mel Gorman
2010-03-09 10:00   ` Nick Piggin
2010-03-09 10:00     ` Nick Piggin
2010-03-09 10:21     ` Mel Gorman
2010-03-09 10:21       ` Mel Gorman
2010-03-09 10:32       ` Nick Piggin
2010-03-09 10:32         ` Nick Piggin
2010-03-11 23:41 ` [RFC PATCH 0/3] Avoid the use of congestion_wait under zone pressure Andrew Morton
2010-03-11 23:41   ` Andrew Morton
2010-03-12  6:39   ` Christian Ehrhardt
2010-03-12  6:39     ` Christian Ehrhardt
2010-03-12  7:05     ` Andrew Morton
2010-03-12  7:05       ` Andrew Morton
2010-03-12 10:47       ` Mel Gorman
2010-03-12 10:47         ` Mel Gorman
2010-03-12 12:15         ` Christian Ehrhardt
2010-03-12 12:15           ` Christian Ehrhardt
2010-03-12 14:37           ` Andrew Morton
2010-03-12 14:37             ` Andrew Morton
2010-03-15 12:29             ` Mel Gorman
2010-03-15 12:29               ` Mel Gorman
2010-03-15 14:45               ` Christian Ehrhardt
2010-03-15 14:45                 ` Christian Ehrhardt
2010-03-15 12:34             ` Christian Ehrhardt
2010-03-15 12:34               ` Christian Ehrhardt
2010-03-15 20:09               ` Andrew Morton
2010-03-15 20:09                 ` Andrew Morton
2010-03-16 10:11                 ` Mel Gorman
2010-03-16 10:11                   ` Mel Gorman
2010-03-18 17:42                 ` Mel Gorman
2010-03-18 17:42                   ` Mel Gorman
2010-03-22 23:50                 ` Mel Gorman
2010-03-22 23:50                   ` Mel Gorman
2010-03-23 14:35                   ` Christian Ehrhardt
2010-03-23 14:35                     ` Christian Ehrhardt
2010-03-23 21:35                   ` Corrado Zoccolo
2010-03-23 21:35                     ` Corrado Zoccolo
2010-03-24 11:48                     ` Mel Gorman
2010-03-24 11:48                       ` Mel Gorman
2010-03-24 12:56                       ` Corrado Zoccolo
2010-03-24 12:56                         ` Corrado Zoccolo
2010-03-23 22:29                   ` Rik van Riel
2010-03-23 22:29                     ` Rik van Riel
2010-03-24 14:50                     ` Mel Gorman
2010-03-24 14:50                       ` Mel Gorman
2010-04-19 12:22                       ` Christian Ehrhardt
2010-04-19 12:22                         ` Christian Ehrhardt
2010-04-19 21:44                         ` Johannes Weiner
2010-04-19 21:44                           ` Johannes Weiner
2010-04-20  7:20                           ` Christian Ehrhardt
2010-04-20  7:20                             ` Christian Ehrhardt
2010-04-20  8:54                             ` Christian Ehrhardt
2010-04-20  8:54                               ` Christian Ehrhardt
2010-04-20 15:32                             ` Johannes Weiner
2010-04-20 15:32                               ` Johannes Weiner
2010-04-20 17:22                               ` Rik van Riel
2010-04-20 17:22                                 ` Rik van Riel
2010-04-21  4:23                                 ` Christian Ehrhardt
2010-04-21  4:23                                   ` Christian Ehrhardt
2010-04-21  7:35                                   ` Christian Ehrhardt
2010-04-21  7:35                                     ` Christian Ehrhardt
2010-04-21 13:19                                     ` Rik van Riel
2010-04-21 13:19                                       ` Rik van Riel
2010-04-22  6:21                                       ` Christian Ehrhardt
2010-04-22  6:21                                         ` Christian Ehrhardt
2010-04-26 10:59                                         ` Subject: [PATCH][RFC] mm: make working set portion that is protected tunable v2 Christian Ehrhardt
2010-04-26 10:59                                           ` Christian Ehrhardt
2010-04-26 11:59                                           ` KOSAKI Motohiro
2010-04-26 11:59                                             ` KOSAKI Motohiro
2010-04-26 12:43                                             ` Christian Ehrhardt
2010-04-26 12:43                                               ` Christian Ehrhardt
2010-04-26 14:20                                               ` Rik van Riel
2010-04-26 14:20                                                 ` Rik van Riel
2010-04-27 14:00                                                 ` Christian Ehrhardt
2010-04-27 14:00                                                   ` Christian Ehrhardt
2010-04-21  9:03                                   ` [RFC PATCH 0/3] Avoid the use of congestion_wait under zone pressure Johannes Weiner
2010-04-21  9:03                                     ` Johannes Weiner
2010-04-21 13:20                                   ` Rik van Riel
2010-04-21 13:20                                     ` Rik van Riel
2010-04-20 14:40                           ` Rik van Riel
2010-04-20 14:40                             ` Rik van Riel
2010-03-24  2:38                   ` Greg KH
2010-03-24  2:38                     ` Greg KH
2010-03-24 11:49                     ` Mel Gorman
2010-03-24 11:49                       ` Mel Gorman
2010-03-24 13:13                   ` Johannes Weiner
2010-03-24 13:13                     ` Johannes Weiner
2010-03-12  9:09   ` Mel Gorman
2010-03-12  9:09     ` Mel Gorman

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=20100309170123.GG4883@csn.ul.ie \
    --to=mel@csn.ul.ie \
    --cc=chris.mason@oracle.com \
    --cc=cl@linux-foundation.org \
    --cc=ehrhardt@linux.vnet.ibm.com \
    --cc=jens.axboe@oracle.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=npiggin@suse.de \
    /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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.