From: Andrea Arcangeli <andrea@suse.de>
To: Christoph Lameter <clameter@sgi.com>
Cc: linux-mm@kvack.org
Subject: Re: [PATCH 10 of 16] stop useless vm trashing while we wait the TIF_MEMDIE task to exit
Date: Sat, 9 Jun 2007 16:05:52 +0200 [thread overview]
Message-ID: <20070609140552.GA7130@v2.random> (raw)
In-Reply-To: <Pine.LNX.4.64.0706082000370.5145@schroedinger.engr.sgi.com>
On Fri, Jun 08, 2007 at 08:01:58PM -0700, Christoph Lameter wrote:
> On Sat, 9 Jun 2007, Andrea Arcangeli wrote:
>
> > I'm sorry to inform you that the oom killing in current mainline has
> > always been a global event not a per-node one, regardless of the fixes
> > I just posted.
>
> Wrong. The oom killling is a local event if we are in a constrained
> allocation. The allocating task is killed not a random task. That call to
> kill the allocating task should not set any global flags.
I just showed the global flag that is being checked. TIF_MEMDIE
affects the whole system, not just your node-constrained allocating
task. If your local constrained task fails to exit because it's
running in the nfs path that loops forever even if NULL is returned
from alloc_pages, it will deadlock the whole system if later a regular
oom happens (alloc_pages isn't guaranteed to be called by a page fault
where we know do_exit will guaranteed to be called if a sigkill is
pending). This is just an example.
Amittedly my fixes made things worse for your "local" oom killing, but
your code was only apparently "local" because TIF_MEMDIE is a _global_
flag in the mainline kernel. So again, I'm very willing to improve the
local oom killing, so that it will really become a local event for the
first time ever. Infact with my fixes applied the whole system will
stop waiting for the TIF_MEMDIE flag to go away, so it'll be much
easier to really make the global oom killing independent from the
local one. I didn't look into the details of the local oom killing yet
(exactly because it wasn't so local in the first place) but it may be
enough to set VM_is_OOM only for tasks that are not being locally
killed and then those new changes will automatically prevent
TIF_MEMDIE being set on a local-oom to affect the global-oom event.
--
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>
next prev parent reply other threads:[~2007-06-09 14:05 UTC|newest]
Thread overview: 77+ messages / expand[flat|nested] mbox.gz Atom feed top
2007-06-08 20:02 [PATCH 00 of 16] OOM related fixes Andrea Arcangeli
2007-06-08 20:02 ` [PATCH 01 of 16] remove nr_scan_inactive/active Andrea Arcangeli
2007-06-10 17:36 ` Rik van Riel
2007-06-10 18:17 ` Andrea Arcangeli
2007-06-11 14:58 ` Rik van Riel
2007-06-26 17:08 ` Rik van Riel
2007-06-26 17:55 ` Andrew Morton
2007-06-26 19:02 ` Rik van Riel
2007-06-28 22:44 ` Rik van Riel
2007-06-28 22:57 ` Andrew Morton
2007-06-28 23:04 ` Rik van Riel
2007-06-28 23:13 ` Andrew Morton
2007-06-28 23:16 ` Rik van Riel
2007-06-28 23:29 ` Andrew Morton
2007-06-29 0:00 ` Rik van Riel
2007-06-29 0:19 ` Andrew Morton
2007-06-29 0:45 ` Rik van Riel
2007-06-29 1:12 ` Andrew Morton
2007-06-29 1:20 ` Rik van Riel
2007-06-29 1:29 ` Andrew Morton
2007-06-28 23:25 ` Andrea Arcangeli
2007-06-29 0:12 ` Andrew Morton
2007-06-29 13:38 ` Lee Schermerhorn
2007-06-29 14:12 ` Andrea Arcangeli
2007-06-29 14:59 ` Rik van Riel
2007-06-29 22:39 ` "Noreclaim Infrastructure" [was Re: [PATCH 01 of 16] remove nr_scan_inactive/active] Lee Schermerhorn
2007-06-29 22:42 ` RFC "Noreclaim Infrastructure - patch 1/3 basic infrastructure" Lee Schermerhorn
2007-06-29 22:44 ` RFC "Noreclaim Infrastructure patch 2/3 - noreclaim statistics..." Lee Schermerhorn
2007-06-29 22:49 ` "Noreclaim - client patch 3/3 - treat pages w/ excessively references anon_vma as nonreclaimable" Lee Schermerhorn
2007-06-26 20:37 ` [PATCH 01 of 16] remove nr_scan_inactive/active Andrea Arcangeli
2007-06-26 20:57 ` Rik van Riel
2007-06-26 22:21 ` Andrea Arcangeli
2007-06-08 20:03 ` [PATCH 02 of 16] avoid oom deadlock in nfs_create_request Andrea Arcangeli
2007-06-10 17:38 ` Rik van Riel
2007-06-10 18:27 ` Andrea Arcangeli
2007-06-08 20:03 ` [PATCH 03 of 16] prevent oom deadlocks during read/write operations Andrea Arcangeli
2007-06-08 20:03 ` [PATCH 04 of 16] serialize oom killer Andrea Arcangeli
2007-06-09 6:43 ` Peter Zijlstra
2007-06-09 15:27 ` Andrea Arcangeli
2007-06-08 20:03 ` [PATCH 05 of 16] avoid selecting already killed tasks Andrea Arcangeli
2007-06-08 20:03 ` [PATCH 06 of 16] reduce the probability of an OOM livelock Andrea Arcangeli
2007-06-08 20:03 ` [PATCH 07 of 16] balance_pgdat doesn't return the number of pages freed Andrea Arcangeli
2007-06-08 20:03 ` [PATCH 08 of 16] don't depend on PF_EXITING tasks to go away Andrea Arcangeli
2007-06-08 20:03 ` [PATCH 09 of 16] fallback killing more tasks if tif-memdie doesn't " Andrea Arcangeli
2007-06-08 21:57 ` Christoph Lameter
2007-06-08 20:03 ` [PATCH 10 of 16] stop useless vm trashing while we wait the TIF_MEMDIE task to exit Andrea Arcangeli
2007-06-08 21:48 ` Christoph Lameter
2007-06-09 1:59 ` Andrea Arcangeli
2007-06-09 3:01 ` Christoph Lameter
2007-06-09 14:05 ` Andrea Arcangeli [this message]
2007-06-09 14:38 ` Andrea Arcangeli
2007-06-11 16:07 ` Christoph Lameter
2007-06-11 16:50 ` Andrea Arcangeli
2007-06-11 16:57 ` Christoph Lameter
2007-06-11 17:51 ` Andrea Arcangeli
2007-06-11 17:56 ` Christoph Lameter
2007-06-11 18:22 ` Andrea Arcangeli
2007-06-11 18:39 ` Christoph Lameter
2007-06-11 18:58 ` Andrea Arcangeli
2007-06-11 19:25 ` Christoph Lameter
2007-06-11 16:04 ` Christoph Lameter
2007-06-08 20:03 ` [PATCH 11 of 16] the oom schedule timeout isn't needed with the VM_is_OOM logic Andrea Arcangeli
2007-06-08 20:03 ` [PATCH 12 of 16] show mem information only when a task is actually being killed Andrea Arcangeli
2007-06-08 20:03 ` [PATCH 13 of 16] simplify oom heuristics Andrea Arcangeli
2007-06-08 20:03 ` [PATCH 14 of 16] oom select should only take rss into account Andrea Arcangeli
2007-06-10 17:17 ` Rik van Riel
2007-06-10 17:30 ` Andrea Arcangeli
2007-06-08 20:03 ` [PATCH 15 of 16] limit reclaim if enough pages have been freed Andrea Arcangeli
2007-06-10 17:20 ` Rik van Riel
2007-06-10 17:32 ` Andrea Arcangeli
2007-06-10 17:52 ` Rik van Riel
2007-06-11 16:23 ` Christoph Lameter
2007-06-11 16:57 ` Rik van Riel
2007-06-08 20:03 ` [PATCH 16 of 16] avoid some lock operation in vm fast path Andrea Arcangeli
2007-06-08 21:26 ` [PATCH 00 of 16] OOM related fixes William Lee Irwin III
2007-06-09 14:55 ` Andrea Arcangeli
2007-06-12 8:58 ` Petr Tesarik
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=20070609140552.GA7130@v2.random \
--to=andrea@suse.de \
--cc=clameter@sgi.com \
--cc=linux-mm@kvack.org \
/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