linux-kernel - Re: [BUG] fatal hang untarring 90GB file, possibly writeback related.

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <1305048553.6737.28.camel@mulgrave.site>
Date:	Tue, 10 May 2011 12:29:13 -0500
From:	James Bottomley <James.Bottomley@...senPartnership.com>
To:	Mel Gorman <mgorman@...e.de>
Cc:	Mel Gorman <mgorman@...ell.com>, Jan Kara <jack@...e.cz>,
	colin.king@...onical.com, Chris Mason <chris.mason@...cle.com>,
	linux-fsdevel <linux-fsdevel@...r.kernel.org>,
	linux-mm <linux-mm@...ck.org>,
	linux-kernel <linux-kernel@...r.kernel.org>,
	linux-ext4 <linux-ext4@...r.kernel.org>
Subject: Re: [BUG] fatal hang untarring 90GB file, possibly writeback
 related.

On Tue, 2011-05-10 at 18:17 +0100, Mel Gorman wrote:
> On Tue, May 10, 2011 at 12:05:54PM -0500, James Bottomley wrote:
> > On Tue, 2011-05-10 at 15:57 +0000, James Bottomley wrote:
> > > On Tue, 2011-05-10 at 10:29 -0500, James Bottomley wrote:
> > > > On Tue, 2011-05-10 at 15:35 +0100, Mel Gorman wrote:
> > > > > On Tue, May 10, 2011 at 09:01:04AM -0500, James Bottomley wrote:
> > > > > > On Tue, 2011-05-10 at 11:21 +0100, Mel Gorman wrote:
> > > > > > > I really would like to hear if the fix makes a big difference or
> > > > > > > if we need to consider forcing SLUB high-order allocations bailing
> > > > > > > at the first sign of trouble (e.g. by masking out __GFP_WAIT in
> > > > > > > allocate_slab). Even with the fix applied, kswapd might be waking up
> > > > > > > less but processes will still be getting stalled in direct compaction
> > > > > > > and direct reclaim so it would still be jittery.
> > > > > > 
> > > > > > "the fix" being this
> > > > > > 
> > > > > > https://lkml.org/lkml/2011/3/5/121
> > > > > > 
> > > > > 
> > > > > Drop this for the moment. It was a long shot at best and there is little
> > > > > evidence the problem is in this area.
> > > > > 
> > > > > I'm attaching two patches. The first is the NO_KSWAPD one to stop
> > > > > kswapd being woken up by SLUB using speculative high-orders. The second
> > > > > one is more drastic and prevents slub entering direct reclaim or
> > > > > compaction. It applies on top of patch 1. These are both untested and
> > > > > afraid are a bit rushed as well :(
> > > > 
> > > > Preliminary results with both patches applied still show kswapd
> > > > periodically going up to 99% but it doesn't stay there, it comes back
> > > > down again (and, obviously, the system doesn't hang).
> > > 
> > > This is a second run with the watch highorders.
> > > 
> > > At the end of the run, the system hung temporarily and now comes back
> > > with CPU3 spinning in all system time at kswapd shrink_slab
> > 
> > Here's a trace in the same situation with the ftrace stack entries
> > bumped to 16 as requested on IRC.  There was no hang for this one.
> > 
> 
> Ok, so the bulk of the high-order allocs are coming from
> 
> > 140162 instances order=1 normal gfp_flags=GFP_NOWARN|GFP_NORETRY|GFP_COMP|GFP_NOMEMALLOC|
> >  => __alloc_pages_nodemask+0x754/0x792 <ffffffff810dc0de>
> >  => alloc_pages_current+0xbe/0xd8 <ffffffff81105459>
> >  => alloc_slab_page+0x1c/0x4d <ffffffff8110c5fe>
> >  => new_slab+0x50/0x199 <ffffffff8110dc48>
> >  => __slab_alloc+0x24a/0x328 <ffffffff8146ab86>
> >  => kmem_cache_alloc+0x77/0x105 <ffffffff8110e450>
> >  => mempool_alloc_slab+0x15/0x17 <ffffffff810d6e85>
> >  => mempool_alloc+0x68/0x116 <ffffffff810d70fa>
> >  => bio_alloc_bioset+0x35/0xc3 <ffffffff81144dd8>
> >  => bio_alloc+0x15/0x24 <ffffffff81144ef5>
> >  => submit_bh+0x6d/0x105 <ffffffff811409f6>
> >  => __block_write_full_page+0x1e7/0x2d7 <ffffffff81141fac>
> >  => block_write_full_page_endio+0x8a/0x97 <ffffffff81143671>
> >  => block_write_full_page+0x15/0x17 <ffffffff81143693>
> >  => mpage_da_submit_io+0x31a/0x395 <ffffffff811935d8>
> >  => mpage_da_map_and_submit+0x2ca/0x2e0 <ffffffff81196e88>
> > 
> 
> That at least is in line with the large untar and absolves i915
> from being the main cause of trouble. The lack of the hang implies
> that SLUB doing high order allocations is stressing the system too
> much and needs to be more willing to fall back to order-0 although
> it does not adequately explain why it hung as opposed to just being
> incredible slow.
> 
> I'm also still concerned with the reports of getting stuck in a heavy
> loop on the i915 shrinker so will try again reproducing this locally
> with a greater focus on something X related happening at the same time.
> 
> One thing at a time though, SLUB needs to be less aggressive so I'll
> prepare a series in the morning, have another go at generating data
> and see what shakes out.

As a final data point, I tried slub_max_order=0.  kswapd is much more
quiescent (it calms down to nothing).  It shows spikes of up to 70% and
once up to 99% though, but it doesn't go up to 99% and stay there like
it did previously.  This is just with top, so it's a bit inaccurate.

James


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/