From: Vlastimil Babka <vbabka@suse.cz>
To: David Zarzycki <dave@znu.io>, linux-mm@kvack.org
Cc: Mel Gorman <mgorman@techsingularity.net>
Subject: Re: Transparent Huge pages hanging on 5.1.x/5.2.0 kernels?
Date: Thu, 8 Aug 2019 18:08:01 +0200 [thread overview]
Message-ID: <b440d2d6-2184-9538-4453-f1722b91e76c@suse.cz> (raw)
In-Reply-To: <E70C35A9-A757-4507-BAB1-D831A5746BBF@znu.io>
On 7/15/19 12:32 PM, David Zarzycki wrote:
> Hello,
>
> In the last few weeks, one of my build boxes started hanging at the end of a build with a zombie ld.lld process stuck in the kernel:
>
> [97199.634549] CPU: 14 PID: 72214 Comm: ld.lld Kdump: loaded Not tainted 5.2.0-1.fc31.x86_64 #1
> [97199.634550] Hardware name: Supermicro SYS-5038K-i-NF9/K1SPE, BIOS 1.0b 04/13/2017
> [97199.634551] RIP: 0010:compact_zone+0x4d0/0xce0
> [97199.634553] Code: 41 c6 47 78 01 e9 52 fc ff ff 4c 89 f7 48 89 ea 4c 89 e6 e8 22 8e 02 00 49 89 c6 e9 d7 fd ff ff 8b 4c 24 10 4c 89 e2 4c 89 ee <4c> 89 ff e8 e8 e0 ff ff 49 89 c4 48 85 c0 0f 84 bd fe ff ff 45 8b
> [97199.634555] RSP: 0018:ffffac6a53c879c0 EFLAGS: 00000202
> [97199.634557] RAX: 0000000000000001 RBX: 000000000619f200 RCX: 000000000000000c
> [97199.634558] RDX: 000000000619f000 RSI: 000000000619ee20 RDI: ffff95f77ffc8330
> [97199.634559] RBP: ffff95fb7ffd4d00 R08: 0000000000000007 R09: 000000000619f000
> [97199.634561] R10: 0000000000000000 R11: 0000000000000003 R12: 000000000619f000
> [97199.634562] R13: 000000000619ee20 R14: fffffb58467b8000 R15: ffffac6a53c87a90
> [97199.634563] FS: 00007ffff10fd700(0000) GS:ffff95f5fb780000(0000) knlGS:0000000000000000
> [97199.634566] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [97199.634567] CR2: 00007fff08001378 CR3: 00000054737f6000 CR4: 00000000001406e0
> [97199.634568] Call Trace:
> [97199.634569] compact_zone_order+0xde/0x140
This was likely the same as
https://bugzilla.kernel.org/show_bug.cgi?id=204165
Fixed by patch https://marc.info/?l=linux-mm&m=156344023621776&w=2
Now commit 670105a25608 ("mm: compaction: avoid 100% CPU usage during
compaction when a task is killed")
It should hit your distro kernel at some point.
> [97199.634570] try_to_compact_pages+0xcc/0x2a0
> [97199.634570] __alloc_pages_direct_compact+0x8c/0x170
> [97199.634571] __alloc_pages_slowpath+0x248/0xdf0
> [97199.634572] ? get_vtime_delta+0x13/0xe0
> [97199.634573] ? finish_task_switch+0x12f/0x2a0
> [97199.634574] __alloc_pages_nodemask+0x2f2/0x340
> [97199.634575] do_huge_pmd_anonymous_page+0x130/0x910
> [97199.634576] __handle_mm_fault+0xfd7/0x1ac0
> [97199.634577] handle_mm_fault+0xc4/0x1f0
> [97199.634577] do_user_addr_fault+0x1f6/0x450
> [97199.634578] do_page_fault+0x33/0x120
> [97199.634579] ? page_fault+0x8/0x30
> [97199.634580] page_fault+0x1e/0x30
>
> This bug seems to go away if I comment out the following lines from my boot script:
>
> # echo always > /sys/kernel/mm/transparent_hugepage/enabled
> # echo always > /sys/kernel/mm/transparent_hugepage/defrag
>
> What can I do to debug this further?
>
> Dave
>
prev parent reply other threads:[~2019-08-08 16:08 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-07-15 10:32 David Zarzycki
2019-08-08 16:08 ` Vlastimil Babka [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=b440d2d6-2184-9538-4453-f1722b91e76c@suse.cz \
--to=vbabka@suse.cz \
--cc=dave@znu.io \
--cc=linux-mm@kvack.org \
--cc=mgorman@techsingularity.net \
/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