From: David Hildenbrand <david@redhat.com>
To: Vlastimil Babka <vbabka@suse.cz>, Andrea Arcangeli <aarcange@redhat.com>
Cc: Mel Gorman <mgorman@suse.de>,
Andrew Morton <akpm@linux-foundation.org>,
linux-mm@kvack.org, Qian Cai <cai@lca.pw>,
Michal Hocko <mhocko@kernel.org>,
linux-kernel@vger.kernel.org, Mike Rapoport <rppt@linux.ibm.com>,
Baoquan He <bhe@redhat.com>
Subject: Re: [PATCH 1/1] mm: compaction: avoid fast_isolate_around() to set pageblock_skip on reserved pages
Date: Wed, 25 Nov 2020 14:32:02 +0100 [thread overview]
Message-ID: <cd9f0b9f-c4f6-b80c-03cd-12697324bfca@redhat.com> (raw)
In-Reply-To: <1c4c405b-52e0-cf6b-1f82-91a0a1e3dd53@suse.cz>
On 25.11.20 13:08, Vlastimil Babka wrote:
> On 11/25/20 6:34 AM, Andrea Arcangeli wrote:
>> Hello,
>>
>> On Mon, Nov 23, 2020 at 02:01:16PM +0100, Vlastimil Babka wrote:
>>> On 11/21/20 8:45 PM, Andrea Arcangeli wrote:
>>>> A corollary issue was fixed in
>>>> 39639000-39814fff : Unknown E820 type
>>>>
>>>> pfn 0x7a200 -> 0x7a200000 min_pfn hit non-RAM:
>>>>
>>>> 7a17b000-7a216fff : Unknown E820 type
>>>
>>> It would be nice to also provide a /proc/zoneinfo and how exactly the
>>> "zone_spans_pfn" was violated. I assume we end up below zone's
>>> start_pfn, but is it true?
>>
>> Agreed, I was about to grab that info along with all page struct
>> around the pfn 0x7a200 and phys address 0x7a216fff.
>>
>> # grep -A1 E820 /proc/iomem
>> 7a17b000-7a216fff : Unknown E820 type
>> 7a217000-7bffffff : System RAM
>>
>> DMA zone_start_pfn 1 zone_end_pfn() 4096 contiguous 1
>> DMA32 zone_start_pfn 4096 zone_end_pfn() 1048576 contiguous 0
>> Normal zone_start_pfn 1048576 zone_end_pfn() 4715392 contiguous 1
>> Movable zone_start_pfn 0 zone_end_pfn() 0 contiguous 0
>
> So the above means that around the "Unknown E820 type" we have:
>
> pfn 499712 - start of pageblock in ZONE_DMA32
> pfn 500091 - start of the "Unknown E820 type" range
> pfn 500224 - start of another pageblock
> pfn 500246 - end of "Unknown E820 type"
>
> So this is indeed not a zone boundary issue, but basically a hole not
> aligned to pageblock boundary and really unexpected.
> We have CONFIG_HOLES_IN_ZONE (that x86 doesn't set) for architectures
> that do this, and even that config only affects pfn_valid_within(). But
> here pfn_valid() is true, but the zone/node linkage is unexpected.
>
>> However the real bug seems that reserved pages have a zero zone_id in
>> the page->flags when it should have the real zone id/nid. The patch I
>> sent earlier to validate highest would only be needed to deal with
>> pfn_valid.
>>
>> Something must have changed more recently than v5.1 that caused the
>> zoneid of reserved pages to be wrong, a possible candidate for the
>> real would be this change below:
>>
>> + __init_single_page(pfn_to_page(pfn), pfn, 0, 0);
>>
>> Even if it may not be it, at the light of how the reserved page
>> zoneid/nid initialized went wrong, the above line like it's too flakey
>> to stay.
>>
>> It'd be preferable if the pfn_valid fails and the
>> pfn_to_section_nr(pfn) returns an invalid section for the intermediate
>> step. Even better memset 0xff over the whole page struct until the
>> second stage comes around.
>>
>> Whenever pfn_valid is true, it's better that the zoneid/nid is correct
>> all times, otherwise if the second stage fails we end up in a bug with
>> weird side effects.
>
> Yeah I guess it would be simpler if zoneid/nid was correct for
> pfn_valid() pfns within a zone's range, even if they are reserved due
> not not being really usable memory.
>
> I don't think we want to introduce CONFIG_HOLES_IN_ZONE to x86. If the
> chosen solution is to make this to a real hole, the hole should be
> extended to MAX_ORDER_NR_PAGES aligned boundaries.
As we don't punch out pages of the memmap on x86-64, pfn_valid() keeps
working as expected. There is a memmap that can be accessed and that was
initialized. It's really just a matter of how to handle memory holes in
this scenario.
a) Try initializing them to the covering node/zone (I gave one example
that might be tricky with hotplug)
b) Mark such pages (either special node/zone or pagetype) and make pfn
walkers ignore these holes. For now, this can only be done via the
reserved flag.
--
Thanks,
David / dhildenb
next prev parent reply other threads:[~2020-11-25 13:32 UTC|newest]
Thread overview: 61+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-04-23 21:25 compaction: VM_BUG_ON_PAGE(!zone_spans_pfn(page_zone(page), pfn)) Qian Cai
2020-04-24 3:43 ` Baoquan He
2020-04-24 13:45 ` Qian Cai
2020-05-05 12:43 ` Baoquan He
2020-05-05 13:20 ` Qian Cai
2020-05-11 1:21 ` Baoquan He
2020-04-26 14:41 ` Mike Rapoport
2020-04-27 13:45 ` Qian Cai
2020-11-21 19:45 ` [PATCH 0/1] VM_BUG_ON_PAGE(!zone_spans_pfn) in set_pfnblock_flags_mask Andrea Arcangeli
2020-11-21 19:45 ` [PATCH 1/1] mm: compaction: avoid fast_isolate_around() to set pageblock_skip on reserved pages Andrea Arcangeli
2020-11-21 19:53 ` Andrea Arcangeli
2020-11-23 11:26 ` David Hildenbrand
2020-11-23 13:01 ` Vlastimil Babka
2020-11-24 13:32 ` Mel Gorman
2020-11-24 20:56 ` Andrea Arcangeli
2020-11-25 10:30 ` Mel Gorman
2020-11-25 17:59 ` Andrea Arcangeli
2020-11-26 10:47 ` Mel Gorman
2020-12-06 2:26 ` Andrea Arcangeli
2020-12-06 23:47 ` Mel Gorman
2020-11-25 5:34 ` Andrea Arcangeli
2020-11-25 6:45 ` David Hildenbrand
2020-11-25 8:51 ` Mike Rapoport
2020-11-25 10:39 ` Mel Gorman
2020-11-25 11:04 ` David Hildenbrand
2020-11-25 11:41 ` David Hildenbrand
2020-11-25 18:47 ` Andrea Arcangeli
2020-11-25 13:33 ` Mel Gorman
2020-11-25 13:41 ` David Hildenbrand
2020-11-25 18:28 ` Andrea Arcangeli
2020-11-25 19:27 ` David Hildenbrand
2020-11-25 20:41 ` Andrea Arcangeli
2020-11-25 21:13 ` David Hildenbrand
2020-11-25 21:04 ` Mike Rapoport
2020-11-25 21:38 ` Andrea Arcangeli
2020-11-26 9:36 ` Mike Rapoport
2020-11-26 10:05 ` David Hildenbrand
2020-11-26 17:46 ` Mike Rapoport
2020-11-29 12:32 ` Mike Rapoport
2020-12-02 0:44 ` Andrea Arcangeli
2020-12-02 17:39 ` Mike Rapoport
2020-12-03 6:23 ` Andrea Arcangeli
2020-12-03 10:51 ` Mike Rapoport
2020-12-03 17:31 ` Andrea Arcangeli
2020-12-06 8:09 ` Mike Rapoport
2020-11-26 18:15 ` Andrea Arcangeli
2020-11-26 18:29 ` Andrea Arcangeli
2020-11-26 19:44 ` Mike Rapoport
2020-11-26 20:30 ` Andrea Arcangeli
2020-11-26 21:03 ` Mike Rapoport
2020-11-26 19:21 ` Andrea Arcangeli
2020-11-25 12:08 ` Vlastimil Babka
2020-11-25 13:32 ` David Hildenbrand [this message]
2020-11-25 14:13 ` Mike Rapoport
2020-11-25 14:42 ` David Hildenbrand
2020-11-26 10:51 ` Mel Gorman
2020-11-25 19:14 ` Andrea Arcangeli
2020-11-25 19:01 ` Andrea Arcangeli
2020-11-25 19:33 ` David Hildenbrand
2020-11-26 3:40 ` Andrea Arcangeli
2020-11-26 10:43 ` Mike Rapoport
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=cd9f0b9f-c4f6-b80c-03cd-12697324bfca@redhat.com \
--to=david@redhat.com \
--cc=aarcange@redhat.com \
--cc=akpm@linux-foundation.org \
--cc=bhe@redhat.com \
--cc=cai@lca.pw \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mgorman@suse.de \
--cc=mhocko@kernel.org \
--cc=rppt@linux.ibm.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).