linux-kernel - Re: [PATCH 1/1] mm: compaction: avoid fast_isolate_around() to set pageblock

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: <X76tKLSpxKqwrmin@redhat.com>
Date:   Wed, 25 Nov 2020 14:14:48 -0500
From:   Andrea Arcangeli <aarcange@...hat.com>
To:     Mike Rapoport <rppt@...ux.ibm.com>
Cc:     David Hildenbrand <david@...hat.com>,
        Vlastimil Babka <vbabka@...e.cz>, Mel Gorman <mgorman@...e.de>,
        Andrew Morton <akpm@...ux-foundation.org>, linux-mm@...ck.org,
        Qian Cai <cai@....pw>, Michal Hocko <mhocko@...nel.org>,
        linux-kernel@...r.kernel.org, Baoquan He <bhe@...hat.com>
Subject: Re: [PATCH 1/1] mm: compaction: avoid fast_isolate_around() to set
 pageblock_skip on reserved pages

On Wed, Nov 25, 2020 at 04:13:25PM +0200, Mike Rapoport wrote:
> I suspect that memmap for the reserved pages is not properly initialized
> after recent changes in free_area_init(). They are cleared at
> init_unavailable_mem() to have zone=0 and node=0, but they seem to be

I'd really like if we would not leave those to 0,0 and if we set the
whole struct page at 0xff, if we miss the second stage that corrects
the uninitialized value. The hope is that it'll crash faster and more
reproducible that way.

> never re-initialized with proper zone and node links which was not the
> case before commit 73a6e474cb37 ("mm: memmap_init: iterate over memblock
> regions rather that check each PFN").

What's strange is that 73a6e474cb37 was suggested as fix for this
bug...

	https://lkml.kernel.org/r/20200505124314.GA5029@MiWiFi-R3L-srv

The addition of "pageblock_pfn_to_page" to validate min_pfn was added
in commit 73a6e474cb37, so I assumed that the first report below
didn't have commit 73a6e474cb37 already applied.

	https://lkml.kernel.org/r/8C537EB7-85EE-4DCF-943E-3CC0ED0DF56D@lca.pw

However if you're correct perhaps the patch was already applied in
5.7.0-rc2-next-20200423+, it landed upstream in v5.8 after all.

> Back then, memmap_init_zone() looped from zone_start_pfn till
> zone_end_pfn and struct page for reserved pages with pfns inside the
> zone would be initialized.
> 
> Now the loop is for interesection of [zone_start_pfn, zone_end_pfn] with
> memblock.memory and for x86 reserved ranges are not in memblock.memory,
> so the memmap for them remains semi-initialized.

That would matches the symptoms. I'll test it as first thing after
confirming older kernels had the right zoneid/nid on the reserved
pages.

Thanks,
Andrea