linux-kernel - [PATCH v12 0/2] introduce memblock_next_valid

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-next>] [day] [month] [year] [list]

Message-ID: <1563861073-47071-1-git-send-email-guohanjun@huawei.com>
Date:   Tue, 23 Jul 2019 13:51:11 +0800
From:   Hanjun Guo <guohanjun@...wei.com>
To:     Ard Biesheuvel <ard.biesheuvel@...aro.org>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Catalin Marinas <catalin.marinas@....com>,
        "Jia He" <hejianet@...il.com>, Mike Rapoport <rppt@...ux.ibm.com>,
        Will Deacon <will@...nel.org>
CC:     <linux-arm-kernel@...ts.infradead.org>, <linux-mm@...ck.org>,
        <linux-kernel@...r.kernel.org>, Hanjun Guo <guohanjun@...wei.com>
Subject: [PATCH v12 0/2] introduce memblock_next_valid_pfn() (again) for arm64

Here is new version of "[PATCH v11 0/3] remain and optimize
memblock_next_valid_pfn on arm and arm64" from Jia He, which is suggested
by Ard to respin this patch set [1].

In the new version, I squashed patch 1/3 and patch 2/3 in v11 into
one patch, fixed a bug for possible out of bound accessing the
regions, and just introduce memblock_next_valid_pfn() for arm64 only
as I don't have a arm32 platform to test.

Ard asked to "with the new data points added for documentation, and
crystal clear about how the meaning of PFN validity differs between
ARM and other architectures, and why the assumptions that the
optimization is based on are guaranteed to hold", to be honest, I
didn't see PFN validity differs between ARM and x86 architecture,
but there is a bug in commit b92df1de5d28 ("mm: page_alloc: skip over
regions of invalid pfns where possible") which has a possible out of
bound accessing the regions as well, so not sure that is the root cause.

Testing on a HiSilicon ARM64 server (a 4 sockets system), I can get
pretty much speedup for bootmem_init() at boot:

with 384G memory,
before: 13310ms
after:  1415ms

with 1T memory,
before: 20s
after:  2s

[1]: https://lkml.org/lkml/2019/6/10/412

Jia He (2):
  mm: page_alloc: introduce memblock_next_valid_pfn() (again) for arm64
  mm: page_alloc: reduce unnecessary binary search in
    memblock_next_valid_pfn

 arch/arm64/Kconfig     |  1 +
 include/linux/mmzone.h |  9 +++++++
 mm/Kconfig             |  3 +++
 mm/memblock.c          | 56 ++++++++++++++++++++++++++++++++++++++++++
 mm/page_alloc.c        |  4 ++-
 5 files changed, 72 insertions(+), 1 deletion(-)

-- 
2.19.1