linux-kernel - Re: [PATCH v7 00/11] kallsyms: Optimizes the performance of lookup symbols

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] [day] [month] [year] [list]

Date:   Wed, 2 Nov 2022 17:18:25 +0800
From:   "Leizhen (ThunderTown)" <thunder.leizhen@...wei.com>
To:     Luis Chamberlain <mcgrof@...nel.org>
CC:     Josh Poimboeuf <jpoimboe@...nel.org>,
        Jiri Kosina <jikos@...nel.org>,
        Miroslav Benes <mbenes@...e.cz>,
        Petr Mladek <pmladek@...e.com>,
        Joe Lawrence <joe.lawrence@...hat.com>,
        <live-patching@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
        Masahiro Yamada <masahiroy@...nel.org>,
        Alexei Starovoitov <ast@...nel.org>,
        Jiri Olsa <jolsa@...nel.org>,
        Kees Cook <keescook@...omium.org>,
        Andrew Morton <akpm@...ux-foundation.org>,
        <linux-modules@...r.kernel.org>,
        Steven Rostedt <rostedt@...dmis.org>,
        "Ingo Molnar" <mingo@...hat.com>
Subject: Re: [PATCH v7 00/11] kallsyms: Optimizes the performance of lookup
 symbols



On 2022/10/31 23:04, Leizhen (ThunderTown) wrote:
> Now, we need to make a decision. Choose one of the two:
> 1. Continue with my current approach. Improve the average performance of
>    kallsyms_lookup_name() by 20 to 30 times. The memory overhead is increased by:
>    arm64 (defconfig):
>      73.5KiB and 4.0% if CONFIG_KALLSYMS_ALL=y.
>      19.8KiB and 2.8% if CONFIG_KALLSYMS_ALL=n.
>    x86 (defconfig):
>      49.0KiB and 3.0% if CONFIG_KALLSYMS_ALL=y.
>      16.8KiB and 2.3% if CONFIG_KALLSYMS_ALL=n.
> 2. Sort names, binary search (The static function causes duplicate names. Additional work is required）
>    2^18=262144, only up to 18 symbol expansions and comparisons are required.
>    The performance is definitely excellent, although I haven't tested it yet.
>    The memory overhead is increased by: 6 * kallsyms_num_syms
>    arm64 (defconfig):
>        1MiB if CONFIG_KALLSYMS_ALL=y.
>      362KiB if CONFIG_KALLSYMS_ALL=n.
>    x86 (defconfig):
>      770KiB if CONFIG_KALLSYMS_ALL=y.
>      356KiB if CONFIG_KALLSYMS_ALL=n.

Hi, Luis:
  I've implemented v8 based on method 2(Sort names, binary search).
The memory overhead is increased by: 3 * kallsyms_num_syms.
kallsyms_offsets_of_names[] is not added in v8 because it does not help
much in performance improvement, so save (3 * kallsyms_num_syms) bytes.
For details about the performance data, please see the commit message.


-- 
Regards,
  Zhen Lei