linux-kernel - Re: [PATCH v2] prctl: Get private anonymous memory region name

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]

Message-ID: <tencent_44E22E4955685EFF479910090D5C84E8B209@qq.com>
Date:   Tue, 28 Nov 2023 09:02:40 +0800
From:   Rong Tao <rtoax@...mail.com>
To:     David Hildenbrand <david@...hat.com>,
        Oleg Nesterov <oleg@...hat.com>
Cc:     Andrew Morton <akpm@...ux-foundation.org>,
        Catalin Marinas <catalin.marinas@....com>,
        Kees Cook <keescook@...omium.org>,
        Florent Revest <revest@...omium.org>,
        Josh Triplett <josh@...htriplett.org>,
        Andy Chiu <andy.chiu@...ive.com>,
        Joey Gouly <joey.gouly@....com>,
        Stefan Roesch <shr@...kernel.io>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Ondrej Mosnacek <omosnace@...hat.com>,
        Miguel Ojeda <ojeda@...nel.org>,
        "open list:MEMORY MANAGEMENT" <linux-mm@...ck.org>,
        open list <linux-kernel@...r.kernel.org>, rongtao@...tc.cn
Subject: Re: [PATCH v2] prctl: Get private anonymous memory region name


On 11/27/23 22:52, David Hildenbrand wrote:
> On 26.11.23 14:56, Oleg Nesterov wrote:
>> On 11/26, Rong Tao wrote:
>>>
>>> then the PR_GET_VMA interface should be provided accordingly,
>>> which is necessary, as the userspace program usually wants to know what
>>> VMA name it has configured for the anonymous page.
>>
>> I don't really understand the use-case for PR_GET_VMA ...
>>
>
> Can't we simply read "/proc/PID/maps" and just have that information 
> from there?


Thank you, David.

The relationship between PR_GET_VMA and /proc/PID/maps is like the 
relationship between

PR_GET_NAME and /proc/PID/comm. Obviously, both methods can obtain the 
corresponding

name. However, prctl(2) can be obtained directly from the code level, 
while reading proc is not

so convenient and efficient. Moreover, reading proc is more like bash 
command line, rather

than C code.


>
> Also, I don't understand the exact use case, that should be clarified 
> -- especially, why the existing way is insufficient.
>

For use-case, in fact, I now want to develop a user-mode patch tool and 
need to map the patch file to the target

process(Use ptrace(2) and pread/pwrite("/proc/self/mem")). I initially 
used shared files

     00400000-00401000 r--p 00000000 08:00 241933181 
/home/sda/git-repos/upatch/tests/hello/hello
     00401000-00402000 r-xp 00001000 08:00 241933181 
/home/sda/git-repos/upatch/tests/hello/hello
     00402000-00403000 r--p 00002000 08:00 241933181 
/home/sda/git-repos/upatch/tests/hello/hello
     00403000-00404000 r--p 00002000 08:00 241933181 
/home/sda/git-repos/upatch/tests/hello/hello
     00404000-00405000 rw-p 00003000 08:00 241933181 
/home/sda/git-repos/upatch/tests/hello/hello
     01136000-01157000 rw-p 00000000 00:00 0 [heap]
     7f21472c0000-7f21472c2000 rw-p 00000000 00:00 0
     7f21472c2000-7f21472e8000 r--p 00000000 103:03 3705 
/usr/lib64/libc.so.6
     7f21472e8000-7f2147448000 r-xp 00026000 103:03 3705 
/usr/lib64/libc.so.6
     7f2147448000-7f2147496000 r--p 00186000 103:03 3705 
/usr/lib64/libc.so.6
     7f2147496000-7f214749a000 r--p 001d3000 103:03 3705 
/usr/lib64/libc.so.6
     7f214749a000-7f214749c000 rw-p 001d7000 103:03 3705 
/usr/lib64/libc.so.6
     7f214749c000-7f21474a6000 rw-p 00000000 00:00 0
     7f21474be000-7f21474bf000 rwxs 00000000 00:27 7794 
/tmp/upatch/62984/map_files/patch-FKSYTp   <<
     7f21474bf000-7f21474c0000 rwxs 00000000 00:27 7793 
/tmp/upatch/62984/map_files/patch-KFaQNU <<
     7f21474c0000-7f21474c1000 r--p 00000000 103:03 3702 
/usr/lib64/ld-linux-x86-64.so.2
     7f21474c1000-7f21474e8000 r-xp 00001000 103:03 3702 
/usr/lib64/ld-linux-x86-64.so.2
     7f21474e8000-7f21474f2000 r--p 00028000 103:03 3702 
/usr/lib64/ld-linux-x86-64.so.2
     7f21474f2000-7f21474f4000 r--p 00031000 103:03 3702 
/usr/lib64/ld-linux-x86-64.so.2
     7f21474f4000-7f21474f6000 rw-p 00033000 103:03 3702 
/usr/lib64/ld-linux-x86-64.so.2
     7ffec158b000-7ffec15ad000 rw-p 00000000 00:00 0 [stack]
     7ffec15cf000-7ffec15d3000 r--p 00000000 00:00 0 [vvar]
     7ffec15d3000-7ffec15d5000 r-xp 00000000 00:00 0 [vdso]
     ffffffffff600000-ffffffffff601000 --xp 00000000 00:00 0 [vsyscall]

However, this is obviously not the best approach, I want to use 
anonymous pages instead,

     7f21474be000-7f21474bf000 rwxp 00000000 00:27 7794 [anon:patch1]
     7f21474bf000-7f21474c0000 rwxp 00000000 00:27 7793 [anon:patch2]

I hope to use the address to directly obtain the vma name, which is 
"patch1". This is very convenient

in the program without parsing /proc/PID/maps in the source code.

Thanks again.

Rong Tao