linux-kernel - Re: [PATCH 3/5] x86/alternative: record .altinstructions section entity size

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, 6 Apr 2022 10:30:32 +0200
From:   Rasmus Villemoes <linux@...musvillemoes.dk>
To:     Thomas Gleixner <tglx@...utronix.de>,
        Alexey Dobriyan <adobriyan@...il.com>,
        Peter Zijlstra <peterz@...radead.org>
Cc:     x86@...nel.org, mingo@...hat.com, bp@...en8.de,
        dave.hansen@...ux.intel.com, hpa@...or.com,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH 3/5] x86/alternative: record .altinstructions section
 entity size

On 05/04/2022 21.24, Thomas Gleixner wrote:
> On Sun, Mar 13 2022 at 21:05, Alexey Dobriyan wrote:

> That does not matter. M merges any duplications in sections with the
> same [section] name, entity size and flags.
> 
>      .pushsection .bar "aM" @progbits, 4
>      .byte 0x01, 0x02, 0x03, 0x04
>      .popsection
> 
>      .pushsection .bar "aM" @progbits, 4
>      .byte 0x01, 0x02, 0x03, 0x04
>      .popsection
> 
> Will create a section .bar with lenght 4 and the content:
>      0x1,0x2,0x3,0x4
> 
> What saves you here is the fact that the altinstruction entries are
> guaranteed to be unique, but that wants a big fat comment.

Actually, I think what saves this is that the linker at least currently
ignores the merge flag for sections with relocations; from binutils
bfd/merge.c:

  if ((sec->flags & SEC_RELOC) != 0)
    {
      /* We aren't prepared to handle relocations in merged sections.  */
      return true;
    }

I do think it is theoretically possible for two altinstruction entries
to end up being identical after relocations have been applied (same
relative offsets to both the .text section and their replacement
instructions).

Rasmus