Skip to content
DevMeme
7256 of 7590
Security Post #7955 · source on Telegram

Copy Fail: 732-Byte Python Script Roots Every Linux Since 2017

Description

A branded vulnerability disclosure card titled 'Copy Fail' (with 'Fail' in blue) on a white background, in the style of marketed CVE landing pages. The text reads: 'Most Linux LPEs need a race window or a kernel-specific offset. Copy Fail is a straight-line logic flaw - it needs neither. The same 732-byte Python script roots every Linux distribution shipped since 2017.' Below in smaller grey text: 'One logic bug in authencesn, chained through AF_ALG and splice() into a 4-byte page-cache write - silently exploitable for nearly a decade.' The image highlights a local privilege escalation in the Linux kernel crypto stack (authencesn via the AF_ALG userspace crypto socket API), chained with splice() to achieve a page-cache write, notable for being deterministic - no race condition or per-kernel offsets required - and for affecting nearly a decade of kernels with a tiny universal exploit

Comments

26
Anonymous ★ Top Pick A decade of KASLR, stack canaries, and lockdown LSMs, defeated by a logic bug small enough to fit in a commit message
  1. Anonymous ★ Top Pick

    A decade of KASLR, stack canaries, and lockdown LSMs, defeated by a logic bug small enough to fit in a commit message

  2. Anonymous

    Linux achieved zero-copy so efficiently it also zero-copied the permission check.

  3. @flopflopflopflopflop 3mo

    Any non-AI summary?

    1. @Rinuuri 3mo

      quote form the official site: What is Copy Fail in one sentence? An unprivileged local user can write 4 controlled bytes into the page cache of any readable file on a Linux system, and use that to gain root.

      1. @RiedleroD 3mo

        I'll say that the official website overhypes it a bit; you need a specific kernel module loaded from what I can tell. From what I know in my circle of friends, some are affected, others are not. and of course most people already got the patched kernels if they updated in the last week or 2

        1. @RiedleroD 3mo

          so this mostly affects servers that can't easily reboot or hotswap kernels

          1. @RiedleroD 3mo

            this friend understands it better than I do https://social.hackerspace.pl/@q3k/116490249266904439

            1. @RiedleroD 3mo

              also: something about "responsible disclosure" and "recent events" or something idk

  4. @blue_bonsai 3mo

    Well-played NSA, well-played.

  5. @TheFloofyFloof 3mo

    responsible disclosure is a myth by big security to sell more security

  6. @purplesyringa 3mo

    where responsible disclosure

    1. @purplesyringa 3mo

      https://www.openwall.com/lists/oss-security/2026/04/29/25

  7. @RiedleroD 3mo

    got some more opinions from my friends: "none of the ~100 servers I admin are affected" as well as "the module was loaded and the exploit functional on all of my arch linux bare metal boxes." so there's a real split going down in the community here

    1. @SamsonovAnton 3mo

      "None are affected" or "None could complete the PoC test with hard-coded paths and pre-compiled binaries for x86"?

      1. @azizhakberdiev 3mo

        I believe the preventive measures themselves were efficient enough

        1. @feedable 3mo

          being?

          1. @azizhakberdiev 3mo

            not anything specific, just no premise for that exploit

            1. @feedable 3mo

              you basically only need an unpatched kernel, that's it

              1. @azizhakberdiev 3mo

                it's the target. The premise is some malware backlog or user with malicious intent

                1. @feedable 3mo

                  that doesn't mean "not affected" tho

      2. @RiedleroD 3mo

        affected - the necessary kernel module isn't loaded

    2. dev_meme 3mo

      There additional chains to widely increase blast radius that were not made public - it’s stated by author and there’s no actual reason to not believe it

      1. @RiedleroD 3mo

        well. but those are yet to be patched, so I consider them to be separate CVEs, no?

        1. @feedable 3mo

          meaning, based on this same exploit

          1. @RiedleroD 3mo

            but the exploit relies entirely on this kernel module. the out-of-bounds access was entirely in that kernel module

            1. @RiedleroD 3mo

              as far as I understood at least

Use J and K for navigation