• Java4Shell

    From =?UTF-8?Q?Arne_Vajh=C3=B8j?=@arne@vajhoej.dk to comp.os.vms on Wed Nov 18 08:34:30 2026
    From Newsgroup: comp.os.vms

    (now we started the *Shell discussion)

    Yesterday the internet started buzzing about a new serious Java
    vulnerability Java4Shell.

    Base on what I have read then the vulnerability is in:
    * base64 decoding (which due to HTTP Basic Authentication means
    that it impacts all HTTP servers among other)
    * impacts x86-64
    * impacts at least Java 17, 21 and 25 - possible also very late
    builds of 8

    The fix is to either update to a patched version or run Java with: -XX:-UseBASE64Intrinsics

    The vulnerability was introduced when the normal Java code for base64
    decoding was replaced by an assembler version utilizing AVX-512. That assembler code has a bug so that certain invalid input causes memory corruption that can lead to RCE.

    Arne

    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Simon Clubley@clubley@remove_me.eisner.decus.org-Earth.UFP to comp.os.vms on Wed Nov 18 20:33:19 2026
    From Newsgroup: comp.os.vms

    On 2026-11-18, Arne Vajhoj <arne@vajhoej.dk> wrote:
    (now we started the *Shell discussion)

    Yesterday the internet started buzzing about a new serious Java vulnerability Java4Shell.

    Base on what I have read then the vulnerability is in:
    * base64 decoding (which due to HTTP Basic Authentication means
    that it impacts all HTTP servers among other)
    * impacts x86-64
    * impacts at least Java 17, 21 and 25 - possible also very late
    builds of 8

    The fix is to either update to a patched version or run Java with: -XX:-UseBASE64Intrinsics

    The vulnerability was introduced when the normal Java code for base64 decoding was replaced by an assembler version utilizing AVX-512. That assembler code has a bug so that certain invalid input causes memory corruption that can lead to RCE.


    Seriously not impressed by this based on what I have read so far.

    Someone has forgotten the purpose of Java and the mission-critical
    environments this language is used in and has made the wrong tradeoff
    here for a little bit of optimisation.

    This is the kind of thing that should have stayed as pure Java in the
    name of robustness and security. Somebody _really_ screwed up here by hand-writing assembly language code to replace this.

    This is the mindset of arrogance over the mindset of security and robustness.

    Very seriously not impressed.

    Simon.
    --
    Simon Clubley, clubley@remove_me.eisner.decus.org-Earth.UFP
    Walking destinations on a map are further away than they appear.
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From =?UTF-8?Q?Arne_Vajh=C3=B8j?=@arne@vajhoej.dk to comp.os.vms on Wed Nov 18 15:55:27 2026
    From Newsgroup: comp.os.vms

    On 11/18/2026 3:33 PM, Simon Clubley wrote:
    On 2026-11-18, Arne Vajh|+j <arne@vajhoej.dk> wrote:
    (now we started the *Shell discussion)

    Yesterday the internet started buzzing about a new serious Java
    vulnerability Java4Shell.

    Base on what I have read then the vulnerability is in:
    * base64 decoding (which due to HTTP Basic Authentication means
    that it impacts all HTTP servers among other)
    * impacts x86-64
    * impacts at least Java 17, 21 and 25 - possible also very late
    builds of 8

    The fix is to either update to a patched version or run Java with:
    -XX:-UseBASE64Intrinsics

    The vulnerability was introduced when the normal Java code for base64
    decoding was replaced by an assembler version utilizing AVX-512. That
    assembler code has a bug so that certain invalid input causes memory
    corruption that can lead to RCE.

    Seriously not impressed by this based on what I have read so far.

    Someone has forgotten the purpose of Java and the mission-critical environments this language is used in and has made the wrong tradeoff
    here for a little bit of optimisation.

    This is the kind of thing that should have stayed as pure Java in the
    name of robustness and security. Somebody _really_ screwed up here by hand-writing assembly language code to replace this.

    This is the mindset of arrogance over the mindset of security and robustness.

    Very seriously not impressed.
    I find it very hard to imagine any application where base64 decoding
    is performance critical.

    So likely the change made the base64 decoding 10 times faster
    and the overall application 0.01% faster.

    Arne

    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Lawrence =?iso-8859-13?q?D=FFOliveiro?=@ldo@nz.invalid to comp.os.vms on Thu Nov 19 00:52:43 2026
    From Newsgroup: comp.os.vms

    Apparently the vulnerability is only triggered if your CPU has certain
    hardware capabilities that are needed by the buggy code.

    The capabilities have something to do with rCLAVXrCY or rCLBMIrCY. HererCOs my little command for checking for the relevant flags:

    lscpu -J | jq '.lscpu | map(select(.field == "Flags:")) | .[0].data | split(" ") | map(select(contains("avx") or contains("bmi")))'

    Others have posted commands involving grepping /proc/cpuinfo, but me,
    I got to take it to the next level. EfyA
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From =?UTF-8?Q?Arne_Vajh=C3=B8j?=@arne@vajhoej.dk to comp.os.vms on Wed Nov 18 20:11:24 2026
    From Newsgroup: comp.os.vms

    On 11/18/2026 7:52 PM, Lawrence DrCOOliveiro wrote:
    Apparently the vulnerability is only triggered if your CPU has certain hardware capabilities that are needed by the buggy code.

    The capabilities have something to do with rCLAVXrCY or rCLBMIrCY. HererCOs my
    little command for checking for the relevant flags:

    lscpu -J | jq '.lscpu | map(select(.field == "Flags:")) | .[0].data | split(" ") | map(select(contains("avx") or contains("bmi")))'

    Others have posted commands involving grepping /proc/cpuinfo, but me,
    I got to take it to the next level. EfyA

    I believe it is AVX-512 VBMI.

    And all newer x86-64 CPU's should have it.

    (Intel 2018- and AMD 2022- per some wikipedia browsing)

    Arne




    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Michael S@already5chosen@yahoo.com to comp.os.vms on Thu Nov 19 22:41:33 2026
    From Newsgroup: comp.os.vms

    On Wed, 18 Nov 2026 20:11:24 -0500
    Arne Vajh|+j <arne@vajhoej.dk> wrote:
    On 11/18/2026 7:52 PM, Lawrence DrCOOliveiro wrote:
    Apparently the vulnerability is only triggered if your CPU has
    certain hardware capabilities that are needed by the buggy code.

    The capabilities have something to do with rCLAVXrCY or rCLBMIrCY. HererCOs my
    little command for checking for the relevant flags:

    lscpu -J | jq '.lscpu | map(select(.field == "Flags:")) |
    .[0].data | split(" ") | map(select(contains("avx") or
    contains("bmi")))'

    Others have posted commands involving grepping /proc/cpuinfo, but
    me, I got to take it to the next level. EfyA

    I believe it is AVX-512 VBMI.

    And all newer x86-64 CPU's should have it.

    (Intel 2018- and AMD 2022- per some wikipedia browsing)

    Arne




    Intel client CPUs had AVX512 for relatively short period (gen 11 and
    12). No AVX512 in gen 13 and newer. Should re-appear in Nova lake.
    Most Intel CPUs had it since long time ago, but there is exception -
    Xeon-6E (Sierra Forest).
    AMD CPUs (both client and server) did not have it until 2022 Q4
    (Zen4/EPYC 4).
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From =?UTF-8?Q?Arne_Vajh=C3=B8j?=@arne@vajhoej.dk to comp.os.vms on Thu Nov 19 16:23:22 2026
    From Newsgroup: comp.os.vms

    On 11/19/2026 3:41 PM, Michael S wrote:
    On Wed, 18 Nov 2026 20:11:24 -0500
    Arne Vajh|+j <arne@vajhoej.dk> wrote:

    On 11/18/2026 7:52 PM, Lawrence DrCOOliveiro wrote:
    Apparently the vulnerability is only triggered if your CPU has
    certain hardware capabilities that are needed by the buggy code.

    The capabilities have something to do with rCLAVXrCY or rCLBMIrCY. HererCOs my
    little command for checking for the relevant flags:

    lscpu -J | jq '.lscpu | map(select(.field == "Flags:")) |
    .[0].data | split(" ") | map(select(contains("avx") or
    contains("bmi")))'

    Others have posted commands involving grepping /proc/cpuinfo, but
    me, I got to take it to the next level. EfyA

    I believe it is AVX-512 VBMI.

    And all newer x86-64 CPU's should have it.

    (Intel 2018- and AMD 2022- per some wikipedia browsing)

    Intel client CPUs had AVX512 for relatively short period (gen 11 and
    12). No AVX512 in gen 13 and newer. Should re-appear in Nova lake.
    Most Intel CPUs had it since long time ago, but there is exception -
    Xeon-6E (Sierra Forest).
    AMD CPUs (both client and server) did not have it until 2022 Q4
    (Zen4/EPYC 4).
    Sorry. Most server x84-64 CPU's.

    Arne

    --- Synchronet 3.22a-Linux NewsLink 1.2