systemd VS linux

https://github.com/systemd/systemd/issues/32028#issuecomment...

There are some very compelling arguments made there if you care to read them

Maybe it was because you weren't pointing out anything new?

There was a pull request to stop linking libzma to systemd before the attack even took place

https://github.com/systemd/systemd/pull/31550

This was likely one of many things that pushed the attackers to work faster, and forced them into making mistakes.

The PR for changing compression libraries to use dlopen() was opened several weeks before the xz-utils backdoor was revealed.

https://github.com/systemd/systemd/pull/31550

I find this the most plausible explanation by far:

* The highly professional outfit simply did not see teknoraver's commit to remove liblzma as standard dependency of systemd build scripts coming.

* The race was on between their compromised code and that commit. They had to win it, with as large a window as possible.

* This caused serious errors.

* The performance regression is __not__ big. It's lucky Andres caught it at all. It's also not necessarily all that simple to remove it. It's not simply a bug in a loop or some such.

* The payload of the 'hack' contains fairly easy ways for the xz hackers to update the payload. They actually used it to remove a real issue where their hackery causes issues with valgrind that might lead to discovering it, and they also used it to release 5.6.1 which rewrites significant chunks; I've as yet not read, nor know of any analysis, as to why they changed so much.

Extra info for those who don't know:

https://github.com/systemd/systemd/commit/3fc72d54132151c131...

That's a commit that changes how liblzma is a dependency of systemd. Not because the author of this commit knew anything was wrong with it. But, pretty much entirely by accident (although removing deps was part of the point of that commit), almost entirely eliminates the value of all those 2 years of hard work.

And that was with the finish line in sight for the xz hackers: On 24 feb 2024, the xz hackers release liblzma 5.6.0 which is the first fully operational compromised version. __12 days later systemd merges a commit that means it won't work__.

So now the race is on. Can they get 5.6.0 integrated into stable releases of major OSes _before_ teknoraver's commit that removes liblzma's status as direct dep of systemd?

I find it plausible that they knew about teknoraver's commit _just before_ Feb 24th 2024 (when liblzma v5.6.0 was released, the first backdoored release), and rushed to release ASAP, before doing the testing you describe. Buoyed by their efforts to add ways to update the payload which they indeed used - March 8th (after teknoraver's commit was accepted) it was used to fix the valgrind issue.

So, no, I don't find this weird, and I don't think the amateurish aspects should be taken as some sort of indication that parts of the outfit were amateuristic. As long as it's plausible that the amateuristic aspects were simply due to time pressure, it sounds like a really bad idea to make assumptions in this regard.

sshd is started by systemd.

systemd has several ways of starting programs and waiting until they're "ready" before starting other programs that depend on them: Type=oneshot, simple, exec, forking, dbus, notify, ...

A while back, several distro maintainers found problems with using Type=exec (?) and chose Type=notify instead. When sshd is ready, it notifies systemd. How you do notification is you send a datagram to systemd's unix domain socket. That's about 10 lines of C code. But to make life even simpler, systemd also provides the one-line sd_notify() call, which is in libsystemd.so. This library is so other programmers can easily integrate with systemd.

So the distro maintainers patched sshd to use the sd_notify() function from libsystemd.so

What else is in libsystemd.so? That's right, systemd also does logging. All the logging functions are in there, so user programs can do logging the systemd way. You can even _read_ logs, using the functions in libsystemd.so. For example, sd_journal_open_files().

By the way... systemd supports the environment variable SYSTEMD_JOURNAL_COMPRESS which can be LZ4, XZ or ZSTD, to allow systemd log files to be compressed.

So, if you're a client program, that needs to read systemd logs, you'll call sd_journal_open_files() in libsystemd.so, which may then need liblz4, liblzma or libzstd functions.

These compression libraries could be dynamically loaded, should sd_journal_open_files() need them - which is what https://github.com/systemd/systemd/pull/31550 submitted on the 29th February this year did. But clearly that's not in common use. No, right now, most libsystemd.so libraries have headers saying "you'll need to load liblz4.so, liblzma.so and libzstd before you can load me!", so liblzma.so gets loaded for the logging functions that sshd doesn't use, so the distro maintainers of sshd can add 1 line instead of 10 to notify systemd that sshd is ready.

They just added an example to the documentation[0] of how to implement the sd_notify protocol without linking to libsystemd, so a little bit of discarding systemd (or at least parts of it) does seem to be part of the solution.

[0] https://github.com/systemd/systemd/pull/32030/files

I think this analysis is more interesting if you consider these two events in particular:

2024-02-29: On GitHub, @teknoraver sends pull request to stop linking liblzma into libsystemd.[1]

2024-03-20: The attacker is now a co-contributor for a patchset proposed to the Linux kernel, with the patchset adding the attacker as a maintainer and mirroring activity with xz-utils.

A theory is that the attacker saw the sshd/libsystemd/xz-utils vector as closing soon with libsystemd removing its dependency on xz-utils. When building a Linux kernel image, the resulting image is compressed by default with gzip [3], but can also be optionally compressed using xz-utils (amongst other compression utilities). There's a lot of distributions of Linux which have chosen xz-utils as the method used to compress kernel images, particularly embedded Linux distributions.[4] xz-utils is even the recommended mode of compression if a small kernel build image is desired.[5] If the attacker can execute code during the process of building a new kernel image, they can cause even more catastrophic impacts than targeting sshd. Targeting sshd was always going to be limited due to targets not exposing sshd over accessible networks, or implementing passive optical taps and real time behavioural analysis, or receiving real time alerts from servers indicative of unusual activity or data transfers. Targeting the Linux kernel would have far worse consequences possible, particularly if the attacker was targeting embedded systems (such as military transport vehicles [6]) where the chance of detection is reduced due to lack of eyeballs looking over it.

[1] https://github.com/systemd/systemd/pull/31550

[2] https://lkml.org/lkml/2024/3/20/1004

[3] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/lin...

[4] https://github.com/search?q=CONFIG_KERNEL_XZ%3Dy&type=code

[5] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/lin...

[6] https://linuxdevices.org/large-military-truck-runs-embedded-...

systemd merged a change to using dlopen for compression libraries recently https://github.com/systemd/systemd/pull/31550 which is a safer linking method in that sense.

1) Debian includes this downstream patch, also.

2) A potential explanation for "why now" is that systemd DID prevent these dependencies from loading automatically in a patch one month ago, and the patches enabling the backdoor merged a few days later. It could be a total coincidence or it could be that the attacker was trying to catch the window before it was closed on them https://github.com/systemd/systemd/pull/31550#issuecomment-1...

These are a bit easier to see what's going on:

https://github.com/torvalds/linux/commit/d5cf50dafc9dd5faa1e...

https://github.com/torvalds/linux/blob/d5cf50dafc9dd5faa1e61...

Unfortunately Github doesn't have a way to render symbols for whitespace, but you can tell by selecting the spaces that the previous version had leading tabs. Linus changed it so that the tokens `default` and the number e.g. `12` are also separated by a tab. This is tricky, because the token "default" is seven characters, it will always give this added tab a width of 1 char which makes it always layout the same as if it were a space no matter if you use tab widths of 1, 2, 4, or 8.

There was also a Gentoo effort to run atop FreeBSD[0]. The challenge of course is that afaik none of the BSD kernel ABIs are considered stable. The stable interface is the BSD libc. That said, with binfmt_misc, I don't see a reason you couldn't just run (at least some) FreeBSD binaries on Linux with a thin syscall translation layer (rather something like qemu-system) and then your layer hooked via binfmt_misc. I'm not aware of anyone who has done this for FreeBSD, but prior efforts existed as alternate binfmts for SysVr4/5 ELF binaries[2]. Either way would take some elbow grease, but you *might* even be able just reuse binfmt_elf and just have a new interpreter for FreeBSD elf.

[0] https://wiki.gentoo.org/wiki/Gentoo_FreeBSD

[1] https://docs.kernel.org/admin-guide/binfmt-misc.html

[2] https://github.com/torvalds/linux/blob/master/fs/binfmt_elf....

> The original less-than check was deemed incorrect

It was only deemed incorrect because of an information leak. Not because it's a valid use-case for user space to copy smaller portions of *hwrpb into user space. https://github.com/torvalds/linux/commit/21c5977a836e399fc71...

Correct. And the combined work needs to carry the MIT license text and copyright attributions for the MIT software authors. With binary distribution it must also be overt, not hidden in some source code drop, but directly accompanying the binary.

Many people who talk about relicensing never credit the MIT developers or distribute the MIT license text. "Because it's GPL now."

I don't think that you believe that, but many developers do.

Some don't see the need for source code scans for Open Source compliance, because the license.txt says GPL, so it's GPL. Prime example is the Linux kernel. There is code under different licenses in there, but people don't even read https://github.com/torvalds/linux/blob/master/COPYING till the end ("In addition, other licenses may also apply.") and conclude it's simply GPL 2 and nothing else.

Also be aware that sublicensing is not the same as relicensing.

> Does he have something against it?

He notoriously hates GNU Emacs, yes.

https://marc.info/?m=122955159617722

https://github.com/torvalds/linux/blob/master/Documentation/...

So If we would only count code and not comments, it is only 9489 LoC Rust. Which would be about 0.03% and if we take all lines and not only LoC it would be around 0.05%

[0] https://github.com/XAMPPRocky/tokei

[1] https://github.com/torvalds/linux/commit/b401b621758e46812da...

AIUI fsync is built on futex_waitv which has been upstreamed. So this has to be more than that.

https://github.com/torvalds/linux/commit/a0eb2da92b715d0c97b...

git clone --no-checkout --depth 1 https://github.com/torvalds/linux.git $dir