lwn.net

There is a blog post on sockpuppet.org arguing that we are not prepared for the upcoming flood of high-quality, LLM-generated vulnerability reports and exploits.

Now consider the poor open source developers who, for the last 18 months, have complained about a torrent of slop vulnerability reports. I'd had mixed sympathies, but the complaints were at least empirically correct. That could change real fast. The new models find real stuff. Forget the slop; will projects be able to keep up with a steady feed of verified, reproducible, reliably-exploitable sev:hi vulnerabilities? That's what's coming down the pipe.

Everything is up in the air. The industry is sold on memory-safe software, but the shift is slow going. We've bought time with sandboxing and attack surface restriction. How well will these countermeasures hold up? A 4 layer system of sandboxes, kernels, hypervisors, and IPC schemes are, to an agent, an iterated version of the same problem. Agents will generate full-chain exploits, and they will do so soon.

Meanwhile, no defense looks flimsier now than closed source code. Reversing was already mostly a speed-bump even for entry-level teams, who lift binaries into IR or decompile them all the way back to source. Agents can do this too, but they can also reason directly from assembly. If you want a problem better suited to LLMs than bug hunting, program translation is a good place to start.

Security updates have been issued by AlmaLinux (firefox, kernel, and kernel-rt), Debian (phpseclib and roundcube), Fedora (bind, bind-dyndb-ldap, dotnet8.0, dotnet9.0, firefox, freerdp, mingw-expat, musescore, nss, ntpd-rs, perl-YAML-Syck, php-phpseclib3, polkit, pyOpenSSL, python3.12, rust, rust-cargo-rpmstatus, rust-cargo-vendor-filterer, stgit, webkitgtk, and xen), SUSE (dovecot24, ImageMagick, jupyter-nbclassic, kernel, libjxl, libsuricata8_0_4, obs-service-recompress, obs-service-tar_scm, obs-service-set_version, openbao, perl-Crypt-URandom, plexus-utils, python-pyasn1, python-PyJWT, strongswan, traefik, traefik2, and webkit2gtk3), and Ubuntu (gst-plugins-base1.0, gst-plugins-good1.0, imagemagick, pillow, pyasn1, pyjwt, and roundcube).

SystemRescue 13.00 has been released. The SystemRescue distribution is a live boot system-rescue toolkit, based on Arch Linux, for repairing systems in the event of a crash. This release includes the 6.18.20 LTS kernel, updates bcachefs tools and kernel module to 1.37.3, and many upgraded packages. See the step-by-step guide for instructions on performing common operations such as recovering files, creating disk clones, and resetting lost passwords.

Version 4.0.0 of the Rspamd spam-filtering system has been released. Notable new features include HTML fuzzy phishing detection, support for up to eight flags with fuzzy hashes, and more. See the changelog for more on improvements, breaking changes, and bug fixes.

Rust's compiler team has been working on a long-term project to rewrite the trait solver — the part of the compiler that determines which concrete function should be called when a programmer uses a trait method that is implemented for multiple types. The rewrite is intended to simplify future changes to the trait system, fix a handful of tricky soundness bugs, and provide faster compile times. It's also nearly finished, with a relatively small number of remaining blocking bugs.

Security updates have been issued by AlmaLinux (freerdp, golang, and ncurses), Debian (asterisk, bind9, gst-plugins-base1.0, gst-plugins-ugly1.0, gvfs, incus, libxml-parser-perl, nodejs, php-phpseclib, php-phpseclib3, phpseclib, and strongswan), Fedora (bcftools, bind, bind-dyndb-ldap, chromium, dotnet10.0, dotnet8.0, dotnet9.0, giflib, htslib, libsoup3, libtasn1, maturin, mingw-expat, mingw-freetype, mongo-c-driver, perl-XML-Parser, php-phpseclib, php-phpseclib3, pypy, pypy3.10, pypy3.11, python-cryptography, python-fastar, python-ply, python-pycparser, python-uv-build, python3.11, python3.12, python3.13, python3.6, roundcubemail, rubygem-json, rust-ambient-id, rust-astral-reqwest-middleware, rust-astral-reqwest-retry, rust-astral-tokio-tar, rust-astral_async_http_range_reader, rust-cargo-c, rust-ingredients, rust-native-tls, rust-nix, rust-openssl-probe, rust-openssl-probe0.1, rust-pty-process, rust-reqsign, rust-reqsign-aliyun-oss, rust-reqsign-aws-v4, rust-reqsign-azure-storage, rust-reqsign-command-execute-tokio, rust-reqsign-core, rust-reqsign-file-read-tokio, rust-reqsign-google, rust-reqsign-http-send-reqwest, rust-reqsign-huaweicloud-obs, rust-reqsign-tencent-cos, rust-rustls-native-certs, rust-sequoia-chameleon-gnupg, rust-tar, rust-webpki-root-certs, rustup, samtools, suricata, uv, and vim), Mageia (cmake, libpng, nodejs, python-ujson, and strongswan), Red Hat (python3 and python3.9), SUSE (389-ds, amazon-cloudwatch-agent, capstone, chromium, containerd, cosign, curl, docker-compose, docker-stable, exiv2, expat, firefox, freeipmi, freerdp, gimp, glusterfs, govulncheck-vulndb, gstreamer-plugins-ugly, jupyter-bqplot-jupyterlab, jupyter-jupyterlab-templates, jupyter-matplotlib, kea, kernel, libsodium, libtpms-devel, LibVNCServer, nghttp2, nginx, poppler, python-dynaconf, python-ldap, python-nltk, python-orjson, python-pyasn1, python-pydicom, python-PyJWT, python-pyopenssl, python-tornado6, python311, python311-cbor2, python311-deepdiff, python311-intake, python311-jsonpath-ng, python311-lmdb, python311-oci-sdk, python312, rclone, redis, salt, tomcat11, v2ray-core, and vim), and Ubuntu (linux-ibm-5.4).

The 7.0-rc6 kernel prepatch is out for testing.

Anyway, exactly because it's just "more than usual" rather than feeling *worse* than usual, I don't currently feel this merits extending the release, and I still hope that next weekend will be the last rc. But it's just a bit unnerving how this release doesn't want to calm down, so no promises.

LiteLLM is a gateway library providing access to a number of large language models (LLMs); it is popular and widely used. On March 24, the word went out that the version of LiteLLM found in the Python Package Index (PyPI) repository had been compromised with information-stealing malware and downloaded thousands of times, sparking concern across the net. This may look like just another supply-chain attack — and it is — but the way it came about reveals just how many weak links there are in the software supply chains that we all depend on.

The SafeDep blog reports that compromised versions of the telnyx package have been found in the PyPI repository:

Two versions of telnyx (4.87.1 and 4.87.2) published to PyPI on March 27, 2026 contain malicious code injected into telnyx/_client.py. The telnyx package averages over 1 million downloads per month (~30,000/day), making this a high-impact supply chain compromise. The payload downloads a second-stage binary hidden inside WAV audio files from a remote server, then either drops a persistent executable on Windows or harvests credentials on Linux/macOS.

Greg Kroah-Hartman has announced the release of the 6.12.79 stable kernel. This release only reverts a patch that caused a regression on the LoongArch platform; users who could not build 6.12.78 on LoongArch need to upgrade.

Security updates have been issued by AlmaLinux (389-ds:1.4, gnutls, mysql:8.0, mysql:8.4, nginx, nginx:1.24, opencryptoki, python3, vim, and virt:rhel and virt-devel:rhel), Debian (firefox-esr, ruby-rack, and thunderbird), Fedora (fontforge, headscale, kryoptic, libopenmpt, pyOpenSSL, python-cryptography, rubygem-json, rust-asn1, rust-asn1_derive, rust-cryptoki, rust-cryptoki-sys, rust-wycheproof, vim, and vtk), Oracle (freerdp, golang, mysql:8.0, and ncurses), Red Hat (osbuild-composer), Slackware (libpng and tigervnc), SUSE (chromium, frr, kea, kernel, nghttp2, pgvector, python-deepdiff, python-pyasn1, python-tornado6, python-urllib3, python3, python310, ruby2.5, salt, sqlite3, systemd, tomcat, vim, and xen), and Ubuntu (libcryptx-perl).

Tomáš Hrčka has announced that the Forgejo-based Fedora Forge is now a fully operational collaborative-development platform; it is ready for use by the larger Fedora community, which means the homegrown Pagure platform's days are numbered:

While pagure.io has been a vital part of our community for many years, the time has come to retire our homegrown forge and transition to this powerful new tool.

The final cutover is planned for Flock to Fedora 2026. We strongly encourage teams to migrate their projects well before the conference to ensure a smooth transition. The pagure.io migration is only the first step in a broader infrastructure modernization effort. By the 2027 Fedora 46 release, we plan to retire all remaining Pagure instances across the project, including the package source repositories on src.fedoraproject.org. Getting familiar with Fedora Forge now will help ensure your team is ready as the rest of the Fedora ecosystem transitions.

There is a migration guide for Fedora community members that own projects hosted on Pagure and need to move to the new forge.

A number of projects have been struggling with the question of which submissions created by large language models (LLMs), if any, should be accepted into their code base. This discussion has been further muddied by efforts to use LLM-driven reimplemention as a way to remove copyleft restrictions from a body of existing code, as recently happened with the Python chardet module. In this context, an attempt to introduce an LLM-generated implementation of the Linux ext4 filesystem into OpenBSD was always going to create some fireworks, but that project has its own, clearly defined reasons for looking askance at such submissions.

Security updates have been issued by Debian (awstats, firefox-esr, and nss), Fedora (chromium, dotnet10.0, dotnet8.0, dotnet9.0, freerdp, and wireshark), Mageia (graphicsmagick and xen), Oracle (mysql:8.4 and nginx), Red Hat (podman), Slackware (bind and tigervnc), SUSE (azure-storage-azcopy, firefox-esr, giflib, glances-common, govulncheck-vulndb, grafana, kernel, libpng16, libsoup, mumble, net-snmp, perl-Crypt-URandom, pgvector-devel, pnpm, postgresql17, Prometheus, protobuf, python-cbor2, python-Jinja2, python-simpleeval, python311-dynaconf, python311-pydicom, python313-PyMuPDF, salt, snpguest, systemd, and vim), and Ubuntu (bind9, linux-azure, linux-azure, linux-azure-6.17, linux-azure-6.8, and mbedtls).

Inside this week's LWN.net Weekly Edition:

• Front: Security collaboration; Manjaro governance; kernel development tools; PHP licensing; kernel direct map patches; sleepable BPF.
• Briefs: LiteLLM compromise; Tor in Taiwan; b4 v0.15.0; 24-hour sideloading; Agama 19; Firefox 149.0; GNOME 50; Krita 5.3.0 and 6.0.0; Quotes; ...
• Announcements: Newsletters, conferences, security updates, patches, and more.

The keynote for Sun Security Con 2026 (SunSecCon) was given by Farzan Karimi on how incident handling can go awry because of a lack of collaboration between the "good guys"—which stands in contrast to how attackers collaboratively operate. He provided some "war stories" where security incident handling had benefited from collaboration and others where it was hampered by its lack. SunSecCon was held in conjunction with SCALE 23x in Pasadena in early March.

The Tor Blog has an interesting article about the non-technical side of setting up a Tor Relay. It documents how a computer science student at National Taiwan Normal University worked with the university system to set up a relay and provides a template for future attempts:

In Taiwan, anonymous networks do not lack technical documentation or ideological support. The real scarcity is experience from actually working through the real institutional system once. Especially in an environment where academic networks are highly centralized and outbound connectivity is tightly controlled, distributed anonymous infrastructure like Tor Relays is inherently difficult to sustain.

This implementation at National Taiwan Normal University was not meant to provide a final answer for anonymous networks. It was a concrete attempt made within real-world institutions. It may not immediately improve the performance or security of anonymous networks, and it was not intended to become a directly reproducible standard process. What it did achieve was leaving behind a clearly visible path of practice—one that can be understood, referenced, and built upon.

Version 2.0 of the LibreQoS traffic-management and network operations platform has been released.

This release makes LibreQoS easier to operate, easier to understand, and much more useful for day-to-day network work. Now users can see more of what is happening across the network, troubleshoot subscriber issues with better tools, and work from a much stronger local WebUI.

This release includes many capabilities that reflect ideas and direction long championed by our late colleague, Dave Täht.

Dave's work helped shape the understanding of bufferbloat and the importance of latency under load across the networking community. His influence continues to guide both LibreQoS and the broader effort to improve Internet quality.

The project has also announced the release of the LibreQoS Bufferbloat Test v2, also dedicated to Täht. It runs in a user's browser to look at "latency under load, jitter, loss, and what those things mean for the kinds of traffic people actually care about: browsing, streaming, video calls, audio calls, backups, and gaming".

The kernel's direct map provides code running in kernel mode with direct access to all physical memory installed in the system — on 64-bit systems, at least. It obviously makes life easier for kernel developers, but the direct map also brings some problems of its own, most of which are security-related. Interest in removing at least some pages from the direct map has been simmering for years; a couple of patch sets under discussion show some use cases for memory that has been removed from the direct map, and how such memory might be efficiently managed.

Greg Kroah-Hartman has announced the release of the 6.19.10, 6.18.20, 6.12.78, 6.6.130, and 6.1.167 stable kernels. Each contains important fixes throughout the tree. Users are advised to upgrade.