Email: email@example.com • Twitter: @benhutchingsuk • Debian: benh • Gitweb: git.decadent.org.uk • Github: github.com/bwhacks
Over the last week the Debian kernel team has worked to mitigate Meltdown in all suites. This mitigation is currently limited to kernels running in 64-bit mode (amd64 architecture), but the issue affects 32-bit mode as well.
You can see where this mitigation is applied on the security tracker. As of today, wheezy, jessie, jessie-backports, stretch and unstable/sid are fixed while stretch-backports, testing/buster and experimental are not.
Spectre needs to be mitigated in the kernel, browsers, and potentially other software. Currently the kernel changes to mitigate it are still under discussion upstream. Mozilla has started mitigating Spectre in Firefox and some of these changes are now in Debian unstable (version 57.0.4-1). Chromium has also started mitigating Spectre but no such changes have landed in Debian yet.
I was assigned 14 hours of work by Freexian's Debian LTS initiative, but only worked 6 hours so I carried over 8 hours to January.
I prepared and uploaded an update to the Linux kernel to fix various security issues. I issued DLA-1200-1 for this update. I also prepared another update on the Linux 3.2 longterm stable branch, though most of that work was done while on holiday so I didn't count the hours. I spent some time following the closed mailing list used to coordinate backports of KPTI/KAISER.
Since Debian 9 "stretch", we've shipped a Linux kernel supporting the "enhanced BPF" feature which allows unprivileged user space to upload code into the kernel. This code is written in a restricted language, but one that's much richer than the older "classic" BPF. The kernel verifies that the code is safe (doesn't loop, only accesses memory it is supposed to, etc.) before running it. However, this means that bugs in the verifier could allow unsafe programs to compromise the kernel's security.
Unfortunately, Jann Horn and others recently found many such bugs in Linux 4.14, and some of them affect older versions too. As a mitigation, consider setting the sysctl kernel.unprivileged_bpf_disabled=1. Updated packages will be available shortly.
Update: There is a public exploit that uses several of these bugs to get root privileges. It doesn't work as-is on stretch with the Linux 4.9 kernel, but is easy to adapt. I recommend applying the above mitigation as soon as possible to all systems running Linux 4.4 or later.
I was assigned 13 hours of work by Freexian's Debian LTS initiative
and carried over 4 hours from
September. I worked all 17 hours.
I prepared and released two updates on the Linux 3.2 longterm stable branch (3.2.95, 3.2.96), but I didn't upload an update to Debian. However, I have rebased the Debian package on 3.2.96 and expect to make a new upload soon.
Last week I attended Cambridge's annual mini-DebConf. It's slightly strange to visit a place one has lived in for a long time but which is no longer home. I joined Nattie in the 'video team house' which was rented for the whole week; I only went for four days.
I travelled down on Wednesday night, and spent a long time (rather longer than planned) on trains and in waiting rooms. I used this time to catch up on discussions about signing infrastructure for Secure Boot, explaining my concerns with the most recent proposal and proposing some changes that might alleviate those. Sorry to everyone who was waiting for that; I should have replied earlier.
On the Thursday and Friday I prepared for my talk, and had some conversations with Steve McIntyre and others about SB signing infrastructure. Nattie and Andy respectively organised group dinners at the Polish club on Thursday and a curry house on Friday, both of which I enjoyed.
The mini-DebConf proper took place on the Saturday and Sunday, and I presented my now annual talk on "What's new in the Linux kernel". As usual, the video team did a fine job of recording and publishing video of the talks.