Linux 4.17 Released

Linux v4.17 was released on Sunday, June 3rd; this is a quick summary of the SELinux and audit changes. After a small set of changes in Linux v4.16 we are back to a more normal, perhaps even a bit larger than normal, set of changes in Linux v4.17.

SELinux

  • While we added basic support for SCTP sockets in Linux v4.11, via the "sctp_socket" object class, proper SELinux support for SCTP was not available until this release. Richard Haines was responsible for the development of the SCTP support, and documented the new controls in the Linux Kernel's "Documentation/security" directory; I encourage you to check out the LSM-sctp.rst and SELinux-sctp.rst files in that directory for more details.
    As a reminder, ensure that your currently loaded SELinux policy has the "extended_socket_class" policy capability enabled before attempting to use the new SELinux SCTP controls.

  • Stephen Smalley submitted a number of patches to encapsulate much of the kernel's global SELinux state into a number of new data structures. While this was a very large change to the code, none of it should be user visible. It is simply an important piece of housekeeping that paves the way for a number of improvements and new features that are currently under development.

  • Fixed a small bug with setxattr(2) where we were not adequately protecting against malformed file security labels that were written by privileged applications. In addition to Linux v4.17, this fix should also start appearing in the different stable kernel releases; check the changelogs.

  • Various small code cleanups and simplifications.

Audit

  • After being marked as deprecated for several years, the audit "entry" filter was removed from the kernel. The "entry" filter was removed from the audit userspace tools in v2.6.7.

  • Fixed a small bug where creating a session ID audit filter could cause the "arch" field to be misrepresented in the various different audit records.

  • The code which interpreted the "audit=" configuration parameter on the kernel command line was improved to make it more robust and accept a wider range of user input, e.g. both "on" and "off" are now valid options.

  • Fix a number of issues relating to the auditing of symlinks. Fixes include honoring the current audit enabled status, as well proper auditing of the related PATH records.

  • Various other small bug fixes, code cleanups, and simplifications.

Linux 4.16 Released

Linux v4.16 was released on Sunday, April 1st; this is a quick summary of the SELinux and audit changes. Fair warning, this was a very small release from a SELinux and audit perspective, but v4.17 should be back to normal and perhaps even a bit larger than normal if everything works out.

SELinux

  • Fixed two small problems identified by syzbot/syzkaller that could only be triggered during early boot (SELinux enabled, but no policy loaded). It is extremely unlikely that these bugs would ever be seen by normal users as there is little point of running a system with SELinux enabled and not loading a security policy.

  • Updated the SELinux kernel copyright notices to reference the FSF's website address and not their postal address.

Audit

There were no audit fixes or improvements in Linux v4.16.

Linux 4.15 Released

Linux v4.15 was released on Sunday, January 28th; this is a quick summary of the SELinux and audit changes.

SELinux

  • A new SELinux object class was added, "bpf", to control access to the extended Berkeley Packet Filter (eBPF) virtual machine in the Linux Kernel, which can be accessed via the bpf(2) system call. The new object class is defined with the following permissions:

    bpf:{ map_create map_read map_write prog_load prog_run }
    The "map_create" permission controls the ability to create new eBPF maps, where the "map_read" and "map_write" permissions control read and write access to eBPF maps. The "prog_load" permission controls the ability to load an eBPF program into the kernel and the "prog_run" permission controls the ability execute a previously loaded eBPF program. Like most system resources, eBPF maps and programs inherit their labels from the originating SELinux domain.

  • The file extended attribute permission code was changed to use the common capability code for any non-SELinux extended attributes. This fixed a problem with file capabilities in user namespaces.

  • Converted the SELinux internal hash table implementation to use kmem_cache, reducing runtime memory consumption by a small amount.

  • Various small code cleanups to remove build warnings and unneeded variable assignments.

Audit

  • Due to a large number of PATH records created by tracefs and debugfs, a new mechanism was introduced which allows the filtering of PATH records based on the filesystem type. Richard Guy Briggs provides a good explanation of this new feature in the patch's description:

    From: Richard Guy Briggs

    audit: filter PATH records keyed on filesystem magic

    Tracefs or debugfs were causing hundreds to thousands of PATH records to be associated with the init_module and finit_module SYSCALL records on a few modules when the following rule was in place for startup:
      -a always,exit -F arch=x86_64 -S init_module -F key=mod-load

    Provide a method to ignore these large number of PATH records from overwhelming the logs if they are not of interest. Introduce a new filter list "AUDIT_FILTER_FS", with a new field type AUDIT_FSTYPE, which keys off the filesystem 4-octet hexadecimal magic identifier to filter specific filesystem PATH records.

    An example rule would look like:
      -a never,filesystem -F fstype=0x74726163 -F key=ignore_tracefs
      -a never,filesystem -F fstype=0x64626720 -F key=ignore_debugfs

    Arguably the better way to address this issue is to disable tracefs and debugfs on boot from production systems.

  • Fixed a bug where the kernel was not initializing the audit subsystem early enough at boot and was missing audit records generated by PID 1 (init/systemd/etc.). In addition to fixing this particular bug there were a number of small code cleanups and general improvements to the audit initialization code.

  • Fixed a bug where the kernel did not recognize when the audit daemon was attempting to shut down the kernel/auditd connection by sending an AUDIT_SET message with a PID 0 value.

  • The audit file and directory write syscall filters were updated with additional syscalls.

  • An old and unused function, audit_log_secctx(), was removed.

UPDATE: Added mention of the SELinux "bpf" object class that was added in Linux v4.15.