Releases: google/nsjail
Releases · google/nsjail
nsjail-3.6
- Added support for embedding the pasta userspace networking binary directly into nsjail (memfd)
- Improved mount handling so both legacy and new mount APIs more reliably enforce read-only remounts
- Fixed seccomp policy assembly so repeated seccomp_string entries are combined and compiled together
- Expanded built-in test coverage for chroot/read-write mount behavior
nsjail-3.5
BROKEN - seccomp_string(s) from config files are not additive, leading to weaker seccomp policies
- Based on the latest Kafel version (support for conditional system calls, updated system call arguments, and m68k support)
- Support for userland networking via pasta/passt
- Experimental support for the new Linux mount API (fsopen/fspick/fsmount)
- More examples in configs/
- Support for configurations in JSON format
nsjail-3.4
- Improved cgroups2 support
- Improved cgroups2 + docker interoperability
- New configs: hexchat, telegram
- Better support for
clone3 - New signals displayed:
SIGPWR - Support for nvim+.clangd
- Improved .clang-format rules
- Print help to stdout if
-h | --helpwas used
nsjail-3.3
- Build fix: Unset LDFLAGS for kafel
- Setup cgroup.subtree_control controllers when necessary in cgroupsv2
nsjail-3.2
- switch to C++14
- various example configs improvements
- using atomics in signal handlers
- improve debug logging
- use only CPUs from current affinity set
- add option to forward fatal signals to the child
nsjail-3.1
- config proto fields remunerated
- various compilation/build error fixes
- process is killed in listen mode once tcp connection is closed
- Support for newly added Linux capabilities (CAP_BPF, CAP_PERFMON, CAP_CHECKPOINT_RESTORE)
- Added global connection limit for listen mode
- Added support for rlimit_mlock, rlimit_rtpr, rlimit_msgq
- Added switch_root option useful for embedded systems that use rootfs
- Fix setting CPU CFS limit
- Allow mount options to contain colons
- Added support for setting cgroup memory.memsw.limit_in_bytes
- Added option to disable TSC
nsjail-3.0
- the TCP proxy mode a socketpair proxy now
- fixes for some configs/ (e.g. for xchat and for znc)
- fixes to the Dockerfile and to the dockerpush.yml
- new clone option recognized (CLONE_NEWPID)
- fixed max_conns_per_ip
- clarification of units for cgroups_mem_max
nsjail-2.9
- even more C++-isms (e.g. RETURN_ON_FAILURE)
- improved EINTR handling
- improved configs for some tools
- changed default RLIMIT_AS to 4GiB
- rudimentary support for cgroups2
- added option to ignore rlimits
- fixed setcwd() w/o CLONE_NEWNS
nsjail-2.8
- even more C++-isms
- clearer main process loop
- refactored cgroup setting code
- ability to specify noexec/nodev/nosuid in mounts
- updated kafel
- added --macvlan_vs_ma option
- better configs/
- changed behavior of --env - empty var means passing it from parent
nsjail-2.7
- More C++'isms across the code
- Removed 'tmpfs_size', '-m none:dest:tmpfs:size=....' can be used for that
- Added support for SECCOMP_FILTER_FLAG_LOG
- Save and restore console state before/after running the subprocesses
- Make use of newer kafel version
- '--iface_own' can be used to put some interface into a jail
- Updated some configs/ (e.g. for Firefox)
- '-s' can be used to specify symlinks via the cmd-line