summaryrefslogtreecommitdiffhomepage
path: root/src/nxt_cgroup.c (follow)
AgeCommit message (Collapse)AuthorFilesLines
2024-06-18fs: Rename nxt_fs_mkdir_all() => nxt_fs_mkdir_p()Alejandro Colomar1-1/+1
"all" is too generic of an attribute to be meaningful. In the context of mkdir(), "parents" is used for this meaning, as in mkdir -p, so it should be more straightforward to readers. Tested-by: Andy Postnikov <apostnikov@gmail.com> Tested-by: Andrew Clayton <a.clayton@nginx.com> Reviewed-by: Andrew Clayton <a.clayton@nginx.com> Signed-off-by: Alejandro Colomar <alx@kernel.org>
2022-12-10Isolation: added core cgroup infrastructure.Andrew Clayton1-0/+174
Firstly, this is not to be confused with CLONE_NEWCGROUP which unit already supports and is related to namespaces. To re-cap, namespaces allow processes to have different views of various parts of the system such as filesystem mounts, networking, hostname etc. Whereas cgroup[0] is a Linux kernel facility for collecting a bunch of processes together to perform some task on the group as a whole, for example to implement resource limits. There are two parts to cgroup, the core part of organising processes into a hierarchy and the controllers which are responsible for enforcing resource limits etc. There are currently two versions of the cgroup sub-system, the original cgroup and a version 2[1] introduced in 3.16 (August 2014) and marked stable in 4.5 (March 2016). This commit supports the cgroup V2 API and implements the ability to place applications into their own cgroup on a per-application basis. You can put them each into their own cgroup or you can group some together. The ability to set resource limits can easily be added in future. The initial use case of this would be to aid in observability of unit applications which becomes much easier if you can just monitor them on a per cgroup basis. One thing to note about cgroup, is that unlike namespaces which are controlled via system calls such as clone(2) and unshare(2), cgroups are setup and controlled through the cgroupfs pseudo-filesystem. cgroup is Linux only and this support will only be enabled if configure finds the cgroup2 filesystem mount, e.g cgroup2 on /sys/fs/cgroup type cgroup2 (rw,nosuid,nodev,noexec,relatime,seclabel,nsdelegate,memory_recursiveprot) The cgroups are removed on shutdown or as required on reconfiguration. This commit just adds the basic infrastructure for using cgroups within unit. Subsequent commits will wire up this support. It supports creating cgroups relative to the main cgroup root and also below the cgroup of the main unit process. [0]: <https://man7.org/linux/man-pages/man7/cgroups.7.html> [1]: <https://www.kernel.org/doc/html/latest/admin-guide/cgroup-v2.html> Cc: Alejandro Colomar <alx@nginx.com> Signed-off-by: Andrew Clayton <a.clayton@nginx.com>