diff --git a/.clippy.toml b/.clippy.toml new file mode 100644 index 00000000..cc94ec53 --- /dev/null +++ b/.clippy.toml @@ -0,0 +1 @@ +upper-case-acronyms-aggressive = true diff --git a/.github/ISSUE_TEMPLATE/bug_report.md b/.github/ISSUE_TEMPLATE/bug_report.md new file mode 100644 index 00000000..8d472dcd --- /dev/null +++ b/.github/ISSUE_TEMPLATE/bug_report.md @@ -0,0 +1,17 @@ +--- +name: Bug report +about: Create a report to help us improve +title: '' +labels: 'bug, needs-review' +assignees: '' + +--- + +**Describe the bug** +A clear and concise description of what the bug is. + +**Expected behavior** +A clear and concise description of what you expected to happen. + +**Additional context** +Add any other context about the problem here. diff --git a/.github/ISSUE_TEMPLATE/enhancement-request.md b/.github/ISSUE_TEMPLATE/enhancement-request.md new file mode 100644 index 00000000..16075976 --- /dev/null +++ b/.github/ISSUE_TEMPLATE/enhancement-request.md @@ -0,0 +1,20 @@ +--- +name: Enhancement request +about: Suggest an improvement to an existing feature +title: '' +labels: enhancement, needs-review +assignees: '' + +--- + +**Which feature do you think can be improved?** + +Specify the feature you think could be made better. + +**How can it be improved?** + +Describe how specifically you think it could be improved. + +**Additional Information** + +Anything else to add? diff --git a/.github/ISSUE_TEMPLATE/feature_request.md b/.github/ISSUE_TEMPLATE/feature_request.md new file mode 100644 index 00000000..a8cea795 --- /dev/null +++ b/.github/ISSUE_TEMPLATE/feature_request.md @@ -0,0 +1,20 @@ +--- +name: Feature request +about: Suggest an idea for this project +title: '' +labels: 'feature, needs-review' +assignees: '' + +--- + +**Is your feature request related to a problem? Please describe.** +A clear and concise description of what the problem is. Ex. I'm always frustrated when [...] + +**Describe the solution you'd like** +A clear and concise description of what you want to happen. + +**Describe alternatives you've considered** +A clear and concise description of any alternative solutions or features you've considered. + +**Additional context** +Add any other context or screenshots about the feature request here. diff --git a/.github/workflows/PR-wip-checks.yaml b/.github/workflows/PR-wip-checks.yaml new file mode 100644 index 00000000..16f8167b --- /dev/null +++ b/.github/workflows/PR-wip-checks.yaml @@ -0,0 +1,21 @@ +name: Pull request WIP checks +on: + pull_request: + types: + - opened + - synchronize + - reopened + - edited + - labeled + - unlabeled + +jobs: + pr_wip_check: + runs-on: ubuntu-latest + name: WIP Check + steps: + - name: WIP Check + uses: tim-actions/wip-check@1c2a1ca6c110026b3e2297bb2ef39e1747b5a755 + with: + labels: '["do-not-merge", "wip", "rfc"]' + keywords: '["WIP", "wip", "RFC", "rfc", "dnm", "DNM", "do-not-merge"]' diff --git a/.github/workflows/bvt.yaml b/.github/workflows/bvt.yaml new file mode 100644 index 00000000..b2bb741c --- /dev/null +++ b/.github/workflows/bvt.yaml @@ -0,0 +1,34 @@ +name: BVT +on: [pull_request] +jobs: + build: + name: Build + runs-on: ubuntu-latest + steps: + - name: Checkout + uses: actions/checkout@v2 + - name: Build + run: make debug + + fmt: + name: Format Check + runs-on: ubuntu-latest + steps: + - uses: actions/checkout@v2 + - run: rustup component add rustfmt + - run: make fmt + clippy: + name: Clippy Check + runs-on: ubuntu-latest + steps: + - uses: actions/checkout@v2 + - run: rustup component add clippy + - run: make clippy + test: + name: Run Unit Test + runs-on: ubuntu-latest + steps: + - uses: actions/checkout@v2 + - run: make test + + diff --git a/.github/workflows/commit-message-check.yaml b/.github/workflows/commit-message-check.yaml new file mode 100644 index 00000000..4d1c57ef --- /dev/null +++ b/.github/workflows/commit-message-check.yaml @@ -0,0 +1,53 @@ +name: Commit Message Check +on: + pull_request: + types: + - opened + - reopened + - synchronize + +env: + error_msg: |+ + See the document below for help on formatting commits for the project. + + https://github.com/kata-containers/community/blob/master/CONTRIBUTING.md#patch-forma + +jobs: + commit-message-check: + runs-on: ubuntu-latest + name: Commit Message Check + steps: + - name: Get PR Commits + id: 'get-pr-commits' + uses: tim-actions/get-pr-commits@v1.0.0 + with: + token: ${{ secrets.GITHUB_TOKEN }} + + - name: DCO Check + uses: tim-actions/dco@2fd0504dc0d27b33f542867c300c60840c6dcb20 + with: + commits: ${{ steps.get-pr-commits.outputs.commits }} + + - name: Commit Body Missing Check + if: ${{ success() || failure() }} + uses: tim-actions/commit-body-check@v1.0.2 + with: + commits: ${{ steps.get-pr-commits.outputs.commits }} + + - name: Check Subject Line Length + if: ${{ success() || failure() }} + uses: tim-actions/commit-message-checker-with-regex@v0.3.1 + with: + commits: ${{ steps.get-pr-commits.outputs.commits }} + pattern: '^.{0,75}(\n.*)*$' + error: 'Subject too long (max 75)' + post_error: ${{ env.error_msg }} + + - name: Check Body Line Length + if: ${{ success() || failure() }} + uses: tim-actions/commit-message-checker-with-regex@v0.3.1 + with: + commits: ${{ steps.get-pr-commits.outputs.commits }} + pattern: '^.+(\n.{0,72})*$|^.+\n\s*[^a-zA-Z\s\n]|^.+\n\S+$' + error: 'Body line too long (max 72)' + post_error: ${{ env.error_msg }} diff --git a/.travis.yml b/.travis.yml deleted file mode 100644 index 20eb418b..00000000 --- a/.travis.yml +++ /dev/null @@ -1,11 +0,0 @@ -language: rust -rust: - - stable - - beta - - nightly -matrix: - allow_failures: - - rust: nightly - fast_finish: true -script: - - cargo build --verbose --all diff --git a/Cargo.toml b/Cargo.toml index 524e0063..86d211ba 100644 --- a/Cargo.toml +++ b/Cargo.toml @@ -1,17 +1,21 @@ [package] -name = "cgroups" +name = "cgroups-rs" description = "Native Rust crate for managing control groups on Linux" -repository = "https://github.com/levex/cgroups-rs" +repository = "https://github.com/kata-containers/cgroups-rs" keywords = ["linux", "cgroup", "containers", "isolation"] categories = ["os", "api-bindings", "os::unix-apis"] license = "MIT OR Apache-2.0" -version = "0.1.1-alpha.0" -authors = ["Levente Kurusa ", "Sam Wilson "] +version = "0.2.5" +authors = ["The Kata Containers community ", "Levente Kurusa ", "Sam Wilson "] edition = "2018" +homepage = "https://github.com/kata-containers/cgroups-rs" +readme = "README.md" [dependencies] log = "0.4" +regex = "1.1" +nix = "0.20.0" +libc = "0.2" [dev-dependencies] -nix = "0.11.0" -libc = "0.2.43" +libc = "0.2.76" diff --git a/Makefile b/Makefile new file mode 100644 index 00000000..89fe2a57 --- /dev/null +++ b/Makefile @@ -0,0 +1,37 @@ +all: debug fmt test + +# +# Build +# + +.PHONY: debug +debug: + RUSTFLAGS="--deny warnings" cargo build + +.PHONY: release +release: + cargo build --release + +.PHONY: build +build: debug + +# +# Tests and linters +# + +.PHONY: test +test: + cargo test -- --color always --nocapture + +.PHONY: check +check: fmt clippy + + +.PHONY: fmt +fmt: + cargo fmt --all -- --check + +.PHONY: clippy +clippy: + cargo clippy --all-targets --all-features -- -D warnings + diff --git a/README.md b/README.md index da0754e0..41c65399 100644 --- a/README.md +++ b/README.md @@ -1,32 +1,36 @@ -# cgroups-rs ![Build](https://travis-ci.org/levex/cgroups-rs.svg?branch=master) +# cgroups-rs ![Build](https://travis-ci.org/kata-containers/cgroups-rs.svg?branch=master) Native Rust library for managing control groups under Linux -Right now the crate only support the original, V1 hierarchy, however support -is planned for the Unified hierarchy. +Both v1 and v2 of cgroups are supported. # Examples ## Create a control group using the builder pattern ``` rust -// Acquire a handle for the V1 cgroup hierarchy. -let hier = ::hierarchies::V1::new(); + + +use cgroups_rs::*; +use cgroups_rs::cgroup_builder::*; + +// Acquire a handle for the cgroup hierarchy. +let hier = cgroups_rs::hierarchies::auto(); // Use the builder pattern (see the documentation to create the control group) // // This creates a control group named "example" in the V1 hierarchy. -let cg: Cgroup = CgroupBuilder::new("example", &v1) - .cpu() - .shares(85) - .done() - .build(); + let cg: Cgroup = CgroupBuilder::new("example") + .cpu() + .shares(85) + .done() + .build(hier); // Now `cg` is a control group that gets 85% of the CPU time in relative to // other control groups. // Get a handle to the CPU controller. -let cpus: &CpuController = cg.controller_of().unwrap(); -cpus.add_task(1234u64); +let cpus: &cgroups_rs::cpu::CpuController = cg.controller_of().unwrap(); +cpus.add_task(&CgroupPid::from(1234u64)); // [...] diff --git a/src/blkio.rs b/src/blkio.rs index fa55f64e..a399697c 100644 --- a/src/blkio.rs +++ b/src/blkio.rs @@ -1,14 +1,20 @@ +// Copyright (c) 2018 Levente Kurusa +// Copyright (c) 2020 And Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! This module contains the implementation of the `blkio` cgroup subsystem. //! //! See the Kernel's documentation for more information about this subsystem, found at: //! [Documentation/cgroup-v1/blkio-controller.txt](https://www.kernel.org/doc/Documentation/cgroup-v1/blkio-controller.txt) -use std::fs::File; -use std::io::{Read, Write}; +use std::io::Write; use std::path::PathBuf; -use crate::error::*; use crate::error::ErrorKind::*; +use crate::error::*; +use crate::{read_string_from, read_u64_from}; use crate::{ BlkIoResources, ControllIdentifier, ControllerInternal, Controllers, Resources, Subsystem, }; @@ -21,6 +27,7 @@ use crate::{ pub struct BlkIoController { base: PathBuf, path: PathBuf, + v2: bool, } #[derive(Eq, PartialEq, Debug)] @@ -53,16 +60,38 @@ pub struct IoService { pub total: u64, } +#[derive(Eq, PartialEq, Debug)] +/// Per-device activity from the control group. +/// Only for cgroup v2 +pub struct IoStat { + /// The major number of the device. + pub major: i16, + /// The minor number of the device. + pub minor: i16, + /// How many bytes were read from the device. + pub rbytes: u64, + /// How many bytes were written to the device. + pub wbytes: u64, + /// How many iops were read from the device. + pub rios: u64, + /// How many iops were written to the device. + pub wios: u64, + /// How many discard bytes were read from the device. + pub dbytes: u64, + /// How many discard iops were written to the device. + pub dios: u64, +} + fn parse_io_service(s: String) -> Result> { s.lines() - .filter(|x| x.split_whitespace().collect::>().len() == 3) + .filter(|x| x.split_whitespace().count() == 3) .map(|x| { let mut spl = x.split_whitespace(); - (spl.nth(0).unwrap(), spl.nth(0).unwrap(), spl.nth(0).unwrap()) + (spl.next().unwrap(), spl.next().unwrap(), spl.next().unwrap()) }) .map(|(a, b, c)| { - let mut spl = a.split(":"); - (spl.nth(0).unwrap(), spl.nth(0).unwrap(), b, c) + let mut spl = a.split(':'); + (spl.next().unwrap(), spl.next().unwrap(), b, c) }) .collect::>() .chunks(5) @@ -94,9 +123,41 @@ fn parse_io_service(s: String) -> Result> { }) } +fn get_value(s: &str) -> String { + let arr = s.split(':').collect::>(); + if arr.len() != 2 { + return "0".to_string(); + } + arr[1].to_string() +} + +fn parse_io_stat(s: String) -> Vec { + // line: + // 8:0 rbytes=180224 wbytes=0 rios=3 wios=0 dbytes=0 dios=0 + s.lines() + .filter(|x| x.split_whitespace().count() == 7) + .map(|x| { + let arr = x.split_whitespace().collect::>(); + let device = arr[0].split(':').collect::>(); + let (major, minor) = (device[0], device[1]); + + IoStat { + major: major.parse::().unwrap(), + minor: minor.parse::().unwrap(), + rbytes: get_value(arr[1]).parse::().unwrap(), + wbytes: get_value(arr[2]).parse::().unwrap(), + rios: get_value(arr[3]).parse::().unwrap(), + wios: get_value(arr[4]).parse::().unwrap(), + dbytes: get_value(arr[5]).parse::().unwrap(), + dios: get_value(arr[6]).parse::().unwrap(), + } + }) + .collect::>() +} + fn parse_io_service_total(s: String) -> Result { s.lines() - .filter(|x| x.split_whitespace().collect::>().len() == 2) + .filter(|x| x.split_whitespace().count() == 2) .fold(Err(Error::new(ParseError)), |_, x| { match x.split_whitespace().collect::>().as_slice() { ["Total", val] => val.parse::().map_err(|_| Error::new(ParseError)), @@ -133,15 +194,15 @@ fn parse_blkio_data(s: String) -> Result> { }); if err.is_err() { - return Err(Error::new(ParseError)); + Err(Error::new(ParseError)) } else { - return Ok(res); + Ok(res) } } /// Current state and statistics about how throttled are the block devices when accessed from the /// controller's control group. -#[derive(Debug)] +#[derive(Default, Debug)] pub struct BlkIoThrottle { /// Statistics about the bytes transferred between the block devices by the tasks in this /// control group. @@ -176,7 +237,7 @@ pub struct BlkIoThrottle { } /// Statistics and state of the block devices. -#[derive(Debug)] +#[derive(Default, Debug)] pub struct BlkIo { /// The number of BIOS requests merged into I/O requests by the control group's tasks. pub io_merged: Vec, @@ -253,6 +314,9 @@ pub struct BlkIo { pub weight: u64, /// Same as `weight`, but per-block-device. pub weight_device: Vec, + + /// IoStat for cgroup v2 + pub io_stat: Vec, } impl ControllerInternal for BlkIoController { @@ -269,34 +333,44 @@ impl ControllerInternal for BlkIoController { &self.base } + fn is_v2(&self) -> bool { + self.v2 + } + fn apply(&self, res: &Resources) -> Result<()> { // get the resources that apply to this controller let res: &BlkIoResources = &res.blkio; - if res.update_values { - let _ = self.set_weight(res.weight as u64); - let _ = self.set_leaf_weight(res.leaf_weight as u64); + if let Some(weight) = res.weight { + let _ = self.set_weight(weight as u64); + } + if let Some(leaf_weight) = res.leaf_weight { + let _ = self.set_leaf_weight(leaf_weight as u64); + } - for dev in &res.weight_device { - let _ = self.set_weight_for_device(dev.major, dev.minor, dev.weight as u64); - let _ = self.set_leaf_weight_for_device(dev.major, dev.minor, dev.leaf_weight as u64); + for dev in &res.weight_device { + if let Some(weight) = dev.weight { + let _ = self.set_weight_for_device(dev.major, dev.minor, weight as u64); } - - for dev in &res.throttle_read_bps_device { - let _ = self.throttle_read_bps_for_device(dev.major, dev.minor, dev.rate); + if let Some(leaf_weight) = dev.leaf_weight { + let _ = self.set_leaf_weight_for_device(dev.major, dev.minor, leaf_weight as u64); } + } - for dev in &res.throttle_write_bps_device { - let _ = self.throttle_write_bps_for_device(dev.major, dev.minor, dev.rate); - } + for dev in &res.throttle_read_bps_device { + let _ = self.throttle_read_bps_for_device(dev.major, dev.minor, dev.rate); + } - for dev in &res.throttle_read_iops_device { - let _ = self.throttle_read_iops_for_device(dev.major, dev.minor, dev.rate); - } + for dev in &res.throttle_write_bps_device { + let _ = self.throttle_write_bps_for_device(dev.major, dev.minor, dev.rate); + } - for dev in &res.throttle_write_iops_device { - let _ = self.throttle_write_iops_for_device(dev.major, dev.minor, dev.rate); - } + for dev in &res.throttle_read_iops_device { + let _ = self.throttle_read_iops_for_device(dev.major, dev.minor, dev.rate); + } + + for dev in &res.throttle_write_iops_device { + let _ = self.throttle_write_iops_for_device(dev.major, dev.minor, dev.rate); } Ok(()) @@ -316,264 +390,263 @@ impl<'a> From<&'a Subsystem> for &'a BlkIoController { Subsystem::BlkIo(c) => c, _ => { assert_eq!(1, 0); - ::std::mem::uninitialized() + let v = std::mem::MaybeUninit::uninit(); + v.assume_init() } } } } } -fn read_string_from(mut file: File) -> Result { - let mut string = String::new(); - match file.read_to_string(&mut string) { - Ok(_) => Ok(string.trim().to_string()), - Err(e) => Err(Error::with_cause(ReadFailed, e)), - } -} - -fn read_u64_from(mut file: File) -> Result { - let mut string = String::new(); - match file.read_to_string(&mut string) { - Ok(_) => string.trim().parse().map_err(|e| Error::with_cause(ParseError, e)), - Err(e) => Err(Error::with_cause(ReadFailed, e)), - } -} - impl BlkIoController { - /// Constructs a new `BlkIoController` with `oroot` serving as the root of the control group. - pub fn new(oroot: PathBuf) -> Self { - let mut root = oroot; - root.push(Self::controller_type().to_string()); + /// Constructs a new `BlkIoController` with `root` serving as the root of the control group. + pub fn new(root: PathBuf, v2: bool) -> Self { Self { base: root.clone(), path: root, + v2, + } + } + + fn blkio_v2(&self) -> BlkIo { + BlkIo { + io_stat: self + .open_path("io.stat", false) + .and_then(read_string_from) + .map(parse_io_stat) + .unwrap_or_default(), + ..Default::default() } } /// Gathers statistics about and reports the state of the block devices used by the control /// group's tasks. pub fn blkio(&self) -> BlkIo { + if self.v2 { + return self.blkio_v2(); + } BlkIo { io_merged: self .open_path("blkio.io_merged", false) .and_then(read_string_from) .and_then(parse_io_service) - .unwrap_or(Vec::new()), + .unwrap_or_default(), io_merged_total: self .open_path("blkio.io_merged", false) .and_then(read_string_from) .and_then(parse_io_service_total) - .unwrap_or(0), + .unwrap_or_default(), io_merged_recursive: self .open_path("blkio.io_merged_recursive", false) .and_then(read_string_from) .and_then(parse_io_service) - .unwrap_or(Vec::new()), + .unwrap_or_default(), io_merged_recursive_total: self .open_path("blkio.io_merged_recursive", false) .and_then(read_string_from) .and_then(parse_io_service_total) - .unwrap_or(0), + .unwrap_or_default(), io_queued: self .open_path("blkio.io_queued", false) .and_then(read_string_from) .and_then(parse_io_service) - .unwrap_or(Vec::new()), + .unwrap_or_default(), io_queued_total: self .open_path("blkio.io_queued", false) .and_then(read_string_from) .and_then(parse_io_service_total) - .unwrap_or(0), + .unwrap_or_default(), io_queued_recursive: self .open_path("blkio.io_queued_recursive", false) .and_then(read_string_from) .and_then(parse_io_service) - .unwrap_or(Vec::new()), + .unwrap_or_default(), io_queued_recursive_total: self .open_path("blkio.io_queued_recursive", false) .and_then(read_string_from) .and_then(parse_io_service_total) - .unwrap_or(0), + .unwrap_or_default(), io_service_bytes: self .open_path("blkio.io_service_bytes", false) .and_then(read_string_from) .and_then(parse_io_service) - .unwrap_or(Vec::new()), + .unwrap_or_default(), io_service_bytes_total: self .open_path("blkio.io_service_bytes", false) .and_then(read_string_from) .and_then(parse_io_service_total) - .unwrap_or(0), + .unwrap_or_default(), io_service_bytes_recursive: self .open_path("blkio.io_service_bytes_recursive", false) .and_then(read_string_from) .and_then(parse_io_service) - .unwrap_or(Vec::new()), + .unwrap_or_default(), io_service_bytes_recursive_total: self .open_path("blkio.io_service_bytes_recursive", false) .and_then(read_string_from) .and_then(parse_io_service_total) - .unwrap_or(0), + .unwrap_or_default(), io_serviced: self .open_path("blkio.io_serviced", false) .and_then(read_string_from) .and_then(parse_io_service) - .unwrap_or(Vec::new()), + .unwrap_or_default(), io_serviced_total: self .open_path("blkio.io_serviced", false) .and_then(read_string_from) .and_then(parse_io_service_total) - .unwrap_or(0), + .unwrap_or_default(), io_serviced_recursive: self .open_path("blkio.io_serviced_recursive", false) .and_then(read_string_from) .and_then(parse_io_service) - .unwrap_or(Vec::new()), + .unwrap_or_default(), io_serviced_recursive_total: self .open_path("blkio.io_serviced_recursive", false) .and_then(read_string_from) .and_then(parse_io_service_total) - .unwrap_or(0), + .unwrap_or_default(), io_service_time: self .open_path("blkio.io_service_time", false) .and_then(read_string_from) .and_then(parse_io_service) - .unwrap_or(Vec::new()), + .unwrap_or_default(), io_service_time_total: self .open_path("blkio.io_service_time", false) .and_then(read_string_from) .and_then(parse_io_service_total) - .unwrap_or(0), + .unwrap_or_default(), io_service_time_recursive: self .open_path("blkio.io_service_time_recursive", false) .and_then(read_string_from) .and_then(parse_io_service) - .unwrap_or(Vec::new()), + .unwrap_or_default(), io_service_time_recursive_total: self .open_path("blkio.io_service_time_recursive", false) .and_then(read_string_from) .and_then(parse_io_service_total) - .unwrap_or(0), + .unwrap_or_default(), io_wait_time: self .open_path("blkio.io_wait_time", false) .and_then(read_string_from) .and_then(parse_io_service) - .unwrap_or(Vec::new()), + .unwrap_or_default(), io_wait_time_total: self .open_path("blkio.io_wait_time", false) .and_then(read_string_from) .and_then(parse_io_service_total) - .unwrap_or(0), + .unwrap_or_default(), io_wait_time_recursive: self .open_path("blkio.io_wait_time_recursive", false) .and_then(read_string_from) .and_then(parse_io_service) - .unwrap_or(Vec::new()), + .unwrap_or_default(), io_wait_time_recursive_total: self .open_path("blkio.io_wait_time_recursive", false) .and_then(read_string_from) .and_then(parse_io_service_total) - .unwrap_or(0), + .unwrap_or_default(), leaf_weight: self .open_path("blkio.leaf_weight", false) - .and_then(|file| read_u64_from(file)) + .and_then(read_u64_from) .unwrap_or(0u64), leaf_weight_device: self .open_path("blkio.leaf_weight_device", false) .and_then(read_string_from) .and_then(parse_blkio_data) - .unwrap_or(Vec::new()), + .unwrap_or_default(), sectors: self .open_path("blkio.sectors", false) .and_then(read_string_from) .and_then(parse_blkio_data) - .unwrap_or(Vec::new()), + .unwrap_or_default(), sectors_recursive: self .open_path("blkio.sectors_recursive", false) .and_then(read_string_from) .and_then(parse_blkio_data) - .unwrap_or(Vec::new()), + .unwrap_or_default(), throttle: BlkIoThrottle { io_service_bytes: self .open_path("blkio.throttle.io_service_bytes", false) .and_then(read_string_from) .and_then(parse_io_service) - .unwrap_or(Vec::new()), + .unwrap_or_default(), io_service_bytes_total: self .open_path("blkio.throttle.io_service_bytes", false) .and_then(read_string_from) .and_then(parse_io_service_total) - .unwrap_or(0), + .unwrap_or_default(), io_service_bytes_recursive: self .open_path("blkio.throttle.io_service_bytes_recursive", false) .and_then(read_string_from) .and_then(parse_io_service) - .unwrap_or(Vec::new()), + .unwrap_or_default(), io_service_bytes_recursive_total: self .open_path("blkio.throttle.io_service_bytes_recursive", false) .and_then(read_string_from) .and_then(parse_io_service_total) - .unwrap_or(0), + .unwrap_or_default(), io_serviced: self .open_path("blkio.throttle.io_serviced", false) .and_then(read_string_from) .and_then(parse_io_service) - .unwrap_or(Vec::new()), + .unwrap_or_default(), io_serviced_total: self .open_path("blkio.throttle.io_serviced", false) .and_then(read_string_from) .and_then(parse_io_service_total) - .unwrap_or(0), + .unwrap_or_default(), io_serviced_recursive: self .open_path("blkio.throttle.io_serviced_recursive", false) .and_then(read_string_from) .and_then(parse_io_service) - .unwrap_or(Vec::new()), + .unwrap_or_default(), io_serviced_recursive_total: self .open_path("blkio.throttle.io_serviced_recursive", false) .and_then(read_string_from) .and_then(parse_io_service_total) - .unwrap_or(0), + .unwrap_or_default(), read_bps_device: self .open_path("blkio.throttle.read_bps_device", false) .and_then(read_string_from) .and_then(parse_blkio_data) - .unwrap_or(Vec::new()), + .unwrap_or_default(), read_iops_device: self .open_path("blkio.throttle.read_iops_device", false) .and_then(read_string_from) .and_then(parse_blkio_data) - .unwrap_or(Vec::new()), + .unwrap_or_default(), write_bps_device: self .open_path("blkio.throttle.write_bps_device", false) .and_then(read_string_from) .and_then(parse_blkio_data) - .unwrap_or(Vec::new()), + .unwrap_or_default(), write_iops_device: self .open_path("blkio.throttle.write_iops_device", false) .and_then(read_string_from) .and_then(parse_blkio_data) - .unwrap_or(Vec::new()), + .unwrap_or_default(), }, time: self .open_path("blkio.time", false) .and_then(read_string_from) .and_then(parse_blkio_data) - .unwrap_or(Vec::new()), + .unwrap_or_default(), time_recursive: self .open_path("blkio.time_recursive", false) .and_then(read_string_from) .and_then(parse_blkio_data) - .unwrap_or(Vec::new()), + .unwrap_or_default(), weight: self .open_path("blkio.weight", false) - .and_then(|file| read_u64_from(file)) + .and_then(read_u64_from) .unwrap_or(0u64), weight_device: self .open_path("blkio.weight_device", false) .and_then(read_string_from) .and_then(parse_blkio_data) - .unwrap_or(Vec::new()), + .unwrap_or_default(), + io_stat: Vec::new(), } } @@ -588,12 +661,7 @@ impl BlkIoController { } /// Same as `set_leaf_weight()`, but settable per each block device. - pub fn set_leaf_weight_for_device( - &self, - major: u64, - minor: u64, - weight: u64, - ) -> Result<()> { + pub fn set_leaf_weight_for_device(&self, major: u64, minor: u64, weight: u64) -> Result<()> { self.open_path("blkio.leaf_weight_device", true) .and_then(|mut file| { file.write_all(format!("{}:{} {}", major, minor, weight).as_ref()) @@ -612,84 +680,89 @@ impl BlkIoController { /// Throttle the bytes per second rate of read operation affecting the block device /// `major:minor` to `bps`. - pub fn throttle_read_bps_for_device( - &self, - major: u64, - minor: u64, - bps: u64, - ) -> Result<()> { - self.open_path("blkio.throttle.read_bps_device", true) - .and_then(|mut file| { - file.write_all(format!("{}:{} {}", major, minor, bps).to_string().as_ref()) - .map_err(|e| Error::with_cause(WriteFailed, e)) - }) + pub fn throttle_read_bps_for_device(&self, major: u64, minor: u64, bps: u64) -> Result<()> { + let mut file = "blkio.throttle.read_bps_device"; + let mut content = format!("{}:{} {}", major, minor, bps); + if self.v2 { + file = "io.max"; + content = format!("{}:{} rbps={}", major, minor, bps); + } + self.open_path(file, true).and_then(|mut file| { + file.write_all(content.as_ref()) + .map_err(|e| Error::with_cause(WriteFailed, e)) + }) } /// Throttle the I/O operations per second rate of read operation affecting the block device /// `major:minor` to `bps`. - pub fn throttle_read_iops_for_device( - &self, - major: u64, - minor: u64, - iops: u64, - ) -> Result<()> { - self.open_path("blkio.throttle.read_iops_device", true) - .and_then(|mut file| { - file.write_all(format!("{}:{} {}", major, minor, iops).to_string().as_ref()) - .map_err(|e| Error::with_cause(WriteFailed, e)) - }) + pub fn throttle_read_iops_for_device(&self, major: u64, minor: u64, iops: u64) -> Result<()> { + let mut file = "blkio.throttle.read_iops_device"; + let mut content = format!("{}:{} {}", major, minor, iops); + if self.v2 { + file = "io.max"; + content = format!("{}:{} riops={}", major, minor, iops); + } + self.open_path(file, true).and_then(|mut file| { + file.write_all(content.as_ref()) + .map_err(|e| Error::with_cause(WriteFailed, e)) + }) } /// Throttle the bytes per second rate of write operation affecting the block device /// `major:minor` to `bps`. - pub fn throttle_write_bps_for_device( - &self, - major: u64, - minor: u64, - bps: u64, - ) -> Result<()> { - self.open_path("blkio.throttle.write_bps_device", true) - .and_then(|mut file| { - file.write_all(format!("{}:{} {}", major, minor, bps).to_string().as_ref()) - .map_err(|e| Error::with_cause(WriteFailed, e)) - }) + pub fn throttle_write_bps_for_device(&self, major: u64, minor: u64, bps: u64) -> Result<()> { + let mut file = "blkio.throttle.write_bps_device"; + let mut content = format!("{}:{} {}", major, minor, bps); + if self.v2 { + file = "io.max"; + content = format!("{}:{} wbps={}", major, minor, bps); + } + self.open_path(file, true).and_then(|mut file| { + file.write_all(content.as_ref()) + .map_err(|e| Error::with_cause(WriteFailed, e)) + }) } /// Throttle the I/O operations per second rate of write operation affecting the block device /// `major:minor` to `bps`. - pub fn throttle_write_iops_for_device( - &self, - major: u64, - minor: u64, - iops: u64, - ) -> Result<()> { - self.open_path("blkio.throttle.write_iops_device", true) - .and_then(|mut file| { - file.write_all(format!("{}:{} {}", major, minor, iops).to_string().as_ref()) - .map_err(|e| Error::with_cause(WriteFailed, e)) - }) + pub fn throttle_write_iops_for_device(&self, major: u64, minor: u64, iops: u64) -> Result<()> { + let mut file = "blkio.throttle.write_iops_device"; + let mut content = format!("{}:{} {}", major, minor, iops); + if self.v2 { + file = "io.max"; + content = format!("{}:{} wiops={}", major, minor, iops); + } + self.open_path(file, true).and_then(|mut file| { + file.write_all(content.as_ref()) + .map_err(|e| Error::with_cause(WriteFailed, e)) + }) } /// Set the weight of the control group's tasks. pub fn set_weight(&self, w: u64) -> Result<()> { - self.open_path("blkio.weight", true) - .and_then(|mut file| { - file.write_all(w.to_string().as_ref()) - .map_err(|e| Error::with_cause(WriteFailed, e)) - }) + // Attation: may not find in high kernel version. + let mut file = "blkio.weight"; + if self.v2 { + file = "io.bfq.weight"; + } + self.open_path(file, true).and_then(|mut file| { + file.write_all(w.to_string().as_ref()) + .map_err(|e| Error::with_cause(WriteFailed, e)) + }) } /// Same as `set_weight()`, but settable per each block device. - pub fn set_weight_for_device( - &self, - major: u64, - minor: u64, - weight: u64, - ) -> Result<()> { - self.open_path("blkio.weight_device", true) - .and_then(|mut file| { - file.write_all(format!("{}:{} {}", major, minor, weight).as_ref()) - .map_err(|e| Error::with_cause(WriteFailed, e)) - }) + pub fn set_weight_for_device(&self, major: u64, minor: u64, weight: u64) -> Result<()> { + let mut file = "blkio.weight_device"; + if self.v2 { + // Attation: there is no weight for device in runc + // https://github.com/opencontainers/runc/blob/46be7b612e2533c494e6a251111de46d8e286ed5/libcontainer/cgroups/fs2/io.go#L30 + // may depends on IO schedulers https://wiki.ubuntu.com/Kernel/Reference/IOSchedulers + file = "io.bfq.weight"; + } + self.open_path(file, true).and_then(|mut file| { + file.write_all(format!("{}:{} {}", major, minor, weight).as_ref()) + .map_err(|e| Error::with_cause(WriteFailed, e)) + }) } } @@ -755,10 +828,7 @@ Total 61823067136 #[test] fn test_parse_io_service_total() { let ok = parse_io_service_total(TEST_VALUE.to_string()).unwrap(); - assert_eq!( - ok, - 61823067136 - ); + assert_eq!(ok, 61823067136); } #[test] @@ -806,10 +876,7 @@ Total 61823067136 ] ); let err = parse_io_service(TEST_WRONG_VALUE.to_string()).unwrap_err(); - assert_eq!( - err.kind(), - &ErrorKind::ParseError, - ); + assert_eq!(err.kind(), &ErrorKind::ParseError,); } #[test] diff --git a/src/cgroup.rs b/src/cgroup.rs index 140a4f4a..c28f0226 100644 --- a/src/cgroup.rs +++ b/src/cgroup.rs @@ -1,11 +1,20 @@ +// Copyright (c) 2018 Levente Kurusa +// Copyright (c) 2020 Ant Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! This module handles cgroup operations. Start here! +use crate::error::ErrorKind::*; use crate::error::*; use crate::{CgroupPid, ControllIdentifier, Controller, Hierarchy, Resources, Subsystem}; +use std::collections::HashMap; use std::convert::From; -use std::path::Path; +use std::fs; +use std::path::{Path, PathBuf}; /// A control group is the central structure to this crate. /// @@ -19,42 +28,81 @@ use std::path::Path; /// > specialized behaviour. /// /// This crate is an attempt at providing a Rust-native way of managing these cgroups. -pub struct Cgroup<'b> { +#[derive(Debug)] +pub struct Cgroup { /// The list of subsystems that control this cgroup subsystems: Vec, /// The hierarchy. - hier: &'b Hierarchy, + hier: Box, + path: String, +} + +impl Clone for Cgroup { + fn clone(&self) -> Self { + Cgroup { + subsystems: self.subsystems.clone(), + path: self.path.clone(), + hier: crate::hierarchies::auto(), + } + } } -impl<'b> Cgroup<'b> { +impl Default for Cgroup { + fn default() -> Self { + Cgroup { + subsystems: Vec::new(), + hier: crate::hierarchies::auto(), + path: "".to_string(), + } + } +} + +impl Cgroup { /// Create this control group. fn create(&self) { - for subsystem in &self.subsystems { - subsystem.to_controller().create(); + if self.hier.v2() { + let _ret = create_v2_cgroup(self.hier.root(), &self.path); + } else { + for subsystem in &self.subsystems { + subsystem.to_controller().create(); + } } } + pub fn v2(&self) -> bool { + self.hier.v2() + } + /// Create a new control group in the hierarchy `hier`, with name `path`. /// /// Returns a handle to the control group that can be used to manipulate it. - /// - /// Note that if the handle goes out of scope and is dropped, the control group is _not_ - /// destroyed. - pub fn new>(hier: &Hierarchy, path: P) -> Cgroup { + pub fn new>(hier: Box, path: P) -> Cgroup { let cg = Cgroup::load(hier, path); cg.create(); cg } + /// Create a new control group in the hierarchy `hier`, with name `path` and `relative_paths` + /// + /// Returns a handle to the control group that can be used to manipulate it. + /// + /// Note that this method is only meaningful for cgroup v1, call it is equivalent to call `new` in the v2 mode + pub fn new_with_relative_paths>( + hier: Box, + path: P, + relative_paths: HashMap, + ) -> Cgroup { + let cg = Cgroup::load_with_relative_paths(hier, path, relative_paths); + cg.create(); + cg + } + /// Create a handle for a control group in the hierarchy `hier`, with name `path`. /// /// Returns a handle to the control group (that possibly does not exist until `create()` has /// been called on the cgroup. - /// - /// Note that if the handle goes out of scope and is dropped, the control group is _not_ - /// destroyed. - pub fn load>(hier: &Hierarchy, path: P) -> Cgroup { + pub fn load>(hier: Box, path: P) -> Cgroup { let path = path.as_ref(); let mut subsystems = hier.subsystems(); if path.as_os_str() != "" { @@ -64,12 +112,54 @@ impl<'b> Cgroup<'b> { .collect::>(); } - let cg = Cgroup { - subsystems: subsystems, - hier: hier, - }; + Cgroup { + path: path.to_str().unwrap().to_string(), + subsystems, + hier, + } + } - cg + /// Create a handle for a control group in the hierarchy `hier`, with name `path` and `relative_paths` + /// + /// Returns a handle to the control group (that possibly does not exist until `create()` has + /// been called on the cgroup. + /// + /// Note that this method is only meaningful for cgroup v1, call it is equivalent to call `load` in the v2 mode + pub fn load_with_relative_paths>( + hier: Box, + path: P, + relative_paths: HashMap, + ) -> Cgroup { + // relative_paths only valid for cgroup v1 + if hier.v2() { + return Self::load(hier, path); + } + + let path = path.as_ref(); + let mut subsystems = hier.subsystems(); + if path.as_os_str() != "" { + subsystems = subsystems + .into_iter() + .map(|x| { + let cn = x.controller_name(); + if relative_paths.contains_key(&cn) { + let rp = relative_paths.get(&cn).unwrap(); + let valid_path = rp.trim_start_matches('/').to_string(); + let mut p = PathBuf::from(valid_path); + p.push(path); + x.enter(p.as_ref()) + } else { + x.enter(path) + } + }) + .collect::>(); + } + + Cgroup { + subsystems, + hier, + path: path.to_str().unwrap().to_string(), + } } /// The list of subsystems that this control group supports. @@ -83,8 +173,17 @@ impl<'b> Cgroup<'b> { /// system call will fail if there are any descendants. Thus, one should check whether it was /// actually removed, and remove the descendants first if not. In the future, this behavior /// will change. - pub fn delete(self) { - self.subsystems.into_iter().for_each(|sub| match sub { + pub fn delete(&self) -> Result<()> { + if self.v2() { + if !self.path.is_empty() { + let mut p = self.hier.root(); + p.push(self.path.clone()); + return fs::remove_dir(p).map_err(|e| Error::with_cause(RemoveFailed, e)); + } + return Ok(()); + } + + self.subsystems.iter().try_for_each(|sub| match sub { Subsystem::Pid(pidc) => pidc.delete(), Subsystem::Mem(c) => c.delete(), Subsystem::CpuSet(c) => c.delete(), @@ -98,7 +197,8 @@ impl<'b> Cgroup<'b> { Subsystem::NetPrio(c) => c.delete(), Subsystem::HugeTlb(c) => c.delete(), Subsystem::Rdma(c) => c.delete(), - }); + Subsystem::Systemd(c) => c.delete(), + }) } /// Apply a set of resource limits to the control group. @@ -118,7 +218,7 @@ impl<'b> Cgroup<'b> { /// let cpu: &CpuController = control_group.controller_of() /// .expect("No cpu controller attached!"); /// ``` - pub fn controller_of<'a, T>(self: &'a Self) -> Option<&'a T> + pub fn controller_of<'a, T>(&'a self) -> Option<&'a T> where &'a T: From<&'a Subsystem>, T: Controller + ControllIdentifier, @@ -143,25 +243,135 @@ impl<'b> Cgroup<'b> { /// Attach a task to the control group. pub fn add_task(&self, pid: CgroupPid) -> Result<()> { + if self.v2() { + let subsystems = self.subsystems(); + if !subsystems.is_empty() { + let c = subsystems[0].to_controller(); + c.add_task(&pid) + } else { + Ok(()) + } + } else { + self.subsystems() + .iter() + .try_for_each(|sub| sub.to_controller().add_task(&pid)) + } + } + + /// Attach a task to the control group by thread group id. + pub fn add_task_by_tgid(&self, pid: CgroupPid) -> Result<()> { self.subsystems() .iter() - .try_for_each(|sub| sub.to_controller().add_task(&pid)) + .try_for_each(|sub| sub.to_controller().add_task_by_tgid(&pid)) + } + + /// Set notify_on_release to the control group. + pub fn set_notify_on_release(&self, enable: bool) -> Result<()> { + self.subsystems() + .iter() + .try_for_each(|sub| sub.to_controller().set_notify_on_release(enable)) + } + + /// Set release_agent + pub fn set_release_agent(&self, path: &str) -> Result<()> { + self.hier + .root_control_group() + .subsystems() + .iter() + .try_for_each(|sub| sub.to_controller().set_release_agent(path)) } /// Returns an Iterator that can be used to iterate over the tasks that are currently in the /// control group. pub fn tasks(&self) -> Vec { // Collect the tasks from all subsystems - let mut v = self - .subsystems() - .iter() - .map(|x| x.to_controller().tasks()) - .fold(vec![], |mut acc, mut x| { - acc.append(&mut x); - acc - }); + let mut v = if self.v2() { + let subsystems = self.subsystems(); + if !subsystems.is_empty() { + let c = subsystems[0].to_controller(); + c.tasks() + } else { + vec![] + } + } else { + self.subsystems() + .iter() + .map(|x| x.to_controller().tasks()) + .fold(vec![], |mut acc, mut x| { + acc.append(&mut x); + acc + }) + }; + v.sort(); v.dedup(); v } } + +pub const UNIFIED_MOUNTPOINT: &str = "/sys/fs/cgroup"; + +fn enable_controllers(controllers: &[String], path: &Path) { + let f = path.join("cgroup.subtree_control"); + for c in controllers { + let body = format!("+{}", c); + let _rest = fs::write(f.as_path(), body.as_bytes()); + } +} + +fn supported_controllers() -> Vec { + let p = format!("{}/{}", UNIFIED_MOUNTPOINT, "cgroup.controllers"); + let ret = fs::read_to_string(p.as_str()); + ret.unwrap_or_default() + .split(' ') + .map(|x| x.to_string()) + .collect::>() +} + +fn create_v2_cgroup(root: PathBuf, path: &str) -> Result<()> { + // controler list ["memory", "cpu"] + let controllers = supported_controllers(); + let mut fp = root; + + // enable for root + enable_controllers(&controllers, &fp); + + // path: "a/b/c" + let elements = path.split('/').collect::>(); + let last_index = elements.len() - 1; + for (i, ele) in elements.iter().enumerate() { + // ROOT/a + fp.push(ele); + // create dir, need not check if is a file or directory + if !fp.exists() { + if let Err(e) = std::fs::create_dir(fp.clone()) { + return Err(Error::with_cause(ErrorKind::FsError, e)); + } + } + + if i < last_index { + // enable controllers for substree + enable_controllers(&controllers, &fp); + } + } + + Ok(()) +} + +pub fn get_cgroups_relative_paths() -> Result> { + let mut m = HashMap::new(); + let content = + fs::read_to_string("/proc/self/cgroup").map_err(|e| Error::with_cause(ReadFailed, e))?; + for l in content.lines() { + let fl: Vec<&str> = l.split(':').collect(); + if fl.len() != 3 { + continue; + } + + let keys: Vec<&str> = fl[1].split(',').collect(); + for key in &keys { + m.insert(key.to_string(), fl[2].to_string()); + } + } + Ok(m) +} diff --git a/src/cgroup_builder.rs b/src/cgroup_builder.rs index 31999c88..09eca24e 100644 --- a/src/cgroup_builder.rs +++ b/src/cgroup_builder.rs @@ -1,3 +1,9 @@ +// Copyright (c) 2018 Levente Kurusa +// Copyright (c) 2020 Ant Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! This module allows the user to create a control group using the Builder pattern. //! # Example //! @@ -10,11 +16,11 @@ //! by a call to `build()`. //! //! ```rust,no_run -//! # use cgroups::*; -//! # use cgroups::devices::*; -//! # use cgroups::cgroup_builder::*; -//! let v1 = cgroups::hierarchies::V1::new(); -//! let cgroup: Cgroup = CgroupBuilder::new("hello", &v1) +//! # use cgroups_rs::*; +//! # use cgroups_rs::devices::*; +//! # use cgroups_rs::cgroup_builder::*; +//! let h = cgroups_rs::hierarchies::auto(); +//! let cgroup: Cgroup = CgroupBuilder::new("hello") //! .memory() //! .kernel_memory_limit(1024 * 1024) //! .memory_hard_limit(1024 * 1024) @@ -42,8 +48,8 @@ //! .blkio() //! .weight(123) //! .leaf_weight(99) -//! .weight_device(6, 1, 100, 55) -//! .weight_device(6, 1, 100, 55) +//! .weight_device(6, 1, Some(100), Some(55)) +//! .weight_device(6, 1, Some(100), Some(55)) //! .throttle_iops() //! .read(6, 1, 10) //! .write(11, 1, 100) @@ -51,89 +57,76 @@ //! .read(6, 1, 10) //! .write(11, 1, 100) //! .done() -//! .build(); +//! .build(h); //! ``` -use crate::error::*; -use crate::{pid, BlkIoDeviceResource, BlkIoDeviceThrottleResource, Cgroup, DeviceResource, Hierarchy, HugePageResource, NetworkPriority, Resources}; +use crate::{ + BlkIoDeviceResource, BlkIoDeviceThrottleResource, Cgroup, DeviceResource, Hierarchy, + HugePageResource, MaxValue, NetworkPriority, Resources, +}; macro_rules! gen_setter { ($res:ident, $cont:ident, $func:ident, $name:ident, $ty:ty) => { /// See the similarly named function in the respective controller. pub fn $name(mut self, $name: $ty) -> Self { - self.cgroup.resources.$res.update_values = true; - self.cgroup.resources.$res.$name = $name; + self.cgroup.resources.$res.$name = Some($name); self } - } + }; } /// A control group builder instance -pub struct CgroupBuilder<'a> { +pub struct CgroupBuilder { name: String, - hierarchy: &'a Hierarchy, /// Internal, unsupported field: use the associated builders instead. resources: Resources, } -impl<'a> CgroupBuilder<'a> { +impl CgroupBuilder { /// Start building a control group with the supplied hierarchy and name pair. /// /// Note that this does not actually create the control group until `build()` is called. - pub fn new(name: &'a str, hierarchy: &'a Hierarchy) -> CgroupBuilder<'a> { + pub fn new(name: &str) -> CgroupBuilder { CgroupBuilder { name: name.to_owned(), - hierarchy: hierarchy, resources: Resources::default(), } } /// Builds the memory resources of the control group. - pub fn memory(self) -> MemoryResourceBuilder<'a> { - MemoryResourceBuilder { - cgroup: self, - } + pub fn memory(self) -> MemoryResourceBuilder { + MemoryResourceBuilder { cgroup: self } } /// Builds the pid resources of the control group. - pub fn pid(self) -> PidResourceBuilder<'a> { - PidResourceBuilder { - cgroup: self, - } + pub fn pid(self) -> PidResourceBuilder { + PidResourceBuilder { cgroup: self } } /// Builds the cpu resources of the control group. - pub fn cpu(self) -> CpuResourceBuilder<'a> { - CpuResourceBuilder { - cgroup: self, - } + pub fn cpu(self) -> CpuResourceBuilder { + CpuResourceBuilder { cgroup: self } } /// Builds the devices resources of the control group, disallowing or /// allowing access to certain devices in the system. - pub fn devices(self) -> DeviceResourceBuilder<'a> { - DeviceResourceBuilder { - cgroup: self, - } + pub fn devices(self) -> DeviceResourceBuilder { + DeviceResourceBuilder { cgroup: self } } /// Builds the network resources of the control group, setting class id, or /// various priorities on networking interfaces. - pub fn network(self) -> NetworkResourceBuilder<'a> { - NetworkResourceBuilder { - cgroup: self, - } + pub fn network(self) -> NetworkResourceBuilder { + NetworkResourceBuilder { cgroup: self } } /// Builds the hugepage/hugetlb resources available to the control group. - pub fn hugepages(self) -> HugepagesResourceBuilder<'a> { - HugepagesResourceBuilder { - cgroup: self, - } + pub fn hugepages(self) -> HugepagesResourceBuilder { + HugepagesResourceBuilder { cgroup: self } } /// Builds the block I/O resources available for the control group. - pub fn blkio(self) -> BlkIoResourcesBuilder<'a> { + pub fn blkio(self) -> BlkIoResourcesBuilder { BlkIoResourcesBuilder { cgroup: self, throttling_iops: false, @@ -141,55 +134,82 @@ impl<'a> CgroupBuilder<'a> { } /// Finalize the control group, consuming the builder and creating the control group. - pub fn build(self) -> Cgroup<'a> { - let cg = Cgroup::new(self.hierarchy, self.name); - cg.apply(&self.resources); + pub fn build(self, hier: Box) -> Cgroup { + let cg = Cgroup::new(hier, self.name); + let _ret = cg.apply(&self.resources); cg } } /// A builder that configures the memory controller of a control group. -pub struct MemoryResourceBuilder<'a> { - cgroup: CgroupBuilder<'a>, +pub struct MemoryResourceBuilder { + cgroup: CgroupBuilder, } -impl<'a> MemoryResourceBuilder<'a> { - - gen_setter!(memory, MemController, set_kmem_limit, kernel_memory_limit, u64); - gen_setter!(memory, MemController, set_limit, memory_hard_limit, u64); - gen_setter!(memory, MemController, set_soft_limit, memory_soft_limit, u64); - gen_setter!(memory, MemController, set_tcp_limit, kernel_tcp_memory_limit, u64); - gen_setter!(memory, MemController, set_memswap_limit, memory_swap_limit, u64); +impl MemoryResourceBuilder { + gen_setter!( + memory, + MemController, + set_kmem_limit, + kernel_memory_limit, + i64 + ); + gen_setter!(memory, MemController, set_limit, memory_hard_limit, i64); + gen_setter!( + memory, + MemController, + set_soft_limit, + memory_soft_limit, + i64 + ); + gen_setter!( + memory, + MemController, + set_tcp_limit, + kernel_tcp_memory_limit, + i64 + ); + gen_setter!( + memory, + MemController, + set_memswap_limit, + memory_swap_limit, + i64 + ); gen_setter!(memory, MemController, set_swappiness, swappiness, u64); /// Finish the construction of the memory resources of a control group. - pub fn done(self) -> CgroupBuilder<'a> { + pub fn done(self) -> CgroupBuilder { self.cgroup } } /// A builder that configures the pid controller of a control group. -pub struct PidResourceBuilder<'a> { - cgroup: CgroupBuilder<'a>, +pub struct PidResourceBuilder { + cgroup: CgroupBuilder, } -impl<'a> PidResourceBuilder<'a> { - - gen_setter!(pid, PidController, set_pid_max, maximum_number_of_processes, pid::PidMax); +impl PidResourceBuilder { + gen_setter!( + pid, + PidController, + set_pid_max, + maximum_number_of_processes, + MaxValue + ); /// Finish the construction of the pid resources of a control group. - pub fn done(self) -> CgroupBuilder<'a> { + pub fn done(self) -> CgroupBuilder { self.cgroup } } /// A builder that configures the cpuset & cpu controllers of a control group. -pub struct CpuResourceBuilder<'a> { - cgroup: CgroupBuilder<'a>, +pub struct CpuResourceBuilder { + cgroup: CgroupBuilder, } -impl<'a> CpuResourceBuilder<'a> { - +impl CpuResourceBuilder { gen_setter!(cpu, CpuSetController, set_cpus, cpus, String); gen_setter!(cpu, CpuSetController, set_mems, mems, String); gen_setter!(cpu, CpuController, set_shares, shares, u64); @@ -199,170 +219,172 @@ impl<'a> CpuResourceBuilder<'a> { gen_setter!(cpu, CpuController, set_rt_period, realtime_period, u64); /// Finish the construction of the cpu resources of a control group. - pub fn done(self) -> CgroupBuilder<'a> { + pub fn done(self) -> CgroupBuilder { self.cgroup } } /// A builder that configures the devices controller of a control group. -pub struct DeviceResourceBuilder<'a> { - cgroup: CgroupBuilder<'a>, +pub struct DeviceResourceBuilder { + cgroup: CgroupBuilder, } -impl<'a> DeviceResourceBuilder<'a> { - +impl DeviceResourceBuilder { /// Restrict (or allow) a device to the tasks inside the control group. - pub fn device(mut self, - major: i64, - minor: i64, - devtype: crate::devices::DeviceType, - allow: bool, - access: Vec) - -> DeviceResourceBuilder<'a> { - self.cgroup.resources.devices.update_values = true; + pub fn device( + mut self, + major: i64, + minor: i64, + devtype: crate::devices::DeviceType, + allow: bool, + access: Vec, + ) -> DeviceResourceBuilder { self.cgroup.resources.devices.devices.push(DeviceResource { + allow, + devtype, major, minor, - devtype, - allow, - access + access, }); self } /// Finish the construction of the devices resources of a control group. - pub fn done(self) -> CgroupBuilder<'a> { + pub fn done(self) -> CgroupBuilder { self.cgroup } } /// A builder that configures the net_cls & net_prio controllers of a control group. -pub struct NetworkResourceBuilder<'a> { - cgroup: CgroupBuilder<'a>, +pub struct NetworkResourceBuilder { + cgroup: CgroupBuilder, } -impl<'a> NetworkResourceBuilder<'a> { - +impl NetworkResourceBuilder { gen_setter!(network, NetclsController, set_class, class_id, u64); /// Set the priority of the tasks when operating on a networking device defined by `name` to be /// `priority`. - pub fn priority(mut self, name: String, priority: u64) - -> NetworkResourceBuilder<'a> { - self.cgroup.resources.network.update_values = true; - self.cgroup.resources.network.priorities.push(NetworkPriority { - name, - priority, - }); + pub fn priority(mut self, name: String, priority: u64) -> NetworkResourceBuilder { + self.cgroup + .resources + .network + .priorities + .push(NetworkPriority { name, priority }); self } /// Finish the construction of the network resources of a control group. - pub fn done(self) -> CgroupBuilder<'a> { + pub fn done(self) -> CgroupBuilder { self.cgroup } } /// A builder that configures the hugepages controller of a control group. -pub struct HugepagesResourceBuilder<'a> { - cgroup: CgroupBuilder<'a>, +pub struct HugepagesResourceBuilder { + cgroup: CgroupBuilder, } -impl<'a> HugepagesResourceBuilder<'a> { - +impl HugepagesResourceBuilder { /// Limit the usage of certain hugepages (determined by `size`) to be at most `limit` bytes. - pub fn limit(mut self, size: String, limit: u64) - -> HugepagesResourceBuilder<'a> { - self.cgroup.resources.hugepages.update_values = true; - self.cgroup.resources.hugepages.limits.push(HugePageResource { - size, - limit, - }); + pub fn limit(mut self, size: String, limit: u64) -> HugepagesResourceBuilder { + self.cgroup + .resources + .hugepages + .limits + .push(HugePageResource { size, limit }); self } /// Finish the construction of the network resources of a control group. - pub fn done(self) -> CgroupBuilder<'a> { + pub fn done(self) -> CgroupBuilder { self.cgroup } } /// A builder that configures the blkio controller of a control group. -pub struct BlkIoResourcesBuilder<'a> { - cgroup: CgroupBuilder<'a>, +pub struct BlkIoResourcesBuilder { + cgroup: CgroupBuilder, throttling_iops: bool, } -impl<'a> BlkIoResourcesBuilder<'a> { - +impl BlkIoResourcesBuilder { gen_setter!(blkio, BlkIoController, set_weight, weight, u16); gen_setter!(blkio, BlkIoController, set_leaf_weight, leaf_weight, u16); /// Set the weight of a certain device. - pub fn weight_device(mut self, - major: u64, - minor: u64, - weight: u16, - leaf_weight: u16) - -> BlkIoResourcesBuilder<'a> { - self.cgroup.resources.blkio.update_values = true; - self.cgroup.resources.blkio.weight_device.push(BlkIoDeviceResource { - major, - minor, - weight, - leaf_weight, - }); + pub fn weight_device( + mut self, + major: u64, + minor: u64, + weight: Option, + leaf_weight: Option, + ) -> BlkIoResourcesBuilder { + self.cgroup + .resources + .blkio + .weight_device + .push(BlkIoDeviceResource { + major, + minor, + weight, + leaf_weight, + }); self } /// Start configuring the I/O operations per second metric. - pub fn throttle_iops(mut self) -> BlkIoResourcesBuilder<'a> { + pub fn throttle_iops(mut self) -> BlkIoResourcesBuilder { self.throttling_iops = true; self } /// Start configuring the bytes per second metric. - pub fn throttle_bps(mut self) -> BlkIoResourcesBuilder<'a> { + pub fn throttle_bps(mut self) -> BlkIoResourcesBuilder { self.throttling_iops = false; self } /// Limit the read rate of the current metric for a certain device. - pub fn read(mut self, major: u64, minor: u64, rate: u64) - -> BlkIoResourcesBuilder<'a> { - self.cgroup.resources.blkio.update_values = true; - let throttle = BlkIoDeviceThrottleResource { - major, - minor, - rate, - }; + pub fn read(mut self, major: u64, minor: u64, rate: u64) -> BlkIoResourcesBuilder { + let throttle = BlkIoDeviceThrottleResource { major, minor, rate }; if self.throttling_iops { - self.cgroup.resources.blkio.throttle_read_iops_device.push(throttle); + self.cgroup + .resources + .blkio + .throttle_read_iops_device + .push(throttle); } else { - self.cgroup.resources.blkio.throttle_read_bps_device.push(throttle); + self.cgroup + .resources + .blkio + .throttle_read_bps_device + .push(throttle); } self } /// Limit the write rate of the current metric for a certain device. - pub fn write(mut self, major: u64, minor: u64, rate: u64) - -> BlkIoResourcesBuilder<'a> { - self.cgroup.resources.blkio.update_values = true; - let throttle = BlkIoDeviceThrottleResource { - major, - minor, - rate, - }; + pub fn write(mut self, major: u64, minor: u64, rate: u64) -> BlkIoResourcesBuilder { + let throttle = BlkIoDeviceThrottleResource { major, minor, rate }; if self.throttling_iops { - self.cgroup.resources.blkio.throttle_write_iops_device.push(throttle); + self.cgroup + .resources + .blkio + .throttle_write_iops_device + .push(throttle); } else { - self.cgroup.resources.blkio.throttle_write_bps_device.push(throttle); + self.cgroup + .resources + .blkio + .throttle_write_bps_device + .push(throttle); } self } /// Finish the construction of the blkio resources of a control group. - pub fn done(self) -> CgroupBuilder<'a> { + pub fn done(self) -> CgroupBuilder { self.cgroup } } diff --git a/src/cpu.rs b/src/cpu.rs index 3e43ea8e..0c7edd90 100644 --- a/src/cpu.rs +++ b/src/cpu.rs @@ -1,3 +1,9 @@ +// Copyright (c) 2018 Levente Kurusa +// Copyright (c) 2020 Ant Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! This module contains the implementation of the `cpu` cgroup subsystem. //! //! See the Kernel's documentation for more information about this subsystem, found at: @@ -7,11 +13,13 @@ use std::fs::File; use std::io::{Read, Write}; use std::path::PathBuf; -use crate::error::*; use crate::error::ErrorKind::*; +use crate::error::*; +use crate::{parse_max_value, read_i64_from, read_u64_from}; use crate::{ - ControllIdentifier, ControllerInternal, Controllers, CpuResources, Resources, Subsystem, + ControllIdentifier, ControllerInternal, Controllers, CpuResources, CustomizedAttribute, + MaxValue, Resources, Subsystem, }; /// A controller that allows controlling the `cpu` subsystem of a Cgroup. @@ -23,6 +31,7 @@ use crate::{ pub struct CpuController { base: PathBuf, path: PathBuf, + v2: bool, } /// The current state of the control group and its processes. @@ -34,6 +43,13 @@ pub struct Cpu { pub stat: String, } +/// The current state of the control group and its processes. +#[derive(Debug)] +struct CfsQuotaAndPeriod { + quota: MaxValue, + period: u64, +} + impl ControllerInternal for CpuController { fn control_type(&self) -> Controllers { Controllers::Cpu @@ -51,29 +67,23 @@ impl ControllerInternal for CpuController { &self.base } + fn is_v2(&self) -> bool { + self.v2 + } + fn apply(&self, res: &Resources) -> Result<()> { // get the resources that apply to this controller let res: &CpuResources = &res.cpu; - if res.update_values { - // apply pid_max - let _ = self.set_shares(res.shares); - if self.shares()? != res.shares as u64 { - return Err(Error::new(ErrorKind::Other)); - } + update_and_test!(self, set_shares, res.shares, shares); + update_and_test!(self, set_cfs_period, res.period, cfs_period); + update_and_test!(self, set_cfs_quota, res.quota, cfs_quota); - let _ = self.set_cfs_period(res.period); - if self.cfs_period()? != res.period as u64 { - return Err(Error::new(ErrorKind::Other)); - } + res.attrs.iter().for_each(|(k, v)| { + let _ = self.set(k, v); + }); - let _ = self.set_cfs_quota(res.quota as u64); - if self.cfs_quota()? != res.quota as u64 { - return Err(Error::new(ErrorKind::Other)); - } - - // TODO: rt properties (CONFIG_RT_GROUP_SCHED) are not yet supported - } + // TODO: rt properties (CONFIG_RT_GROUP_SCHED) are not yet supported Ok(()) } @@ -92,29 +102,21 @@ impl<'a> From<&'a Subsystem> for &'a CpuController { Subsystem::Cpu(c) => c, _ => { assert_eq!(1, 0); - ::std::mem::uninitialized() + let v = std::mem::MaybeUninit::uninit(); + v.assume_init() } } } } } -fn read_u64_from(mut file: File) -> Result { - let mut string = String::new(); - match file.read_to_string(&mut string) { - Ok(_) => string.trim().parse().map_err(|e| Error::with_cause(ParseError, e)), - Err(e) => Err(Error::with_cause(ReadFailed, e)), - } -} - impl CpuController { - /// Contructs a new `CpuController` with `oroot` serving as the root of the control group. - pub fn new(oroot: PathBuf) -> Self { - let mut root = oroot; - root.push(Self::controller_type().to_string()); + /// Contructs a new `CpuController` with `root` serving as the root of the control group. + pub fn new(root: PathBuf, v2: bool) -> Self { Self { base: root.clone(), path: root, + v2, } } @@ -130,7 +132,8 @@ impl CpuController { Ok(_) => Ok(s), Err(e) => Err(Error::with_cause(ReadFailed, e)), } - }).unwrap_or("".to_string()), + }) + .unwrap_or_default(), } } @@ -141,7 +144,12 @@ impl CpuController { /// `shares` to `200` ensures that control group `B` receives twice as much as CPU bandwidth. /// (Assuming both `A` and `B` are of the same parent) pub fn set_shares(&self, shares: u64) -> Result<()> { - self.open_path("cpu.shares", true).and_then(|mut file| { + let mut file = "cpu.shares"; + if self.v2 { + file = "cpu.weight"; + } + // NOTE: .CpuShares is not used here. Conversion is the caller's responsibility. + self.open_path(file, true).and_then(|mut file| { file.write_all(shares.to_string().as_ref()) .map_err(|e| Error::with_cause(WriteFailed, e)) }) @@ -150,12 +158,19 @@ impl CpuController { /// Retrieve the CPU bandwidth that this control group (relative to other control groups and /// this control group's parent) can use. pub fn shares(&self) -> Result { - self.open_path("cpu.shares", false).and_then(read_u64_from) + let mut file = "cpu.shares"; + if self.v2 { + file = "cpu.weight"; + } + self.open_path(file, false).and_then(read_u64_from) } /// Specify a period (when using the CFS scheduler) of time in microseconds for how often this /// control group's access to the CPU should be reallocated. pub fn set_cfs_period(&self, us: u64) -> Result<()> { + if self.v2 { + return self.set_cfs_quota_and_period(None, Some(us)); + } self.open_path("cpu.cfs_period_us", true) .and_then(|mut file| { file.write_all(us.to_string().as_ref()) @@ -166,13 +181,22 @@ impl CpuController { /// Retrieve the period of time of how often this cgroup's access to the CPU should be /// reallocated in microseconds. pub fn cfs_period(&self) -> Result { + if self.v2 { + let current_value = self + .open_path("cpu.max", false) + .and_then(parse_cfs_quota_and_period)?; + return Ok(current_value.period); + } self.open_path("cpu.cfs_period_us", false) .and_then(read_u64_from) } /// Specify a quota (when using the CFS scheduler) of time in microseconds for which all tasks /// in this control group can run during one period (see: `set_cfs_period()`). - pub fn set_cfs_quota(&self, us: u64) -> Result<()> { + pub fn set_cfs_quota(&self, us: i64) -> Result<()> { + if self.v2 { + return self.set_cfs_quota_and_period(Some(us), None); + } self.open_path("cpu.cfs_quota_us", true) .and_then(|mut file| { file.write_all(us.to_string().as_ref()) @@ -182,8 +206,98 @@ impl CpuController { /// Retrieve the quota of time for which all tasks in this cgroup can run during one period, in /// microseconds. - pub fn cfs_quota(&self) -> Result { + pub fn cfs_quota(&self) -> Result { + if self.v2 { + let current_value = self + .open_path("cpu.max", false) + .and_then(parse_cfs_quota_and_period)?; + return Ok(current_value.quota.to_i64()); + } + self.open_path("cpu.cfs_quota_us", false) - .and_then(read_u64_from) + .and_then(read_i64_from) + } + + pub fn set_cfs_quota_and_period(&self, quota: Option, period: Option) -> Result<()> { + if !self.v2 { + if let Some(q) = quota { + self.set_cfs_quota(q)?; + } + if let Some(p) = period { + self.set_cfs_period(p)?; + } + return Ok(()); + } + + // https://www.kernel.org/doc/html/latest/admin-guide/cgroup-v2.html + + // cpu.max + // A read-write two value file which exists on non-root cgroups. The default is “max 100000”. + // The maximum bandwidth limit. It’s in the following format: + // $MAX $PERIOD + // which indicates that the group may consume upto $MAX in each $PERIOD duration. + // “max” for $MAX indicates no limit. If only one number is written, $MAX is updated. + + let current_value = self + .open_path("cpu.max", false) + .and_then(parse_cfs_quota_and_period)?; + + let new_quota = if let Some(q) = quota { + if q > 0 { + q.to_string() + } else { + "max".to_string() + } + } else { + current_value.quota.to_string() + }; + + let new_period = if let Some(p) = period { + p.to_string() + } else { + current_value.period.to_string() + }; + + let line = format!("{} {}", new_quota, new_period); + self.open_path("cpu.max", true).and_then(|mut file| { + file.write_all(line.as_ref()) + .map_err(|e| Error::with_cause(WriteFailed, e)) + }) } + + pub fn set_rt_runtime(&self, us: i64) -> Result<()> { + self.open_path("cpu.rt_runtime_us", true) + .and_then(|mut file| { + file.write_all(us.to_string().as_ref()) + .map_err(|e| Error::with_cause(WriteFailed, e)) + }) + } + + pub fn set_rt_period_us(&self, us: u64) -> Result<()> { + self.open_path("cpu.rt_period_us", true) + .and_then(|mut file| { + file.write_all(us.to_string().as_ref()) + .map_err(|e| Error::with_cause(WriteFailed, e)) + }) + } +} + +impl CustomizedAttribute for CpuController {} + +fn parse_cfs_quota_and_period(mut file: File) -> Result { + let mut content = String::new(); + file.read_to_string(&mut content) + .map_err(|e| Error::with_cause(ReadFailed, e))?; + + let fields = content.trim().split(' ').collect::>(); + if fields.len() != 2 { + return Err(Error::from_string(format!("invaild format: {}", content))); + } + + let quota = parse_max_value(&fields[0].to_string())?; + let period = fields[1] + .parse::() + .map_err(|e| Error::with_cause(ParseError, e))?; + + Ok(CfsQuotaAndPeriod { quota, period }) } diff --git a/src/cpuacct.rs b/src/cpuacct.rs index ec3b967f..7928e45a 100644 --- a/src/cpuacct.rs +++ b/src/cpuacct.rs @@ -1,14 +1,19 @@ +// Copyright (c) 2018 Levente Kurusa +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! This module contains the implementation of the `cpuacct` cgroup subsystem. //! //! See the Kernel's documentation for more information about this subsystem, found at: //! [Documentation/cgroup-v1/cpuacct.txt](https://www.kernel.org/doc/Documentation/cgroup-v1/cpuacct.txt) -use std::fs::File; -use std::io::{Read, Write}; +use std::io::Write; use std::path::PathBuf; -use crate::error::*; use crate::error::ErrorKind::*; +use crate::error::*; +use crate::{read_string_from, read_u64_from}; use crate::{ControllIdentifier, ControllerInternal, Controllers, Resources, Subsystem}; /// A controller that allows controlling the `cpuacct` subsystem of a Cgroup. @@ -84,38 +89,17 @@ impl<'a> From<&'a Subsystem> for &'a CpuAcctController { Subsystem::CpuAcct(c) => c, _ => { assert_eq!(1, 0); - ::std::mem::uninitialized() + let v = std::mem::MaybeUninit::uninit(); + v.assume_init() } } } } } -fn read_u64_from(mut file: File) -> Result { - let mut string = String::new(); - let res = file.read_to_string(&mut string); - match res { - Ok(_) => match string.trim().parse() { - Ok(e) => Ok(e), - Err(e) => Err(Error::with_cause(ParseError, e)), - }, - Err(e) => Err(Error::with_cause(ReadFailed, e)), - } -} - -fn read_string_from(mut file: File) -> Result { - let mut string = String::new(); - match file.read_to_string(&mut string) { - Ok(_) => Ok(string.trim().to_string()), - Err(e) => Err(Error::with_cause(ReadFailed, e)), - } -} - impl CpuAcctController { - /// Contructs a new `CpuAcctController` with `oroot` serving as the root of the control group. - pub fn new(oroot: PathBuf) -> Self { - let mut root = oroot; - root.push(Self::controller_type().to_string()); + /// Contructs a new `CpuAcctController` with `root` serving as the root of the control group. + pub fn new(root: PathBuf) -> Self { Self { base: root.clone(), path: root, @@ -127,42 +111,44 @@ impl CpuAcctController { CpuAcct { stat: self .open_path("cpuacct.stat", false) - .and_then(|file| read_string_from(file)) - .unwrap_or("".to_string()), + .and_then(read_string_from) + .unwrap_or_default(), usage: self .open_path("cpuacct.usage", false) - .and_then(|file| read_u64_from(file)) + .and_then(read_u64_from) .unwrap_or(0), usage_all: self .open_path("cpuacct.usage_all", false) - .and_then(|file| read_string_from(file)) - .unwrap_or("".to_string()), + .and_then(read_string_from) + .unwrap_or_default(), usage_percpu: self .open_path("cpuacct.usage_percpu", false) - .and_then(|file| read_string_from(file)) - .unwrap_or("".to_string()), + .and_then(read_string_from) + .unwrap_or_default(), usage_percpu_sys: self .open_path("cpuacct.usage_percpu_sys", false) - .and_then(|file| read_string_from(file)) - .unwrap_or("".to_string()), + .and_then(read_string_from) + .unwrap_or_default(), usage_percpu_user: self .open_path("cpuacct.usage_percpu_user", false) - .and_then(|file| read_string_from(file)) - .unwrap_or("".to_string()), + .and_then(read_string_from) + .unwrap_or_default(), usage_sys: self .open_path("cpuacct.usage_sys", false) - .and_then(|file| read_u64_from(file)) + .and_then(read_u64_from) .unwrap_or(0), usage_user: self .open_path("cpuacct.usage_user", false) - .and_then(|file| read_u64_from(file)) + .and_then(read_u64_from) .unwrap_or(0), } } /// Reset the statistics the kernel has gathered about the control group. pub fn reset(&self) -> Result<()> { - self.open_path("cpuacct.usage", true) - .and_then(|mut file| file.write_all(b"0").map_err(|e| Error::with_cause(WriteFailed, e))) + self.open_path("cpuacct.usage", true).and_then(|mut file| { + file.write_all(b"0") + .map_err(|e| Error::with_cause(WriteFailed, e)) + }) } } diff --git a/src/cpuset.rs b/src/cpuset.rs index b7b679a7..4bc5b57c 100644 --- a/src/cpuset.rs +++ b/src/cpuset.rs @@ -1,14 +1,22 @@ +// Copyright (c) 2018 Levente Kurusa +// Copyright (c) 2020 Ant Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! This module contains the implementation of the `cpuset` cgroup subsystem. //! //! See the Kernel's documentation for more information about this subsystem, found at: //! [Documentation/cgroup-v1/cpusets.txt](https://www.kernel.org/doc/Documentation/cgroup-v1/cpusets.txt) -use std::fs::File; -use std::io::{Read, Write}; + +use log::*; +use std::io::Write; use std::path::PathBuf; -use crate::error::*; use crate::error::ErrorKind::*; +use crate::error::*; +use crate::{read_string_from, read_u64_from}; use crate::{ ControllIdentifier, ControllerInternal, Controllers, CpuResources, Resources, Subsystem, }; @@ -21,6 +29,7 @@ use crate::{ pub struct CpuSetController { base: PathBuf, path: PathBuf, + v2: bool, } /// The current state of the `cpuset` controller for this control group. @@ -93,17 +102,85 @@ impl ControllerInternal for CpuSetController { &self.base } + fn is_v2(&self) -> bool { + self.v2 + } + fn apply(&self, res: &Resources) -> Result<()> { // get the resources that apply to this controller let res: &CpuResources = &res.cpu; - if res.update_values { - let _ = self.set_cpus(&res.cpus); - let _ = self.set_mems(&res.mems); - } + update!(self, set_cpus, res.cpus.as_ref()); + update!(self, set_mems, res.mems.as_ref()); Ok(()) } + + fn post_create(&self) { + if self.is_v2() { + return; + } + let current = self.get_path(); + + if current != self.get_base() { + match copy_from_parent(current.to_str().unwrap(), "cpuset.cpus") { + Ok(_) => (), + Err(err) => error!("error create_dir for cpuset.cpus {:?}", err), + } + match copy_from_parent(current.to_str().unwrap(), "cpuset.mems") { + Ok(_) => (), + Err(err) => error!("error create_dir for cpuset.mems {:?}", err), + } + } + } +} + +fn find_no_empty_parent(from: &str, file: &str) -> Result<(String, Vec)> { + let mut current_path = ::std::path::Path::new(from).to_path_buf(); + let mut v = vec![]; + + loop { + let current_value = + match ::std::fs::read_to_string(current_path.clone().join(file).to_str().unwrap()) { + Ok(cpus) => String::from(cpus.trim()), + Err(e) => return Err(Error::with_cause(ReadFailed, e)), + }; + + if !current_value.is_empty() { + return Ok((current_value, v)); + } + v.push(current_path.clone()); + + let parent = match current_path.parent() { + Some(p) => p, + None => return Ok(("".to_string(), v)), + }; + + // next loop, find parent + current_path = parent.to_path_buf(); + } +} + +/// copy_from_parent copy the cpuset.cpus and cpuset.mems from the parent +/// directory to the current directory if the file's contents are 0 +fn copy_from_parent(current: &str, file: &str) -> Result<()> { + // find not empty cpus/memes from current directory. + let (value, parents) = find_no_empty_parent(current, file)?; + + if value.is_empty() || parents.is_empty() { + return Ok(()); + } + + for p in parents.iter().rev() { + let mut pb = p.clone(); + pb.push(file); + match ::std::fs::write(pb.to_str().unwrap(), value.as_bytes()) { + Ok(_) => (), + Err(e) => return Err(Error::with_cause(WriteFailed, e)), + } + } + + Ok(()) } impl ControllIdentifier for CpuSetController { @@ -119,44 +196,29 @@ impl<'a> From<&'a Subsystem> for &'a CpuSetController { Subsystem::CpuSet(c) => c, _ => { assert_eq!(1, 0); - ::std::mem::uninitialized() + let v = std::mem::MaybeUninit::uninit(); + v.assume_init() } } } } } -fn read_string_from(mut file: File) -> Result { - let mut string = String::new(); - match file.read_to_string(&mut string) { - Ok(_) => Ok(string.trim().to_string()), - Err(e) => Err(Error::with_cause(ReadFailed, e)), - } -} - -fn read_u64_from(mut file: File) -> Result { - let mut string = String::new(); - match file.read_to_string(&mut string) { - Ok(_) => string.trim().parse().map_err(|e| Error::with_cause(ParseError, e)), - Err(e) => Err(Error::with_cause(ReadFailed, e)), - } -} - /// Parse a string like "1,2,4-5,8" into a list of (start, end) tuples. fn parse_range(s: String) -> Result> { let mut fin = Vec::new(); - if s == "".to_string() { + if s.is_empty() { return Ok(fin); } // first split by commas - let comma_split = s.split(","); + let comma_split = s.split(','); for sp in comma_split { - if sp.contains("-") { + if sp.contains('-') { // this is a true range - let dash_split = sp.split("-").collect::>(); + let dash_split = sp.split('-').collect::>(); if dash_split.len() != 2 { return Err(Error::new(ParseError)); } @@ -180,13 +242,12 @@ fn parse_range(s: String) -> Result> { } impl CpuSetController { - /// Contructs a new `CpuSetController` with `oroot` serving as the root of the control group. - pub fn new(oroot: PathBuf) -> Self { - let mut root = oroot; - root.push(Self::controller_type().to_string()); + /// Contructs a new `CpuSetController` with `root` serving as the root of the control group. + pub fn new(root: PathBuf, v2: bool) -> Self { Self { base: root.clone(), path: root, + v2, } } @@ -196,7 +257,7 @@ impl CpuSetController { CpuSet { cpu_exclusive: { self.open_path("cpuset.cpu_exclusive", false) - .and_then(|file| read_u64_from(file)) + .and_then(read_u64_from) .map(|x| x == 1) .unwrap_or(false) }, @@ -204,19 +265,19 @@ impl CpuSetController { self.open_path("cpuset.cpus", false) .and_then(read_string_from) .and_then(parse_range) - .unwrap_or(Vec::new()) + .unwrap_or_default() }, effective_cpus: { self.open_path("cpuset.effective_cpus", false) .and_then(read_string_from) .and_then(parse_range) - .unwrap_or(Vec::new()) + .unwrap_or_default() }, effective_mems: { self.open_path("cpuset.effective_mems", false) .and_then(read_string_from) .and_then(parse_range) - .unwrap_or(Vec::new()) + .unwrap_or_default() }, mem_exclusive: { self.open_path("cpuset.mem_exclusive", false) @@ -263,7 +324,7 @@ impl CpuSetController { self.open_path("cpuset.mems", false) .and_then(read_string_from) .and_then(parse_range) - .unwrap_or(Vec::new()) + .unwrap_or_default() }, sched_load_balance: { self.open_path("cpuset.sched_load_balance", false) @@ -285,9 +346,11 @@ impl CpuSetController { self.open_path("cpuset.cpu_exclusive", true) .and_then(|mut file| { if b { - file.write_all(b"1").map_err(|e| Error::with_cause(WriteFailed, e)) + file.write_all(b"1") + .map_err(|e| Error::with_cause(WriteFailed, e)) } else { - file.write_all(b"0").map_err(|e| Error::with_cause(WriteFailed, e)) + file.write_all(b"0") + .map_err(|e| Error::with_cause(WriteFailed, e)) } }) } @@ -298,9 +361,11 @@ impl CpuSetController { self.open_path("cpuset.mem_exclusive", true) .and_then(|mut file| { if b { - file.write_all(b"1").map_err(|e| Error::with_cause(WriteFailed, e)) + file.write_all(b"1") + .map_err(|e| Error::with_cause(WriteFailed, e)) } else { - file.write_all(b"0").map_err(|e| Error::with_cause(WriteFailed, e)) + file.write_all(b"0") + .map_err(|e| Error::with_cause(WriteFailed, e)) } }) } @@ -335,9 +400,11 @@ impl CpuSetController { self.open_path("cpuset.mem_hardwall", true) .and_then(|mut file| { if b { - file.write_all(b"1").map_err(|e| Error::with_cause(WriteFailed, e)) + file.write_all(b"1") + .map_err(|e| Error::with_cause(WriteFailed, e)) } else { - file.write_all(b"0").map_err(|e| Error::with_cause(WriteFailed, e)) + file.write_all(b"0") + .map_err(|e| Error::with_cause(WriteFailed, e)) } }) } @@ -348,9 +415,11 @@ impl CpuSetController { self.open_path("cpuset.sched_load_balance", true) .and_then(|mut file| { if b { - file.write_all(b"1").map_err(|e| Error::with_cause(WriteFailed, e)) + file.write_all(b"1") + .map_err(|e| Error::with_cause(WriteFailed, e)) } else { - file.write_all(b"0").map_err(|e| Error::with_cause(WriteFailed, e)) + file.write_all(b"0") + .map_err(|e| Error::with_cause(WriteFailed, e)) } }) } @@ -372,9 +441,11 @@ impl CpuSetController { self.open_path("cpuset.memory_migrate", true) .and_then(|mut file| { if b { - file.write_all(b"1").map_err(|e| Error::with_cause(WriteFailed, e)) + file.write_all(b"1") + .map_err(|e| Error::with_cause(WriteFailed, e)) } else { - file.write_all(b"0").map_err(|e| Error::with_cause(WriteFailed, e)) + file.write_all(b"0") + .map_err(|e| Error::with_cause(WriteFailed, e)) } }) } @@ -385,9 +456,11 @@ impl CpuSetController { self.open_path("cpuset.memory_spread_page", true) .and_then(|mut file| { if b { - file.write_all(b"1").map_err(|e| Error::with_cause(WriteFailed, e)) + file.write_all(b"1") + .map_err(|e| Error::with_cause(WriteFailed, e)) } else { - file.write_all(b"0").map_err(|e| Error::with_cause(WriteFailed, e)) + file.write_all(b"0") + .map_err(|e| Error::with_cause(WriteFailed, e)) } }) } @@ -398,9 +471,11 @@ impl CpuSetController { self.open_path("cpuset.memory_spread_slab", true) .and_then(|mut file| { if b { - file.write_all(b"1").map_err(|e| Error::with_cause(WriteFailed, e)) + file.write_all(b"1") + .map_err(|e| Error::with_cause(WriteFailed, e)) } else { - file.write_all(b"0").map_err(|e| Error::with_cause(WriteFailed, e)) + file.write_all(b"0") + .map_err(|e| Error::with_cause(WriteFailed, e)) } }) } @@ -417,9 +492,11 @@ impl CpuSetController { self.open_path("cpuset.memory_pressure_enabled", true) .and_then(|mut file| { if b { - file.write_all(b"1").map_err(|e| Error::with_cause(WriteFailed, e)) + file.write_all(b"1") + .map_err(|e| Error::with_cause(WriteFailed, e)) } else { - file.write_all(b"0").map_err(|e| Error::with_cause(WriteFailed, e)) + file.write_all(b"0") + .map_err(|e| Error::with_cause(WriteFailed, e)) } }) } diff --git a/src/devices.rs b/src/devices.rs index 90c386d1..1cd1ea8b 100644 --- a/src/devices.rs +++ b/src/devices.rs @@ -1,3 +1,8 @@ +// Copyright (c) 2018 Levente Kurusa +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! This module contains the implementation of the `devices` cgroup subsystem. //! //! See the Kernel's documentation for more information about this subsystem, found at: @@ -7,8 +12,8 @@ use std::path::PathBuf; use log::*; -use crate::error::*; use crate::error::ErrorKind::*; +use crate::error::*; use crate::{ ControllIdentifier, ControllerInternal, Controllers, DeviceResource, DeviceResources, @@ -97,7 +102,7 @@ impl DevicePermissions { /// Checks whether the string is a valid descriptor of DevicePermissions. pub fn is_valid(s: &str) -> bool { - if s == "" { + if s.is_empty() { return false; } for i in s.chars() { @@ -105,7 +110,7 @@ impl DevicePermissions { return false; } } - return true; + true } /// Returns a Vec will all the permissions that a device can have. @@ -118,14 +123,14 @@ impl DevicePermissions { } /// Convert a string into DevicePermissions. + #[allow(clippy::should_implement_trait)] pub fn from_str(s: &str) -> Result> { let mut v = Vec::new(); - if s == "" { + if s.is_empty() { return Ok(v); } for e in s.chars() { - let perm = DevicePermissions::from_char(e) - .ok_or_else(|| Error::new(ParseError))?; + let perm = DevicePermissions::from_char(e).ok_or_else(|| Error::new(ParseError))?; v.push(perm); } @@ -151,13 +156,11 @@ impl ControllerInternal for DevicesController { // get the resources that apply to this controller let res: &DeviceResources = &res.devices; - if res.update_values { - for i in &res.devices { - if i.allow { - let _ = self.allow_device(i.devtype, i.major, i.minor, &i.access); - } else { - let _ = self.deny_device(i.devtype, i.major, i.minor, &i.access); - } + for i in &res.devices { + if i.allow { + let _ = self.allow_device(i.devtype, i.major, i.minor, &i.access); + } else { + let _ = self.deny_device(i.devtype, i.major, i.minor, &i.access); } } @@ -178,7 +181,8 @@ impl<'a> From<&'a Subsystem> for &'a DevicesController { Subsystem::Devices(c) => c, _ => { assert_eq!(1, 0); - ::std::mem::uninitialized() + let v = std::mem::MaybeUninit::uninit(); + v.assume_init() } } } @@ -186,10 +190,8 @@ impl<'a> From<&'a Subsystem> for &'a DevicesController { } impl DevicesController { - /// Constructs a new `DevicesController` with `oroot` serving as the root of the control group. - pub fn new(oroot: PathBuf) -> Self { - let mut root = oroot; - root.push(Self::controller_type().to_string()); + /// Constructs a new `DevicesController` with `root` serving as the root of the control group. + pub fn new(root: PathBuf) -> Self { Self { base: root.clone(), path: root, @@ -205,7 +207,7 @@ impl DevicesController { devtype: DeviceType, major: i64, minor: i64, - perm: &Vec, + perm: &[DevicePermissions], ) -> Result<()> { let perms = perm .iter() @@ -237,7 +239,7 @@ impl DevicesController { devtype: DeviceType, major: i64, minor: i64, - perm: &Vec, + perm: &[DevicePermissions], ) -> Result<()> { let perms = perm .iter() @@ -273,13 +275,13 @@ impl DevicesController { error!("allowed_devices: acc: {:?}, ls: {:?}", acc, ls); Err(Error::new(ParseError)) } else { - let devtype = DeviceType::from_char(ls[0].chars().nth(0)); + let devtype = DeviceType::from_char(ls[0].chars().next()); let mut major = ls[1].parse::(); let mut minor = ls[2].parse::(); - if major.is_err() && ls[1] == "*".to_string() { + if major.is_err() && ls[1] == "*" { major = Ok(-1); } - if minor.is_err() && ls[2] == "*".to_string() { + if minor.is_err() && ls[2] == "*" { minor = Ok(-1); } if devtype.is_none() || major.is_err() || minor.is_err() || !DevicePermissions::is_valid(&ls[3]) { @@ -294,7 +296,7 @@ impl DevicesController { devtype: devtype.unwrap(), major: major.unwrap(), minor: minor.unwrap(), - access: access, + access, }); Ok(acc) } diff --git a/src/error.rs b/src/error.rs index 70eb4eca..cea6a39b 100644 --- a/src/error.rs +++ b/src/error.rs @@ -1,15 +1,27 @@ +// Copyright (c) 2018 Levente Kurusa +// Copyright (c) 2020 Ant Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + use std::error::Error as StdError; use std::fmt; /// The different types of errors that can occur while manipulating control groups. #[derive(Debug, Eq, PartialEq)] pub enum ErrorKind { + FsError, + Common(String), + /// An error occured while writing to a control group file. WriteFailed, /// An error occured while trying to read from a control group file. ReadFailed, + /// An error occured while trying to remove a control group. + RemoveFailed, + /// An error occured while trying to parse a value from a control group file. /// /// In the future, there will be some information attached to this field. @@ -27,6 +39,8 @@ pub enum ErrorKind { /// This crate checks against this and operations will fail with this error. InvalidPath, + InvalidBytesSize, + /// An unknown error has occured. Other, } @@ -34,26 +48,35 @@ pub enum ErrorKind { #[derive(Debug)] pub struct Error { kind: ErrorKind, - cause: Option>, + cause: Option>, } impl fmt::Display for Error { fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { - let msg = match self.kind { - ErrorKind::WriteFailed => "unable to write to a control group file", - ErrorKind::ReadFailed => "unable to read a control group file", - ErrorKind::ParseError => "unable to parse control group file", - ErrorKind::InvalidOperation => "the requested operation is invalid", - ErrorKind::InvalidPath => "the given path is invalid", - ErrorKind::Other => "an unknown error", + let msg = match &self.kind { + ErrorKind::FsError => "fs error".to_string(), + ErrorKind::Common(s) => s.clone(), + ErrorKind::WriteFailed => "unable to write to a control group file".to_string(), + ErrorKind::ReadFailed => "unable to read a control group file".to_string(), + ErrorKind::RemoveFailed => "unable to remove a control group".to_string(), + ErrorKind::ParseError => "unable to parse control group file".to_string(), + ErrorKind::InvalidOperation => "the requested operation is invalid".to_string(), + ErrorKind::InvalidPath => "the given path is invalid".to_string(), + ErrorKind::InvalidBytesSize => "invalid bytes size".to_string(), + ErrorKind::Other => "an unknown error".to_string(), }; - write!(f, "{}", msg) + if let Some(cause) = &self.cause { + write!(f, "{} caused by: {:?}", msg, cause) + } else { + write!(f, "{}", msg) + } } } impl StdError for Error { - fn cause(&self) -> Option<&StdError> { + fn cause(&self) -> Option<&dyn StdError> { + #[allow(clippy::manual_map)] match self.cause { Some(ref x) => Some(&**x), None => None, @@ -62,16 +85,19 @@ impl StdError for Error { } impl Error { - pub(crate) fn new(kind: ErrorKind) -> Self { + pub(crate) fn from_string(s: String) -> Self { Self { - kind, + kind: ErrorKind::Common(s), cause: None, } } + pub(crate) fn new(kind: ErrorKind) -> Self { + Self { kind, cause: None } + } pub(crate) fn with_cause(kind: ErrorKind, cause: E) -> Self where - E: 'static + Send + StdError, + E: 'static + Send + Sync + StdError, { Self { kind, diff --git a/src/events.rs b/src/events.rs new file mode 100644 index 00000000..08c25dee --- /dev/null +++ b/src/events.rs @@ -0,0 +1,87 @@ +// Copyright (c) 2020 Ant Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + +use eventfd::{eventfd, EfdFlags}; +use nix::sys::eventfd; +use std::fs::{self, File}; +use std::io::Read; +use std::os::unix::io::{AsRawFd, FromRawFd}; +use std::path::Path; +use std::sync::mpsc::{self, Receiver}; +use std::thread; + +use crate::error::ErrorKind::*; +use crate::error::*; + +// notify_on_oom returns channel on which you can expect event about OOM, +// if process died without OOM this channel will be closed. +pub fn notify_on_oom_v2(key: &str, dir: &Path) -> Result> { + register_memory_event(key, dir, "memory.oom_control", "") +} + +// notify_on_oom returns channel on which you can expect event about OOM, +// if process died without OOM this channel will be closed. +pub fn notify_on_oom_v1(key: &str, dir: &Path) -> Result> { + register_memory_event(key, dir, "memory.oom_control", "") +} + +// level is one of "low", "medium", or "critical" +pub fn notify_memory_pressure(key: &str, dir: &Path, level: &str) -> Result> { + if level != "low" && level != "medium" && level != "critical" { + return Err(Error::from_string(format!( + "invalid pressure level {}", + level + ))); + } + + register_memory_event(key, dir, "memory.pressure_level", level) +} + +fn register_memory_event( + key: &str, + cg_dir: &Path, + event_name: &str, + arg: &str, +) -> Result> { + let path = cg_dir.join(event_name); + let event_file = File::open(path).map_err(|e| Error::with_cause(ReadFailed, e))?; + + let eventfd = + eventfd(0, EfdFlags::EFD_CLOEXEC).map_err(|e| Error::with_cause(ReadFailed, e))?; + + let event_control_path = cg_dir.join("cgroup.event_control"); + let data; + if arg.is_empty() { + data = format!("{} {}", eventfd, event_file.as_raw_fd()); + } else { + data = format!("{} {} {}", eventfd, event_file.as_raw_fd(), arg); + } + + // write to file and set mode to 0700(FIXME) + fs::write(&event_control_path, data).map_err(|e| Error::with_cause(WriteFailed, e))?; + + let mut eventfd_file = unsafe { File::from_raw_fd(eventfd) }; + + let (sender, receiver) = mpsc::channel(); + let key = key.to_string(); + + thread::spawn(move || { + loop { + let mut buf = [0; 8]; + if eventfd_file.read(&mut buf).is_err() { + return; + } + + // When a cgroup is destroyed, an event is sent to eventfd. + // So if the control path is gone, return instead of notifying. + if !Path::new(&event_control_path).exists() { + return; + } + sender.send(key.clone()).unwrap(); + } + }); + + Ok(receiver) +} diff --git a/src/freezer.rs b/src/freezer.rs index 632ee272..1382b824 100644 --- a/src/freezer.rs +++ b/src/freezer.rs @@ -1,3 +1,9 @@ +// Copyright (c) 2018 Levente Kurusa +// Copyright (c) 2020 Ant Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! This module contains the implementation of the `freezer` cgroup subsystem. //! //! See the Kernel's documentation for more information about this subsystem, found at: @@ -5,8 +11,8 @@ use std::io::{Read, Write}; use std::path::PathBuf; -use crate::error::*; use crate::error::ErrorKind::*; +use crate::error::*; use crate::{ControllIdentifier, ControllerInternal, Controllers, Resources, Subsystem}; @@ -22,6 +28,7 @@ use crate::{ControllIdentifier, ControllerInternal, Controllers, Resources, Subs pub struct FreezerController { base: PathBuf, path: PathBuf, + v2: bool, } /// The current state of the control group @@ -66,7 +73,8 @@ impl<'a> From<&'a Subsystem> for &'a FreezerController { Subsystem::Freezer(c) => c, _ => { assert_eq!(1, 0); - ::std::mem::uninitialized() + let v = std::mem::MaybeUninit::uninit(); + v.assume_init() } } } @@ -74,41 +82,59 @@ impl<'a> From<&'a Subsystem> for &'a FreezerController { } impl FreezerController { - /// Contructs a new `FreezerController` with `oroot` serving as the root of the control group. - pub fn new(oroot: PathBuf) -> Self { - let mut root = oroot; - root.push(Self::controller_type().to_string()); + /// Contructs a new `FreezerController` with `root` serving as the root of the control group. + pub fn new(root: PathBuf, v2: bool) -> Self { Self { base: root.clone(), path: root, + v2, } } /// Freezes the processes in the control group. pub fn freeze(&self) -> Result<()> { - self.open_path("freezer.state", true).and_then(|mut file| { - file.write_all("FROZEN".to_string().as_ref()) + let mut file = "freezer.state"; + let mut content = "FROZEN".to_string(); + if self.v2 { + file = "cgroup.freeze"; + content = "1".to_string(); + } + + self.open_path(file, true).and_then(|mut file| { + file.write_all(content.as_ref()) .map_err(|e| Error::with_cause(WriteFailed, e)) }) } /// Thaws, that is, unfreezes the processes in the control group. pub fn thaw(&self) -> Result<()> { - self.open_path("freezer.state", true).and_then(|mut file| { - file.write_all("THAWED".to_string().as_ref()) + let mut file = "freezer.state"; + let mut content = "THAWED".to_string(); + if self.v2 { + file = "cgroup.freeze"; + content = "0".to_string(); + } + self.open_path(file, true).and_then(|mut file| { + file.write_all(content.as_ref()) .map_err(|e| Error::with_cause(WriteFailed, e)) }) } /// Retrieve the state of processes in the control group. pub fn state(&self) -> Result { - self.open_path("freezer.state", false).and_then(|mut file| { + let mut file = "freezer.state"; + if self.v2 { + file = "cgroup.freeze"; + } + self.open_path(file, false).and_then(|mut file| { let mut s = String::new(); let res = file.read_to_string(&mut s); match res { - Ok(_) => match s.as_ref() { + Ok(_) => match s.trim() { "FROZEN" => Ok(FreezerState::Frozen), "THAWED" => Ok(FreezerState::Thawed), + "1" => Ok(FreezerState::Frozen), + "0" => Ok(FreezerState::Thawed), "FREEZING" => Ok(FreezerState::Freezing), _ => Err(Error::new(ParseError)), }, diff --git a/src/hierarchies.rs b/src/hierarchies.rs index 368251a5..0cd2e0e8 100644 --- a/src/hierarchies.rs +++ b/src/hierarchies.rs @@ -1,14 +1,18 @@ +// Copyright (c) 2018 Levente Kurusa +// Copyright (c) 2020 Ant Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! This module represents the various control group hierarchies the Linux kernel supports. //! //! Currently, we only support the cgroupv1 hierarchy, but in the future we will add support for //! the Unified Hierarchy. +use std::fs; use std::fs::File; -use std::io::BufRead; -use std::io::BufReader; -use std::path::{Path, PathBuf}; - -use log::*; +use std::io::{BufRead, BufReader}; +use std::path::PathBuf; use crate::blkio::BlkIoController; use crate::cpu::CpuController; @@ -23,116 +27,344 @@ use crate::net_prio::NetPrioController; use crate::perf_event::PerfEventController; use crate::pid::PidController; use crate::rdma::RdmaController; +use crate::systemd::SystemdController; use crate::{Controllers, Hierarchy, Subsystem}; use crate::cgroup::Cgroup; +/// Process mounts information. +/// +/// See `proc(5)` for format details. +#[derive(Debug, PartialEq, Eq, Hash, Clone)] +pub struct Mountinfo { + /// Mount pathname relative to the process's root. + pub mount_point: PathBuf, + /// Filesystem type (main type with optional sub-type). + pub fs_type: (String, Option), + /// Superblock options. + pub super_opts: Vec, +} + +pub(crate) fn parse_mountinfo_for_line(line: &str) -> Option { + let s_values: Vec<_> = line.split(" - ").collect(); + if s_values.len() != 2 { + return None; + } + + let s0_values: Vec<_> = s_values[0].trim().split(' ').collect(); + let s1_values: Vec<_> = s_values[1].trim().split(' ').collect(); + if s0_values.len() < 6 || s1_values.len() < 3 { + return None; + } + let mount_point = PathBuf::from(s0_values[4]); + let fs_type_values: Vec<_> = s1_values[0].trim().split('.').collect(); + let fs_type = match fs_type_values.len() { + 1 => (fs_type_values[0].to_string(), None), + 2 => ( + fs_type_values[0].to_string(), + Some(fs_type_values[1].to_string()), + ), + _ => return None, + }; + + let super_opts: Vec = s1_values[2].trim().split(',').map(String::from).collect(); + Some(Mountinfo { + mount_point, + fs_type, + super_opts, + }) +} + +/// Parses the provided mountinfo file. +fn mountinfo_file(file: &mut File) -> Vec { + let mut r = Vec::new(); + for line in BufReader::new(file).lines() { + match line { + Ok(line) => { + if let Some(mi) = parse_mountinfo_for_line(&line) { + if mi.fs_type.0 == "cgroup" { + r.push(mi); + } + } + } + Err(_) => continue, + } + } + r +} + +/// Returns mounts information for the current process. +pub fn mountinfo_self() -> Vec { + match File::open("/proc/self/mountinfo") { + Ok(mut file) => mountinfo_file(&mut file), + Err(_) => vec![], + } +} + /// The standard, original cgroup implementation. Often referred to as "cgroupv1". +#[derive(Debug, Clone)] pub struct V1 { - mount_point: String, + mountinfo: Vec, +} + +#[derive(Debug, Clone)] +pub struct V2 { + root: String, } impl Hierarchy for V1 { + fn v2(&self) -> bool { + false + } + fn subsystems(&self) -> Vec { let mut subs = vec![]; - if self.check_support(Controllers::Pids) { - subs.push(Subsystem::Pid(PidController::new(self.root()))); + + // The cgroup writeback feature requires cooperation between memcgs and blkcgs + // To avoid exceptions, we should add_task for blkcg before memcg(push BlkIo before Mem) + // For more Information: https://www.alibabacloud.com/help/doc-detail/155509.htm + if let Some(root) = self.get_mount_point(Controllers::BlkIo) { + subs.push(Subsystem::BlkIo(BlkIoController::new(root, false))); + } + if let Some(root) = self.get_mount_point(Controllers::Mem) { + subs.push(Subsystem::Mem(MemController::new(root, false))); } - if self.check_support(Controllers::Mem) { - subs.push(Subsystem::Mem(MemController::new(self.root()))); + if let Some(root) = self.get_mount_point(Controllers::Pids) { + subs.push(Subsystem::Pid(PidController::new(root, false))); } - if self.check_support(Controllers::CpuSet) { - subs.push(Subsystem::CpuSet(CpuSetController::new(self.root()))); + if let Some(root) = self.get_mount_point(Controllers::CpuSet) { + subs.push(Subsystem::CpuSet(CpuSetController::new(root, false))); } - if self.check_support(Controllers::CpuAcct) { - subs.push(Subsystem::CpuAcct(CpuAcctController::new(self.root()))); + if let Some(root) = self.get_mount_point(Controllers::CpuAcct) { + subs.push(Subsystem::CpuAcct(CpuAcctController::new(root))); } - if self.check_support(Controllers::Cpu) { - subs.push(Subsystem::Cpu(CpuController::new(self.root()))); + if let Some(root) = self.get_mount_point(Controllers::Cpu) { + subs.push(Subsystem::Cpu(CpuController::new(root, false))); } - if self.check_support(Controllers::Devices) { - subs.push(Subsystem::Devices(DevicesController::new(self.root()))); + if let Some(root) = self.get_mount_point(Controllers::Devices) { + subs.push(Subsystem::Devices(DevicesController::new(root))); } - if self.check_support(Controllers::Freezer) { - subs.push(Subsystem::Freezer(FreezerController::new(self.root()))); + if let Some(root) = self.get_mount_point(Controllers::Freezer) { + subs.push(Subsystem::Freezer(FreezerController::new(root, false))); } - if self.check_support(Controllers::NetCls) { - subs.push(Subsystem::NetCls(NetClsController::new(self.root()))); + if let Some(root) = self.get_mount_point(Controllers::NetCls) { + subs.push(Subsystem::NetCls(NetClsController::new(root))); } - if self.check_support(Controllers::BlkIo) { - subs.push(Subsystem::BlkIo(BlkIoController::new(self.root()))); + if let Some(root) = self.get_mount_point(Controllers::PerfEvent) { + subs.push(Subsystem::PerfEvent(PerfEventController::new(root))); } - if self.check_support(Controllers::PerfEvent) { - subs.push(Subsystem::PerfEvent(PerfEventController::new(self.root()))); + if let Some(root) = self.get_mount_point(Controllers::NetPrio) { + subs.push(Subsystem::NetPrio(NetPrioController::new(root))); } - if self.check_support(Controllers::NetPrio) { - subs.push(Subsystem::NetPrio(NetPrioController::new(self.root()))); + if let Some(root) = self.get_mount_point(Controllers::HugeTlb) { + subs.push(Subsystem::HugeTlb(HugeTlbController::new(root, false))); } - if self.check_support(Controllers::HugeTlb) { - subs.push(Subsystem::HugeTlb(HugeTlbController::new(self.root()))); + if let Some(root) = self.get_mount_point(Controllers::Rdma) { + subs.push(Subsystem::Rdma(RdmaController::new(root))); } - if self.check_support(Controllers::Rdma) { - subs.push(Subsystem::Rdma(RdmaController::new(self.root()))); + if let Some(root) = self.get_mount_point(Controllers::Systemd) { + subs.push(Subsystem::Systemd(SystemdController::new(root, false))); } subs } fn root_control_group(&self) -> Cgroup { - Cgroup::load(self, "".to_string()) + Cgroup::load(auto(), "".to_string()) } - fn check_support(&self, sub: Controllers) -> bool { - let root = self.root().read_dir().unwrap(); - for entry in root { - if let Ok(entry) = entry { - if entry.file_name().into_string().unwrap() == sub.to_string() { - return true; + fn root(&self) -> PathBuf { + self.mountinfo + .iter() + .find_map(|m| { + if m.fs_type.0 == "cgroup" { + return Some(m.mount_point.parent().unwrap()); + } + None + }) + .unwrap() + .to_path_buf() + } +} + +impl Hierarchy for V2 { + fn v2(&self) -> bool { + true + } + + fn subsystems(&self) -> Vec { + let p = format!("{}/{}", UNIFIED_MOUNTPOINT, "cgroup.controllers"); + let ret = fs::read_to_string(p.as_str()); + if ret.is_err() { + return vec![]; + } + + let mut subs = vec![]; + + let controllers = ret.unwrap().trim().to_string(); + let controller_list: Vec<&str> = controllers.split(' ').collect(); + + for s in controller_list { + match s { + "cpu" => { + subs.push(Subsystem::Cpu(CpuController::new(self.root(), true))); + } + "io" => { + subs.push(Subsystem::BlkIo(BlkIoController::new(self.root(), true))); + } + "cpuset" => { + subs.push(Subsystem::CpuSet(CpuSetController::new(self.root(), true))); + } + "memory" => { + subs.push(Subsystem::Mem(MemController::new(self.root(), true))); } + "pids" => { + subs.push(Subsystem::Pid(PidController::new(self.root(), true))); + } + "freezer" => { + subs.push(Subsystem::Freezer(FreezerController::new( + self.root(), + true, + ))); + } + "hugetlb" => { + subs.push(Subsystem::HugeTlb(HugeTlbController::new( + self.root(), + true, + ))); + } + _ => {} } } - return false; + + subs + } + + fn root_control_group(&self) -> Cgroup { + Cgroup::load(auto(), "".to_string()) } fn root(&self) -> PathBuf { - PathBuf::from(self.mount_point.clone()) + PathBuf::from(self.root.clone()) } } impl V1 { /// Finds where control groups are mounted to and returns a hierarchy in which control groups /// can be created. - pub fn new() -> Self { - let mount_point = find_v1_mount().unwrap(); + pub fn new() -> V1 { V1 { - mount_point: mount_point, + mountinfo: mountinfo_self(), } } + + pub fn get_mount_point(&self, controller: Controllers) -> Option { + self.mountinfo.iter().find_map(|m| { + if m.fs_type.0 == "cgroup" && m.super_opts.contains(&controller.to_string()) { + return Some(m.mount_point.clone()); + } + None + }) + } } -fn find_v1_mount() -> Option { - // Open mountinfo so we can get a parseable mount list - let mountinfo_path = Path::new("/proc/self/mountinfo"); +impl Default for V1 { + fn default() -> Self { + Self::new() + } +} - // If /proc isn't mounted, or something else happens, then bail out - if mountinfo_path.exists() == false { - return None; +impl V2 { + /// Finds where control groups are mounted to and returns a hierarchy in which control groups + /// can be created. + pub fn new() -> V2 { + V2 { + root: String::from(UNIFIED_MOUNTPOINT), + } } +} + +impl Default for V2 { + fn default() -> Self { + Self::new() + } +} - let mountinfo_file = File::open(mountinfo_path).unwrap(); - let mountinfo_reader = BufReader::new(&mountinfo_file); - for _line in mountinfo_reader.lines() { - let line = _line.unwrap(); - let mut fields = line.split_whitespace(); - let index = line.find(" - ").unwrap(); - let mut more_fields = line[index + 3..].split_whitespace().collect::>(); - let fstype = more_fields[0]; - if fstype == "tmpfs" && more_fields[2].contains("ro") { - let cgroups_mount = fields.nth(4).unwrap(); - info!("found cgroups at {:?}", cgroups_mount); - return Some(cgroups_mount.to_string()); +pub const UNIFIED_MOUNTPOINT: &str = "/sys/fs/cgroup"; + +#[cfg(all(target_os = "linux", not(target_env = "musl")))] +pub fn is_cgroup2_unified_mode() -> bool { + use nix::sys::statfs; + + let path = std::path::Path::new(UNIFIED_MOUNTPOINT); + let fs_stat = statfs::statfs(path); + if fs_stat.is_err() { + return false; + } + + // FIXME notwork, nix will not compile CGROUP2_SUPER_MAGIC because not(target_env = "musl") + fs_stat.unwrap().filesystem_type() == statfs::CGROUP2_SUPER_MAGIC +} + +pub const INIT_CGROUP_PATHS: &str = "/proc/1/cgroup"; + +#[cfg(all(target_os = "linux", target_env = "musl"))] +pub fn is_cgroup2_unified_mode() -> bool { + let lines = fs::read_to_string(INIT_CGROUP_PATHS); + if lines.is_err() { + return false; + } + + for line in lines.unwrap().lines() { + let fields: Vec<&str> = line.split(':').collect(); + if fields.len() != 3 { + continue; } + if fields[0] != "0" { + return false; + } + } + + true +} + +pub fn auto() -> Box { + if is_cgroup2_unified_mode() { + Box::new(V2::new()) + } else { + Box::new(V1::new()) } +} + +#[cfg(test)] +mod tests { + use super::*; + + #[test] + fn test_parse_mount() { + let mountinfo = vec![ + ("29 26 0:26 / /sys/fs/cgroup/cpuset,cpu,cpuacct rw,nosuid,nodev,noexec,relatime shared:10 - cgroup cgroup rw,cpuset,cpu,cpuacct", + Mountinfo{mount_point: PathBuf::from("/sys/fs/cgroup/cpuset,cpu,cpuacct"), fs_type: ("cgroup".to_string(), None), super_opts: vec![ + "rw".to_string(), + "cpuset".to_string(), + "cpu".to_string(), + "cpuacct".to_string(), + ]}), + ("121 1731 0:42 / /shm rw,nosuid,nodev,noexec,relatime shared:68 master:66 - tmpfs shm rw,size=65536k", + Mountinfo{mount_point: PathBuf::from("/shm"), fs_type: ("tmpfs".to_string(), None), super_opts: vec![ + "rw".to_string(), + "size=65536k".to_string(), + ]}), + ("121 1731 0:42 / /shm rw,nosuid,nodev,noexec,relatime shared:68 master:66 - tmpfs.123 shm rw,size=65536k", + Mountinfo{mount_point: PathBuf::from("/shm"), fs_type: ("tmpfs".to_string(), Some("123".to_string())), super_opts: vec![ + "rw".to_string(), + "size=65536k".to_string(), + ]}), + ]; - None + for mi in mountinfo { + let info = parse_mountinfo_for_line(mi.0).unwrap(); + assert_eq!(info, mi.1) + } + } } diff --git a/src/hugetlb.rs b/src/hugetlb.rs index 5c3b1b73..b6888790 100644 --- a/src/hugetlb.rs +++ b/src/hugetlb.rs @@ -1,17 +1,22 @@ +// Copyright (c) 2018 Levente Kurusa +// Copyright (c) 2020 Ant Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! This module contains the implementation of the `hugetlb` cgroup subsystem. //! //! See the Kernel's documentation for more information about this subsystem, found at: //! [Documentation/cgroup-v1/hugetlb.txt](https://www.kernel.org/doc/Documentation/cgroup-v1/hugetlb.txt) -use std::fs::File; -use std::io::{Read, Write}; +use std::io::Write; use std::path::PathBuf; -use crate::error::*; use crate::error::ErrorKind::*; +use crate::error::*; +use crate::{flat_keyed_to_vec, read_u64_from}; use crate::{ - ControllIdentifier, ControllerInternal, Controllers, HugePageResources, Resources, - Subsystem, + ControllIdentifier, ControllerInternal, Controllers, HugePageResources, Resources, Subsystem, }; /// A controller that allows controlling the `hugetlb` subsystem of a Cgroup. @@ -22,6 +27,8 @@ use crate::{ pub struct HugeTlbController { base: PathBuf, path: PathBuf, + sizes: Vec, + v2: bool, } impl ControllerInternal for HugeTlbController { @@ -38,18 +45,21 @@ impl ControllerInternal for HugeTlbController { &self.base } + fn is_v2(&self) -> bool { + self.v2 + } + fn apply(&self, res: &Resources) -> Result<()> { // get the resources that apply to this controller let res: &HugePageResources = &res.hugepages; - if res.update_values { - for i in &res.limits { - let _ = self.set_limit_in_bytes(&i.size, i.limit); - if self.limit_in_bytes(&i.size)? != i.limit { - return Err(Error::new(Other)); - } + for i in &res.limits { + let _ = self.set_limit_in_bytes(&i.size, i.limit); + if self.limit_in_bytes(&i.size)? != i.limit { + return Err(Error::new(Other)); } } + Ok(()) } } @@ -67,40 +77,59 @@ impl<'a> From<&'a Subsystem> for &'a HugeTlbController { Subsystem::HugeTlb(c) => c, _ => { assert_eq!(1, 0); - ::std::mem::uninitialized() + let v = std::mem::MaybeUninit::uninit(); + v.assume_init() } } } } } -fn read_u64_from(mut file: File) -> Result { - let mut string = String::new(); - match file.read_to_string(&mut string) { - Ok(_) => string.trim().parse().map_err(|e| Error::with_cause(ParseError, e)), - Err(e) => Err(Error::with_cause(ReadFailed, e)), - } -} - impl HugeTlbController { - /// Constructs a new `HugeTlbController` with `oroot` serving as the root of the control group. - pub fn new(oroot: PathBuf) -> Self { - let mut root = oroot; - root.push(Self::controller_type().to_string()); + /// Constructs a new `HugeTlbController` with `root` serving as the root of the control group. + pub fn new(root: PathBuf, v2: bool) -> Self { + let sizes = get_hugepage_sizes().unwrap(); Self { base: root.clone(), path: root, + sizes, + v2, } } /// Whether the system supports `hugetlb_size` hugepages. - pub fn size_supported(&self, _hugetlb_size: &str) -> bool { - // TODO - true + pub fn size_supported(&self, hugetlb_size: &str) -> bool { + for s in &self.sizes { + if s == hugetlb_size { + return true; + } + } + false + } + + pub fn get_sizes(&self) -> Vec { + self.sizes.clone() + } + + fn failcnt_v2(&self, hugetlb_size: &str) -> Result { + self.open_path(&format!("hugetlb.{}.events", hugetlb_size), false) + .and_then(flat_keyed_to_vec) + .and_then(|x| { + if x.is_empty() { + return Err(Error::from_string(format!( + "get empty from hugetlb.{}.events", + hugetlb_size + ))); + } + Ok(x[0].1 as u64) + }) } /// Check how many times has the limit of `hugetlb_size` hugepages been hit. pub fn failcnt(&self, hugetlb_size: &str) -> Result { + if self.v2 { + return self.failcnt_v2(hugetlb_size); + } self.open_path(&format!("hugetlb.{}.failcnt", hugetlb_size), false) .and_then(read_u64_from) } @@ -115,8 +144,11 @@ impl HugeTlbController { /// Get the current usage of memory that is backed by hugepages of a certain size /// (`hugetlb_size`). pub fn usage_in_bytes(&self, hugetlb_size: &str) -> Result { - self.open_path(&format!("hugetlb.{}.usage_in_bytes", hugetlb_size), false) - .and_then(read_u64_from) + let mut file = format!("hugetlb.{}.usage_in_bytes", hugetlb_size); + if self.v2 { + file = format!("hugetlb.{}.current", hugetlb_size); + } + self.open_path(&file, false).and_then(read_u64_from) } /// Get the maximum observed usage of memory that is backed by hugepages of a certain size @@ -125,16 +157,144 @@ impl HugeTlbController { self.open_path( &format!("hugetlb.{}.max_usage_in_bytes", hugetlb_size), false, - ).and_then(read_u64_from) + ) + .and_then(read_u64_from) } /// Set the limit (in bytes) of how much memory can be backed by hugepages of a certain size /// (`hugetlb_size`). pub fn set_limit_in_bytes(&self, hugetlb_size: &str, limit: u64) -> Result<()> { - self.open_path(&format!("hugetlb.{}.limit_in_bytes", hugetlb_size), true) - .and_then(|mut file| { - file.write_all(limit.to_string().as_ref()) - .map_err(|e| Error::with_cause(WriteFailed, e)) - }) + let mut file = format!("hugetlb.{}.limit_in_bytes", hugetlb_size); + if self.v2 { + file = format!("hugetlb.{}.max", hugetlb_size); + } + self.open_path(&file, true).and_then(|mut file| { + file.write_all(limit.to_string().as_ref()) + .map_err(|e| Error::with_cause(WriteFailed, e)) + }) + } +} + +pub const HUGEPAGESIZE_DIR: &str = "/sys/kernel/mm/hugepages"; +use regex::Regex; +use std::collections::HashMap; +use std::fs; + +fn get_hugepage_sizes() -> Result> { + let mut m = Vec::new(); + let dirs = fs::read_dir(HUGEPAGESIZE_DIR); + if dirs.is_err() { + return Ok(m); + } + + for e in dirs.unwrap() { + let entry = e.unwrap(); + let name = entry.file_name().into_string().unwrap(); + let parts: Vec<&str> = name.split('-').collect(); + if parts.len() != 2 { + continue; + } + let bmap = get_binary_size_map(); + let size = parse_size(parts[1], &bmap)?; + let dabbrs = get_decimal_abbrs(); + m.push(custom_size(size as f64, 1024.0, &dabbrs)); + } + + Ok(m) +} + +pub const KB: u128 = 1000; +pub const MB: u128 = 1000 * KB; +pub const GB: u128 = 1000 * MB; +pub const TB: u128 = 1000 * GB; +pub const PB: u128 = 1000 * TB; + +#[allow(non_upper_case_globals)] +pub const KiB: u128 = 1024; +#[allow(non_upper_case_globals)] +pub const MiB: u128 = 1024 * KiB; +#[allow(non_upper_case_globals)] +pub const GiB: u128 = 1024 * MiB; +#[allow(non_upper_case_globals)] +pub const TiB: u128 = 1024 * GiB; +#[allow(non_upper_case_globals)] +pub const PiB: u128 = 1024 * TiB; + +pub fn get_binary_size_map() -> HashMap { + let mut m = HashMap::new(); + m.insert("k".to_string(), KiB); + m.insert("m".to_string(), MiB); + m.insert("g".to_string(), GiB); + m.insert("t".to_string(), TiB); + m.insert("p".to_string(), PiB); + m +} + +pub fn get_decimal_size_map() -> HashMap { + let mut m = HashMap::new(); + m.insert("k".to_string(), KB); + m.insert("m".to_string(), MB); + m.insert("g".to_string(), GB); + m.insert("t".to_string(), TB); + m.insert("p".to_string(), PB); + m +} + +pub fn get_decimal_abbrs() -> Vec { + let m = vec![ + "B".to_string(), + "KB".to_string(), + "MB".to_string(), + "GB".to_string(), + "TB".to_string(), + "PB".to_string(), + "EB".to_string(), + "ZB".to_string(), + "YB".to_string(), + ]; + m +} + +fn parse_size(s: &str, m: &HashMap) -> Result { + let re = Regex::new(r"(?P\d+)(?P[kKmMgGtTpP]?)[bB]?$"); + + if re.is_err() { + return Err(Error::new(InvalidBytesSize)); } + let caps = re.unwrap().captures(s).unwrap(); + + let num = caps.name("num"); + let size: u128 = if let Some(num) = num { + let n = num.as_str().trim().parse::(); + if n.is_err() { + return Err(Error::new(InvalidBytesSize)); + } + n.unwrap() + } else { + return Err(Error::new(InvalidBytesSize)); + }; + + let q = caps.name("mul"); + let mul: u128 = if let Some(q) = q { + let t = m.get(q.as_str()); + if let Some(t) = t { + *t + } else { + return Err(Error::new(InvalidBytesSize)); + } + } else { + return Err(Error::new(InvalidBytesSize)); + }; + + Ok(size * mul) +} + +fn custom_size(mut size: f64, base: f64, m: &[String]) -> String { + let mut i = 0; + while size >= base && i < m.len() - 1 { + size /= base; + i += 1; + } + + format!("{}{}", size, m[i].as_str()) } diff --git a/src/lib.rs b/src/lib.rs index e6ed6459..b8b2385d 100644 --- a/src/lib.rs +++ b/src/lib.rs @@ -1,16 +1,47 @@ +// Copyright (c) 2018 Levente Kurusa +// Copyright (c) 2020 Ant Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + +#![allow(clippy::unnecessary_unwrap)] use log::*; -use std::fs::File; -use std::io::{BufRead, BufReader, Write}; +use std::collections::HashMap; +use std::fmt; +use std::fs::{self, File}; +use std::io::{BufRead, BufReader, Read, Write}; use std::path::{Path, PathBuf}; +use std::str::FromStr; + +macro_rules! update_and_test { + ($self: ident, $set_func:ident, $value:expr, $get_func:ident) => { + if let Some(v) = $value { + $self.$set_func(v)?; + if $self.$get_func()? != v { + return Err(Error::new(Other)); + } + } + }; +} + +macro_rules! update { + ($self: ident, $set_func:ident, $value:expr) => { + if let Some(v) = $value { + let _ = $self.$set_func(v); + } + }; +} pub mod blkio; pub mod cgroup; +pub mod cgroup_builder; pub mod cpu; pub mod cpuacct; pub mod cpuset; pub mod devices; pub mod error; +pub mod events; pub mod freezer; pub mod hierarchies; pub mod hugetlb; @@ -20,13 +51,14 @@ pub mod net_prio; pub mod perf_event; pub mod pid; pub mod rdma; -pub mod cgroup_builder; +pub mod systemd; use crate::blkio::BlkIoController; use crate::cpu::CpuController; use crate::cpuacct::CpuAcctController; use crate::cpuset::CpuSetController; use crate::devices::DevicesController; +use crate::error::ErrorKind::*; use crate::error::*; use crate::freezer::FreezerController; use crate::hugetlb::HugeTlbController; @@ -36,11 +68,13 @@ use crate::net_prio::NetPrioController; use crate::perf_event::PerfEventController; use crate::pid::PidController; use crate::rdma::RdmaController; +use crate::systemd::SystemdController; +#[doc(inline)] pub use crate::cgroup::Cgroup; /// Contains all the subsystems that are available in this crate. -#[derive(Debug)] +#[derive(Debug, Clone)] pub enum Subsystem { /// Controller for the `Pid` subsystem, see `PidController` for more information. Pid(PidController), @@ -68,10 +102,12 @@ pub enum Subsystem { HugeTlb(HugeTlbController), /// Controller for the `Rdma` subsystem, see `RdmaController` for more information. Rdma(RdmaController), + /// Controller for the `Systemd` subsystem, see `SystemdController` for more information. + Systemd(SystemdController), } #[doc(hidden)] -#[derive(Eq, PartialEq, Debug)] +#[derive(Eq, PartialEq, Debug, Clone)] pub enum Controllers { Pids, Mem, @@ -86,24 +122,26 @@ pub enum Controllers { NetPrio, HugeTlb, Rdma, + Systemd, } -impl Controllers { - pub fn to_string(&self) -> String { +impl fmt::Display for Controllers { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { match self { - Controllers::Pids => return "pids".to_string(), - Controllers::Mem => return "memory".to_string(), - Controllers::CpuSet => return "cpuset".to_string(), - Controllers::CpuAcct => return "cpuacct".to_string(), - Controllers::Cpu => return "cpu".to_string(), - Controllers::Devices => return "devices".to_string(), - Controllers::Freezer => return "freezer".to_string(), - Controllers::NetCls => return "net_cls".to_string(), - Controllers::BlkIo => return "blkio".to_string(), - Controllers::PerfEvent => return "perf_event".to_string(), - Controllers::NetPrio => return "net_prio".to_string(), - Controllers::HugeTlb => return "hugetlb".to_string(), - Controllers::Rdma => return "rdma".to_string(), + Controllers::Pids => write!(f, "pids"), + Controllers::Mem => write!(f, "memory"), + Controllers::CpuSet => write!(f, "cpuset"), + Controllers::CpuAcct => write!(f, "cpuacct"), + Controllers::Cpu => write!(f, "cpu"), + Controllers::Devices => write!(f, "devices"), + Controllers::Freezer => write!(f, "freezer"), + Controllers::NetCls => write!(f, "net_cls"), + Controllers::BlkIo => write!(f, "blkio"), + Controllers::PerfEvent => write!(f, "perf_event"), + Controllers::NetPrio => write!(f, "net_prio"), + Controllers::HugeTlb => write!(f, "hugetlb"), + Controllers::Rdma => write!(f, "rdma"), + Controllers::Systemd => write!(f, "name=systemd"), } } } @@ -120,6 +158,13 @@ mod sealed { fn get_path_mut(&mut self) -> &mut PathBuf; fn get_base(&self) -> &PathBuf; + /// Hooks running after controller crated, if have + fn post_create(&self) {} + + fn is_v2(&self) -> bool { + false + } + fn verify_path(&self) -> Result<()> { if self.get_path().starts_with(self.get_base()) { Ok(()) @@ -136,30 +181,59 @@ mod sealed { if w { match File::create(&path) { - Err(e) => return Err(Error::with_cause(ErrorKind::WriteFailed, e)), - Ok(file) => return Ok(file), + Err(e) => Err(Error::with_cause(ErrorKind::WriteFailed, e)), + Ok(file) => Ok(file), } } else { match File::open(&path) { - Err(e) => return Err(Error::with_cause(ErrorKind::ReadFailed, e)), - Ok(file) => return Ok(file), + Err(e) => Err(Error::with_cause(ErrorKind::ReadFailed, e)), + Ok(file) => Ok(file), } } } + fn get_max_value(&self, f: &str) -> Result { + self.open_path(f, false).and_then(|mut file| { + let mut string = String::new(); + let res = file.read_to_string(&mut string); + match res { + Ok(_) => parse_max_value(&string), + Err(e) => Err(Error::with_cause(ReadFailed, e)), + } + }) + } + #[doc(hidden)] fn path_exists(&self, p: &str) -> bool { - if let Err(_) = self.verify_path() { + if self.verify_path().is_err() { return false; } std::path::Path::new(p).exists() } + } + pub trait CustomizedAttribute: ControllerInternal { + fn set(&self, key: &str, value: &str) -> Result<()> { + self.open_path(key, true).and_then(|mut file| { + file.write_all(value.as_ref()) + .map_err(|e| Error::with_cause(WriteFailed, e)) + }) + } + + fn get(&self, key: &str) -> Result { + self.open_path(key, false).and_then(|mut file: File| { + let mut string = String::new(); + match file.read_to_string(&mut string) { + Ok(_) => Ok(string.trim().to_owned()), + Err(e) => Err(Error::with_cause(ReadFailed, e)), + } + }) + } } } -pub(crate) use crate::sealed::ControllerInternal; +pub(crate) use crate::sealed::{ControllerInternal, CustomizedAttribute}; /// A Controller is a subsystem attached to the control group. /// @@ -181,17 +255,31 @@ pub trait Controller { /// Does this controller already exist? fn exists(&self) -> bool; + /// Set notify_on_release + fn set_notify_on_release(&self, enable: bool) -> Result<()>; + + /// Set release_agent + fn set_release_agent(&self, path: &str) -> Result<()>; + /// Delete the controller. - fn delete(&self); + fn delete(&self) -> Result<()>; /// Attach a task to this controller. fn add_task(&self, pid: &CgroupPid) -> Result<()>; + /// Attach a task to this controller. + fn add_task_by_tgid(&self, pid: &CgroupPid) -> Result<()>; + /// Get the list of tasks that this controller has. fn tasks(&self) -> Vec; + + fn v2(&self) -> bool; } -impl Controller for T where T: ControllerInternal { +impl Controller for T +where + T: ControllerInternal, +{ fn control_type(&self) -> Controllers { ControllerInternal::control_type(self) } @@ -208,29 +296,60 @@ impl Controller for T where T: ControllerInternal { /// Create this controller fn create(&self) { - self.verify_path().expect("path should be valid"); + self.verify_path() + .unwrap_or_else(|_| panic!("path should be valid: {:?}", self.path())); - match ::std::fs::create_dir(self.get_path()) { - Ok(_) => (), - Err(e) => warn!("error create_dir {:?}", e), + match ::std::fs::create_dir_all(self.get_path()) { + Ok(_) => self.post_create(), + Err(e) => warn!("error create_dir: {:?} error: {:?}", self.get_path(), e), } } + /// Set notify_on_release + fn set_notify_on_release(&self, enable: bool) -> Result<()> { + self.open_path("notify_on_release", true) + .and_then(|mut file| { + write!(file, "{}", enable as i32) + .map_err(|e| Error::with_cause(ErrorKind::WriteFailed, e)) + }) + } + + /// Set release_agent + fn set_release_agent(&self, path: &str) -> Result<()> { + self.open_path("release_agent", true).and_then(|mut file| { + file.write_all(path.as_bytes()) + .map_err(|e| Error::with_cause(ErrorKind::WriteFailed, e)) + }) + } /// Does this controller already exist? fn exists(&self) -> bool { self.get_path().exists() } /// Delete the controller. - fn delete(&self) { - if self.get_path().exists() { - let _ = ::std::fs::remove_dir(self.get_path()); + fn delete(&self) -> Result<()> { + if !self.get_path().exists() { + return Ok(()); } + + remove_dir(self.get_path()) } /// Attach a task to this controller. fn add_task(&self, pid: &CgroupPid) -> Result<()> { - self.open_path("tasks", true).and_then(|mut file| { + let mut file = "tasks"; + if self.is_v2() { + file = "cgroup.procs"; + } + self.open_path(file, true).and_then(|mut file| { + file.write_all(pid.pid.to_string().as_ref()) + .map_err(|e| Error::with_cause(ErrorKind::WriteFailed, e)) + }) + } + + /// Attach a task to this controller by thread group id. + fn add_task_by_tgid(&self, pid: &CgroupPid) -> Result<()> { + self.open_path("cgroup.procs", true).and_then(|mut file| { file.write_all(pid.pid.to_string().as_ref()) .map_err(|e| Error::with_cause(ErrorKind::WriteFailed, e)) }) @@ -238,19 +357,48 @@ impl Controller for T where T: ControllerInternal { /// Get the list of tasks that this controller has. fn tasks(&self) -> Vec { - self.open_path("tasks", false) - .and_then(|file| { + let mut file = "tasks"; + if self.is_v2() { + file = "cgroup.procs"; + } + self.open_path(file, false) + .map(|file| { let bf = BufReader::new(file); let mut v = Vec::new(); - for line in bf.lines() { - if let Ok(line) = line { - let n = line.trim().parse().unwrap_or(0u64); - v.push(n); - } + for line in bf.lines().flatten() { + let n = line.trim().parse().unwrap_or(0u64); + v.push(n); } - Ok(v.into_iter().map(CgroupPid::from).collect()) - }).unwrap_or(vec![]) + v.into_iter().map(CgroupPid::from).collect() + }) + .unwrap_or_default() } + + fn v2(&self) -> bool { + self.is_v2() + } +} + +// remove_dir aims to remove cgroup path. It does so recursively, +// by removing any subdirectories (sub-cgroups) first. +fn remove_dir(dir: &Path) -> Result<()> { + // try the fast path first. + if fs::remove_dir(dir).is_ok() { + return Ok(()); + } + + if dir.exists() && dir.is_dir() { + for entry in fs::read_dir(dir).map_err(|e| Error::with_cause(ReadFailed, e))? { + let entry = entry.map_err(|e| Error::with_cause(ReadFailed, e))?; + let path = entry.path(); + if path.is_dir() { + remove_dir(&path)?; + } + } + fs::remove_dir(dir).map_err(|e| Error::with_cause(RemoveFailed, e))?; + } + + Ok(()) } #[doc(hidden)] @@ -260,7 +408,7 @@ pub trait ControllIdentifier { /// Control group hierarchy (right now, only V1 is supported, but in the future Unified will be /// implemented as well). -pub trait Hierarchy { +pub trait Hierarchy: std::fmt::Debug + Send + Sync { /// Returns what subsystems are supported by the hierarchy. fn subsystems(&self) -> Vec; @@ -270,74 +418,80 @@ pub trait Hierarchy { /// Return a handle to the root control group in the hierarchy. fn root_control_group(&self) -> Cgroup; - /// Checks whether a certain subsystem is supported in the hierarchy. - /// - /// This is an internal function and should not be used. - #[doc(hidden)] - fn check_support(&self, sub: Controllers) -> bool; + fn v2(&self) -> bool; } /// Resource limits for the memory subsystem. #[derive(Debug, Clone, Eq, PartialEq, Default)] pub struct MemoryResources { - /// Whether values should be applied to the controller. - pub update_values: bool, /// How much memory (in bytes) can the kernel consume. - pub kernel_memory_limit: u64, + pub kernel_memory_limit: Option, /// Upper limit of memory usage of the control group's tasks. - pub memory_hard_limit: u64, + pub memory_hard_limit: Option, /// How much memory the tasks in the control group can use when the system is under memory /// pressure. - pub memory_soft_limit: u64, + pub memory_soft_limit: Option, /// How much of the kernel's memory (in bytes) can be used for TCP-related buffers. - pub kernel_tcp_memory_limit: u64, + pub kernel_tcp_memory_limit: Option, /// How much memory and swap together can the tasks in the control group use. - pub memory_swap_limit: u64, + pub memory_swap_limit: Option, /// Controls the tendency of the kernel to swap out parts of the address space of the tasks to /// disk. Lower value implies less likely. /// /// Note, however, that a value of zero does not mean the process is never swapped out. Use the /// traditional `mlock(2)` system call for that purpose. - pub swappiness: u64, + pub swappiness: Option, + /// Customized key-value attributes + /// + /// # Usage: + /// ``` + /// let resource = &mut cgroups_rs::Resources::default(); + /// resource.memory.attrs.insert("memory.numa_balancing", "true".to_string()); + /// // apply here + pub attrs: std::collections::HashMap<&'static str, String>, } /// Resources limits on the number of processes. #[derive(Debug, Clone, Eq, PartialEq, Default)] pub struct PidResources { - /// Whether values should be applied to the controller. - pub update_values: bool, /// The maximum number of processes that can exist in the control group. /// /// Note that attaching processes to the control group will still succeed _even_ if the limit /// would be violated, however forks/clones inside the control group will have with `EAGAIN` if /// they would violate the limit set here. - pub maximum_number_of_processes: pid::PidMax, + pub maximum_number_of_processes: Option, } /// Resources limits about how the tasks can use the CPU. #[derive(Debug, Clone, Eq, PartialEq, Default)] pub struct CpuResources { - /// Whether values should be applied to the controller. - pub update_values: bool, // cpuset /// A comma-separated list of CPU IDs where the task in the control group can run. Dashes /// between numbers indicate ranges. - pub cpus: String, + pub cpus: Option, /// Same syntax as the `cpus` field of this structure, but applies to memory nodes instead of /// processors. - pub mems: String, + pub mems: Option, // cpu /// Weight of how much of the total CPU time should this control group get. Note that this is /// hierarchical, so this is weighted against the siblings of this control group. - pub shares: u64, + pub shares: Option, /// In one `period`, how much can the tasks run in nanoseconds. - pub quota: i64, + pub quota: Option, /// Period of time in nanoseconds. - pub period: u64, + pub period: Option, /// This is currently a no-operation. - pub realtime_runtime: i64, + pub realtime_runtime: Option, /// This is currently a no-operation. - pub realtime_period: u64, + pub realtime_period: Option, + /// Customized key-value attributes + /// # Usage: + /// ``` + /// let resource = &mut cgroups_rs::Resources::default(); + /// resource.cpu.attrs.insert("cpu.cfs_init_buffer_us", "10".to_string()); + /// // apply here + /// ``` + pub attrs: std::collections::HashMap<&'static str, String>, } /// A device resource that can be allowed or denied access to. @@ -358,8 +512,6 @@ pub struct DeviceResource { /// Limit the usage of devices for the control group's tasks. #[derive(Debug, Clone, Eq, PartialEq, Default)] pub struct DeviceResources { - /// Whether values should be applied to the controller. - pub update_values: bool, /// For each device in the list, the limits in the structure are applied. pub devices: Vec, } @@ -377,12 +529,10 @@ pub struct NetworkPriority { /// control group. #[derive(Debug, Clone, Eq, PartialEq, Default)] pub struct NetworkResources { - /// Whether values should be applied to the controller. - pub update_values: bool, /// The networking class identifier to attach to the packets. /// /// This can then later be used in iptables and such to have special rules. - pub class_id: u64, + pub class_id: Option, /// Priority of the egress traffic for each interface. pub priorities: Vec, } @@ -400,8 +550,6 @@ pub struct HugePageResource { /// Provides the ability to set consumption limit on each type of hugepages. #[derive(Debug, Clone, Eq, PartialEq, Default)] pub struct HugePageResources { - /// Whether values should be applied to the controller. - pub update_values: bool, /// Set a limit of consumption for each hugepages type. pub limits: Vec, } @@ -414,9 +562,9 @@ pub struct BlkIoDeviceResource { /// The minor number of the device. pub minor: u64, /// The weight of the device against the descendant nodes. - pub weight: u16, + pub weight: Option, /// The weight of the device against the sibling nodes. - pub leaf_weight: u16, + pub leaf_weight: Option, } /// Provides the ability to throttle a device (both byte/sec, and IO op/s) @@ -433,12 +581,10 @@ pub struct BlkIoDeviceThrottleResource { /// General block I/O resource limits. #[derive(Debug, Clone, Eq, PartialEq, Default)] pub struct BlkIoResources { - /// Whether values should be applied to the controller. - pub update_values: bool, /// The weight of the control group against descendant nodes. - pub weight: u16, + pub weight: Option, /// The weight of the control group against sibling nodes. - pub leaf_weight: u16, + pub leaf_weight: Option, /// For each device, a separate weight (both normal and leaf) can be provided. pub weight_device: Vec, /// Throttled read bytes/second can be provided for each device. @@ -493,75 +639,66 @@ impl<'a> From<&'a std::process::Child> for CgroupPid { impl Subsystem { fn enter(self, path: &Path) -> Self { match self { - Subsystem::Pid(cont) => Subsystem::Pid({ - let mut c = cont.clone(); - c.get_path_mut().push(path); - c + Subsystem::Pid(mut cont) => Subsystem::Pid({ + cont.get_path_mut().push(path); + cont }), - Subsystem::Mem(cont) => Subsystem::Mem({ - let mut c = cont.clone(); - c.get_path_mut().push(path); - c + Subsystem::Mem(mut cont) => Subsystem::Mem({ + cont.get_path_mut().push(path); + cont }), - Subsystem::CpuSet(cont) => Subsystem::CpuSet({ - let mut c = cont.clone(); - c.get_path_mut().push(path); - c + Subsystem::CpuSet(mut cont) => Subsystem::CpuSet({ + cont.get_path_mut().push(path); + cont }), - Subsystem::CpuAcct(cont) => Subsystem::CpuAcct({ - let mut c = cont.clone(); - c.get_path_mut().push(path); - c + Subsystem::CpuAcct(mut cont) => Subsystem::CpuAcct({ + cont.get_path_mut().push(path); + cont }), - Subsystem::Cpu(cont) => Subsystem::Cpu({ - let mut c = cont.clone(); - c.get_path_mut().push(path); - c + Subsystem::Cpu(mut cont) => Subsystem::Cpu({ + cont.get_path_mut().push(path); + cont }), - Subsystem::Devices(cont) => Subsystem::Devices({ - let mut c = cont.clone(); - c.get_path_mut().push(path); - c + Subsystem::Devices(mut cont) => Subsystem::Devices({ + cont.get_path_mut().push(path); + cont }), - Subsystem::Freezer(cont) => Subsystem::Freezer({ - let mut c = cont.clone(); - c.get_path_mut().push(path); - c + Subsystem::Freezer(mut cont) => Subsystem::Freezer({ + cont.get_path_mut().push(path); + cont }), - Subsystem::NetCls(cont) => Subsystem::NetCls({ - let mut c = cont.clone(); - c.get_path_mut().push(path); - c + Subsystem::NetCls(mut cont) => Subsystem::NetCls({ + cont.get_path_mut().push(path); + cont }), - Subsystem::BlkIo(cont) => Subsystem::BlkIo({ - let mut c = cont.clone(); - c.get_path_mut().push(path); - c + Subsystem::BlkIo(mut cont) => Subsystem::BlkIo({ + cont.get_path_mut().push(path); + cont }), - Subsystem::PerfEvent(cont) => Subsystem::PerfEvent({ - let mut c = cont.clone(); - c.get_path_mut().push(path); - c + Subsystem::PerfEvent(mut cont) => Subsystem::PerfEvent({ + cont.get_path_mut().push(path); + cont }), - Subsystem::NetPrio(cont) => Subsystem::NetPrio({ - let mut c = cont.clone(); - c.get_path_mut().push(path); - c + Subsystem::NetPrio(mut cont) => Subsystem::NetPrio({ + cont.get_path_mut().push(path); + cont }), - Subsystem::HugeTlb(cont) => Subsystem::HugeTlb({ - let mut c = cont.clone(); - c.get_path_mut().push(path); - c + Subsystem::HugeTlb(mut cont) => Subsystem::HugeTlb({ + cont.get_path_mut().push(path); + cont }), - Subsystem::Rdma(cont) => Subsystem::Rdma({ - let mut c = cont.clone(); - c.get_path_mut().push(path); - c + Subsystem::Rdma(mut cont) => Subsystem::Rdma({ + cont.get_path_mut().push(path); + cont + }), + Subsystem::Systemd(mut cont) => Subsystem::Systemd({ + cont.get_path_mut().push(path); + cont }), } } - fn to_controller(&self) -> &dyn Controller { + pub fn to_controller(&self) -> &dyn Controller { match self { Subsystem::Pid(cont) => cont, Subsystem::Mem(cont) => cont, @@ -576,6 +713,156 @@ impl Subsystem { Subsystem::NetPrio(cont) => cont, Subsystem::HugeTlb(cont) => cont, Subsystem::Rdma(cont) => cont, + Subsystem::Systemd(cont) => cont, + } + } + + pub fn controller_name(&self) -> String { + self.to_controller().control_type().to_string() + } +} + +/// The values for `memory.hight` or `pids.max` +#[derive(Eq, PartialEq, Copy, Clone, Debug)] +pub enum MaxValue { + /// This value is returned when the text is `"max"`. + Max, + /// When the value is a numerical value, they are returned via this enum field. + Value(i64), +} + +impl Default for MaxValue { + fn default() -> Self { + MaxValue::Max + } +} + +impl MaxValue { + fn to_i64(&self) -> i64 { + match self { + MaxValue::Max => -1, + MaxValue::Value(num) => *num, + } + } +} + +impl fmt::Display for MaxValue { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + match self { + MaxValue::Max => write!(f, "max"), + MaxValue::Value(num) => write!(f, "{}", num.to_string()), + } + } +} + +pub fn parse_max_value(s: &str) -> Result { + if s.trim() == "max" { + return Ok(MaxValue::Max); + } + match s.trim().parse() { + Ok(val) => Ok(MaxValue::Value(val)), + Err(e) => Err(Error::with_cause(ParseError, e)), + } +} + +// Flat keyed +// KEY0 VAL0\n +// KEY1 VAL1\n +pub fn flat_keyed_to_vec(mut file: File) -> Result> { + let mut content = String::new(); + file.read_to_string(&mut content) + .map_err(|e| Error::with_cause(ReadFailed, e))?; + + let mut v = Vec::new(); + for line in content.lines() { + let parts: Vec<&str> = line.split(' ').collect(); + if parts.len() == 2 { + if let Ok(i) = parts[1].parse::() { + v.push((parts[0].to_string(), i)); + } + } + } + Ok(v) +} + +// Flat keyed +// KEY0 VAL0\n +// KEY1 VAL1\n +pub fn flat_keyed_to_hashmap(mut file: File) -> Result> { + let mut content = String::new(); + file.read_to_string(&mut content) + .map_err(|e| Error::with_cause(ReadFailed, e))?; + + let mut h = HashMap::new(); + for line in content.lines() { + let parts: Vec<&str> = line.split(' ').collect(); + if parts.len() == 2 { + if let Ok(i) = parts[1].parse::() { + h.insert(parts[0].to_string(), i); + } } } + Ok(h) +} + +// Nested keyed +// KEY0 SUB_KEY0=VAL00 SUB_KEY1=VAL01... +// KEY1 SUB_KEY0=VAL10 SUB_KEY1=VAL11... +pub fn nested_keyed_to_hashmap(mut file: File) -> Result>> { + let mut content = String::new(); + file.read_to_string(&mut content) + .map_err(|e| Error::with_cause(ReadFailed, e))?; + + let mut h = HashMap::new(); + for line in content.lines() { + let parts: Vec<&str> = line.split(' ').collect(); + if parts.is_empty() { + continue; + } + let mut th = HashMap::new(); + for item in parts[1..].iter() { + let fields: Vec<&str> = item.split('=').collect(); + if fields.len() == 2 { + if let Ok(i) = fields[1].parse::() { + th.insert(fields[0].to_string(), i); + } + } + } + h.insert(parts[0].to_string(), th); + } + + Ok(h) +} + +fn read_from(mut file: File) -> Result +where + T: FromStr, + ::Err: 'static + Send + Sync + std::error::Error, +{ + let mut string = String::new(); + match file.read_to_string(&mut string) { + Ok(_) => string + .trim() + .parse::() + .map_err(|e| Error::with_cause(ParseError, e)), + Err(e) => Err(Error::with_cause(ReadFailed, e)), + } +} + +fn read_string_from(mut file: File) -> Result { + let mut string = String::new(); + match file.read_to_string(&mut string) { + Ok(_) => Ok(string.trim().to_string()), + Err(e) => Err(Error::with_cause(ReadFailed, e)), + } +} + +/// read and parse an u64 data +fn read_u64_from(file: File) -> Result { + read_from::(file) +} + +/// read and parse an i64 data +fn read_i64_from(file: File) -> Result { + read_from::(file) } diff --git a/src/memory.rs b/src/memory.rs index d7acaf81..b682a9cb 100644 --- a/src/memory.rs +++ b/src/memory.rs @@ -1,16 +1,28 @@ +// Copyright (c) 2018 Levente Kurusa +// Copyright (c) 2020 Ant Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! This module contains the implementation of the `memory` cgroup subsystem. //! //! See the Kernel's documentation for more information about this subsystem, found at: //! [Documentation/cgroup-v1/memory.txt](https://www.kernel.org/doc/Documentation/cgroup-v1/memory.txt) -use std::fs::File; -use std::io::{Read, Write}; +use std::collections::HashMap; +use std::io::Write; use std::path::PathBuf; +use std::sync::mpsc::Receiver; -use crate::error::*; use crate::error::ErrorKind::*; +use crate::error::*; +use crate::events; +use crate::{read_i64_from, read_string_from, read_u64_from}; + +use crate::flat_keyed_to_hashmap; use crate::{ - ControllIdentifier, ControllerInternal, Controllers, MemoryResources, Resources, Subsystem, + ControllIdentifier, ControllerInternal, Controllers, CustomizedAttribute, MaxValue, + MemoryResources, Resources, Subsystem, }; /// A controller that allows controlling the `memory` subsystem of a Cgroup. @@ -22,6 +34,15 @@ use crate::{ pub struct MemController { base: PathBuf, path: PathBuf, + v2: bool, +} + +#[derive(Default, Debug, PartialEq, Eq)] +pub struct SetMemory { + pub low: Option, + pub high: Option, + pub min: Option, + pub max: Option, } /// Controls statistics and controls about the OOM killer operating in this control group. @@ -35,13 +56,32 @@ pub struct OomControl { pub oom_kill: u64, } +#[allow(clippy::unnecessary_wraps)] fn parse_oom_control(s: String) -> Result { let spl = s.split_whitespace().collect::>(); + let oom_kill_disable = if spl.len() > 1 { + spl[1].parse::().unwrap() == 1 + } else { + false + }; + + let under_oom = if spl.len() > 3 { + spl[3].parse::().unwrap() == 1 + } else { + false + }; + + let oom_kill = if spl.len() > 5 { + spl[5].parse::().unwrap() + } else { + 0 + }; + Ok(OomControl { - oom_kill_disable: spl[1].parse::().unwrap() == 1, - under_oom: spl[3].parse::().unwrap() == 1, - oom_kill: spl[5].parse::().unwrap(), + oom_kill_disable, + under_oom, + oom_kill, }) } @@ -83,18 +123,19 @@ pub struct NumaStat { pub hierarchical_unevictable_pages_per_node: Vec, } +#[allow(clippy::unnecessary_wraps)] fn parse_numa_stat(s: String) -> Result { // Parse the number of nodes - let _nodes = (s.split_whitespace().collect::>().len() - 8) / 8; + let _nodes = (s.split_whitespace().count() - 8) / 8; let mut ls = s.lines(); let total_line = ls.next().unwrap(); let file_line = ls.next().unwrap(); let anon_line = ls.next().unwrap(); let unevict_line = ls.next().unwrap(); - let hier_total_line = ls.next().unwrap(); - let hier_file_line = ls.next().unwrap(); - let hier_anon_line = ls.next().unwrap(); - let hier_unevict_line = ls.next().unwrap(); + let hier_total_line = ls.next().unwrap_or_default(); + let hier_file_line = ls.next().unwrap_or_default(); + let hier_anon_line = ls.next().unwrap_or_default(); + let hier_unevict_line = ls.next().unwrap_or_default(); Ok(NumaStat { total_pages: total_line @@ -103,13 +144,14 @@ fn parse_numa_stat(s: String) -> Result { .parse::() .unwrap_or(0), total_pages_per_node: { - let spl = &total_line.split(" ").collect::>()[1..]; + let spl = &total_line.split(' ').collect::>()[1..]; spl.iter() .map(|x| { - x.split("=").collect::>()[1] + x.split('=').collect::>()[1] .parse::() .unwrap_or(0) - }).collect() + }) + .collect() }, file_pages: file_line .split(|x| x == ' ' || x == '=') @@ -117,13 +159,14 @@ fn parse_numa_stat(s: String) -> Result { .parse::() .unwrap_or(0), file_pages_per_node: { - let spl = &file_line.split(" ").collect::>()[1..]; + let spl = &file_line.split(' ').collect::>()[1..]; spl.iter() .map(|x| { - x.split("=").collect::>()[1] + x.split('=').collect::>()[1] .parse::() .unwrap_or(0) - }).collect() + }) + .collect() }, anon_pages: anon_line .split(|x| x == ' ' || x == '=') @@ -131,13 +174,14 @@ fn parse_numa_stat(s: String) -> Result { .parse::() .unwrap_or(0), anon_pages_per_node: { - let spl = &anon_line.split(" ").collect::>()[1..]; + let spl = &anon_line.split(' ').collect::>()[1..]; spl.iter() .map(|x| { - x.split("=").collect::>()[1] + x.split('=').collect::>()[1] .parse::() .unwrap_or(0) - }).collect() + }) + .collect() }, unevictable_pages: unevict_line .split(|x| x == ' ' || x == '=') @@ -145,69 +189,114 @@ fn parse_numa_stat(s: String) -> Result { .parse::() .unwrap_or(0), unevictable_pages_per_node: { - let spl = &unevict_line.split(" ").collect::>()[1..]; + let spl = &unevict_line.split(' ').collect::>()[1..]; spl.iter() .map(|x| { - x.split("=").collect::>()[1] + x.split('=').collect::>()[1] .parse::() .unwrap_or(0) - }).collect() + }) + .collect() + }, + hierarchical_total_pages: { + if !hier_total_line.is_empty() { + hier_total_line + .split(|x| x == ' ' || x == '=') + .collect::>()[1] + .parse::() + .unwrap_or(0) + } else { + 0 + } }, - hierarchical_total_pages: hier_total_line - .split(|x| x == ' ' || x == '=') - .collect::>()[1] - .parse::() - .unwrap_or(0), hierarchical_total_pages_per_node: { - let spl = &hier_total_line.split(" ").collect::>()[1..]; - spl.iter() - .map(|x| { - x.split("=").collect::>()[1] - .parse::() - .unwrap_or(0) - }).collect() + if !hier_total_line.is_empty() { + let spl = &hier_total_line.split(' ').collect::>()[1..]; + spl.iter() + .map(|x| { + x.split('=').collect::>()[1] + .parse::() + .unwrap_or(0) + }) + .collect() + } else { + Vec::new() + } + }, + hierarchical_file_pages: { + if !hier_file_line.is_empty() { + hier_file_line + .split(|x| x == ' ' || x == '=') + .collect::>()[1] + .parse::() + .unwrap_or(0) + } else { + 0 + } }, - hierarchical_file_pages: hier_file_line - .split(|x| x == ' ' || x == '=') - .collect::>()[1] - .parse::() - .unwrap_or(0), hierarchical_file_pages_per_node: { - let spl = &hier_file_line.split(" ").collect::>()[1..]; - spl.iter() - .map(|x| { - x.split("=").collect::>()[1] - .parse::() - .unwrap_or(0) - }).collect() + if !hier_file_line.is_empty() { + let spl = &hier_file_line.split(' ').collect::>()[1..]; + spl.iter() + .map(|x| { + x.split('=').collect::>()[1] + .parse::() + .unwrap_or(0) + }) + .collect() + } else { + Vec::new() + } + }, + hierarchical_anon_pages: { + if !hier_anon_line.is_empty() { + hier_anon_line + .split(|x| x == ' ' || x == '=') + .collect::>()[1] + .parse::() + .unwrap_or(0) + } else { + 0 + } }, - hierarchical_anon_pages: hier_anon_line - .split(|x| x == ' ' || x == '=') - .collect::>()[1] - .parse::() - .unwrap_or(0), hierarchical_anon_pages_per_node: { - let spl = &hier_anon_line.split(" ").collect::>()[1..]; - spl.iter() - .map(|x| { - x.split("=").collect::>()[1] - .parse::() - .unwrap_or(0) - }).collect() + if !hier_anon_line.is_empty() { + let spl = &hier_anon_line.split(' ').collect::>()[1..]; + spl.iter() + .map(|x| { + x.split('=').collect::>()[1] + .parse::() + .unwrap_or(0) + }) + .collect() + } else { + Vec::new() + } + }, + hierarchical_unevictable_pages: { + if !hier_unevict_line.is_empty() { + hier_unevict_line + .split(|x| x == ' ' || x == '=') + .collect::>()[1] + .parse::() + .unwrap_or(0) + } else { + 0 + } }, - hierarchical_unevictable_pages: hier_unevict_line - .split(|x| x == ' ' || x == '=') - .collect::>()[1] - .parse::() - .unwrap_or(0), hierarchical_unevictable_pages_per_node: { - let spl = &hier_unevict_line.split(" ").collect::>()[1..]; - spl.iter() - .map(|x| { - x.split("=").collect::>()[1] - .parse::() - .unwrap_or(0) - }).collect() + if !hier_unevict_line.is_empty() { + let spl = &hier_unevict_line.split(' ').collect::>()[1..]; + spl.iter() + .map(|x| { + x.split('=').collect::>()[1] + .parse::() + .unwrap_or(0) + }) + .collect() + } else { + Vec::new() + } }, }) } @@ -231,8 +320,8 @@ pub struct MemoryStat { pub inactive_file: u64, pub active_file: u64, pub unevictable: u64, - pub hierarchical_memory_limit: u64, - pub hierarchical_memsw_limit: u64, + pub hierarchical_memory_limit: i64, + pub hierarchical_memsw_limit: i64, pub total_cache: u64, pub total_rss: u64, pub total_rss_huge: u64, @@ -250,52 +339,64 @@ pub struct MemoryStat { pub total_inactive_file: u64, pub total_active_file: u64, pub total_unevictable: u64, + pub raw: HashMap, } +#[allow(clippy::unnecessary_wraps)] fn parse_memory_stat(s: String) -> Result { - let sp: Vec<&str> = s - .split_whitespace() - .filter(|x| x.parse::().is_ok()) - .collect(); + let mut raw = HashMap::new(); + + for l in s.lines() { + let t: Vec<&str> = l.split(' ').collect(); + if t.len() != 2 { + continue; + } + let n = t[1].trim().parse::(); + if n.is_err() { + continue; + } + + raw.insert(t[0].to_string(), n.unwrap()); + } - let mut spl = sp.iter(); Ok(MemoryStat { - cache: spl.next().unwrap().parse::().unwrap(), - rss: spl.next().unwrap().parse::().unwrap(), - rss_huge: spl.next().unwrap().parse::().unwrap(), - shmem: spl.next().unwrap().parse::().unwrap(), - mapped_file: spl.next().unwrap().parse::().unwrap(), - dirty: spl.next().unwrap().parse::().unwrap(), - writeback: spl.next().unwrap().parse::().unwrap(), - swap: spl.next().unwrap().parse::().unwrap(), - pgpgin: spl.next().unwrap().parse::().unwrap(), - pgpgout: spl.next().unwrap().parse::().unwrap(), - pgfault: spl.next().unwrap().parse::().unwrap(), - pgmajfault: spl.next().unwrap().parse::().unwrap(), - inactive_anon: spl.next().unwrap().parse::().unwrap(), - active_anon: spl.next().unwrap().parse::().unwrap(), - inactive_file: spl.next().unwrap().parse::().unwrap(), - active_file: spl.next().unwrap().parse::().unwrap(), - unevictable: spl.next().unwrap().parse::().unwrap(), - hierarchical_memory_limit: spl.next().unwrap().parse::().unwrap(), - hierarchical_memsw_limit: spl.next().unwrap().parse::().unwrap(), - total_cache: spl.next().unwrap().parse::().unwrap(), - total_rss: spl.next().unwrap().parse::().unwrap(), - total_rss_huge: spl.next().unwrap().parse::().unwrap(), - total_shmem: spl.next().unwrap().parse::().unwrap(), - total_mapped_file: spl.next().unwrap().parse::().unwrap(), - total_dirty: spl.next().unwrap().parse::().unwrap(), - total_writeback: spl.next().unwrap().parse::().unwrap(), - total_swap: spl.next().unwrap().parse::().unwrap(), - total_pgpgin: spl.next().unwrap().parse::().unwrap(), - total_pgpgout: spl.next().unwrap().parse::().unwrap(), - total_pgfault: spl.next().unwrap().parse::().unwrap(), - total_pgmajfault: spl.next().unwrap().parse::().unwrap(), - total_inactive_anon: spl.next().unwrap().parse::().unwrap(), - total_active_anon: spl.next().unwrap().parse::().unwrap(), - total_inactive_file: spl.next().unwrap().parse::().unwrap(), - total_active_file: spl.next().unwrap().parse::().unwrap(), - total_unevictable: spl.next().unwrap().parse::().unwrap(), + cache: *raw.get("cache").unwrap_or(&0), + rss: *raw.get("rss").unwrap_or(&0), + rss_huge: *raw.get("rss_huge").unwrap_or(&0), + shmem: *raw.get("shmem").unwrap_or(&0), + mapped_file: *raw.get("mapped_file").unwrap_or(&0), + dirty: *raw.get("dirty").unwrap_or(&0), + writeback: *raw.get("writeback").unwrap_or(&0), + swap: *raw.get("swap").unwrap_or(&0), + pgpgin: *raw.get("pgpgin").unwrap_or(&0), + pgpgout: *raw.get("pgpgout").unwrap_or(&0), + pgfault: *raw.get("pgfault").unwrap_or(&0), + pgmajfault: *raw.get("pgmajfault").unwrap_or(&0), + inactive_anon: *raw.get("inactive_anon").unwrap_or(&0), + active_anon: *raw.get("active_anon").unwrap_or(&0), + inactive_file: *raw.get("inactive_file").unwrap_or(&0), + active_file: *raw.get("active_file").unwrap_or(&0), + unevictable: *raw.get("unevictable").unwrap_or(&0), + hierarchical_memory_limit: *raw.get("hierarchical_memory_limit").unwrap_or(&0) as i64, + hierarchical_memsw_limit: *raw.get("hierarchical_memsw_limit").unwrap_or(&0) as i64, + total_cache: *raw.get("total_cache").unwrap_or(&0), + total_rss: *raw.get("total_rss").unwrap_or(&0), + total_rss_huge: *raw.get("total_rss_huge").unwrap_or(&0), + total_shmem: *raw.get("total_shmem").unwrap_or(&0), + total_mapped_file: *raw.get("total_mapped_file").unwrap_or(&0), + total_dirty: *raw.get("total_dirty").unwrap_or(&0), + total_writeback: *raw.get("total_writeback").unwrap_or(&0), + total_swap: *raw.get("total_swap").unwrap_or(&0), + total_pgpgin: *raw.get("total_pgpgin").unwrap_or(&0), + total_pgpgout: *raw.get("total_pgpgout").unwrap_or(&0), + total_pgfault: *raw.get("total_pgfault").unwrap_or(&0), + total_pgmajfault: *raw.get("total_pgmajfault").unwrap_or(&0), + total_inactive_anon: *raw.get("total_inactive_anon").unwrap_or(&0), + total_active_anon: *raw.get("total_active_anon").unwrap_or(&0), + total_inactive_file: *raw.get("total_inactive_file").unwrap_or(&0), + total_active_file: *raw.get("total_active_file").unwrap_or(&0), + total_unevictable: *raw.get("total_unevictable").unwrap_or(&0), + raw, }) } @@ -306,7 +407,7 @@ pub struct MemSwap { /// How many times the limit has been hit. pub fail_cnt: u64, /// Memory and swap usage limit in bytes. - pub limit_in_bytes: u64, + pub limit_in_bytes: i64, /// Current usage of memory and swap in bytes. pub usage_in_bytes: u64, /// The maximum observed usage of memory and swap in bytes. @@ -320,7 +421,7 @@ pub struct Memory { /// How many times the limit has been hit. pub fail_cnt: u64, /// The limit in bytes of the memory usage of the control group's tasks. - pub limit_in_bytes: u64, + pub limit_in_bytes: i64, /// The current usage of memory by the control group's tasks. pub usage_in_bytes: u64, /// The maximum observed usage of memory by the control group's tasks. @@ -342,7 +443,7 @@ pub struct Memory { pub oom_control: OomControl, /// Allows setting a limit to memory usage which is enforced when the system (note, _not_ the /// control group) detects memory pressure. - pub soft_limit_in_bytes: u64, + pub soft_limit_in_bytes: i64, /// Contains a wide array of statistics about the memory usage of the tasks in the control /// group. pub stat: MemoryStat, @@ -365,7 +466,7 @@ pub struct Tcp { pub fail_cnt: u64, /// The limit in bytes of the memory usage of the kernel's TCP buffers by control group's /// tasks. - pub limit_in_bytes: u64, + pub limit_in_bytes: i64, /// The current memory used by the kernel's TCP buffers related to these tasks. pub usage_in_bytes: u64, /// The observed maximum usage of memory by the kernel's TCP buffers (that originated from @@ -382,7 +483,7 @@ pub struct Kmem { /// How many times the limit has been hit. pub fail_cnt: u64, /// The limit in bytes of the kernel memory used by the control group's tasks. - pub limit_in_bytes: u64, + pub limit_in_bytes: i64, /// The current usage of kernel memory used by the control group's tasks, in bytes. pub usage_in_bytes: u64, /// The maximum observed usage of kernel memory used by the control group's tasks, in bytes. @@ -405,31 +506,98 @@ impl ControllerInternal for MemController { &self.base } + fn is_v2(&self) -> bool { + self.v2 + } + fn apply(&self, res: &Resources) -> Result<()> { // get the resources that apply to this controller let memres: &MemoryResources = &res.memory; - if memres.update_values { - let _ = self.set_limit(memres.memory_hard_limit); - let _ = self.set_soft_limit(memres.memory_soft_limit); - let _ = self.set_kmem_limit(memres.kernel_memory_limit); - let _ = self.set_memswap_limit(memres.memory_swap_limit); - let _ = self.set_tcp_limit(memres.kernel_tcp_memory_limit); - let _ = self.set_swappiness(memres.swappiness); - } + update!(self, set_limit, memres.memory_hard_limit); + update!(self, set_soft_limit, memres.memory_soft_limit); + update!(self, set_kmem_limit, memres.kernel_memory_limit); + update!(self, set_memswap_limit, memres.memory_swap_limit); + update!(self, set_tcp_limit, memres.kernel_tcp_memory_limit); + update!(self, set_swappiness, memres.swappiness); + + memres.attrs.iter().for_each(|(k, v)| { + let _ = self.set(k, v); + }); Ok(()) } } impl MemController { - /// Contructs a new `MemController` with `oroot` serving as the root of the control group. - pub fn new(oroot: PathBuf) -> Self { - let mut root = oroot; - root.push(Self::controller_type().to_string()); + /// Contructs a new `MemController` with `root` serving as the root of the control group. + pub fn new(root: PathBuf, v2: bool) -> Self { Self { base: root.clone(), path: root, + v2, + } + } + + // for v2 + pub fn set_mem(&self, m: SetMemory) -> Result<()> { + let values = vec![ + (m.high, "memory.high"), + (m.low, "memory.low"), + (m.max, "memory.max"), + (m.min, "memory.min"), + ]; + for value in values { + let v = value.0; + let f = value.1; + if let Some(v) = v { + let v = v.to_string(); + self.open_path(f, true).and_then(|mut file| { + file.write_all(v.as_ref()) + .map_err(|e| Error::with_cause(WriteFailed, e)) + })?; + } + } + Ok(()) + } + + // for v2 + pub fn get_mem(&self) -> Result { + let mut m: SetMemory = Default::default(); + self.get_max_value("memory.high") + .map(|x| m.high = Some(x))?; + self.get_max_value("memory.low").map(|x| m.low = Some(x))?; + self.get_max_value("memory.max").map(|x| m.max = Some(x))?; + self.get_max_value("memory.min").map(|x| m.min = Some(x))?; + + Ok(m) + } + + fn memory_stat_v2(&self) -> Memory { + let set = self.get_mem().unwrap(); + + Memory { + fail_cnt: 0, + limit_in_bytes: set.max.unwrap().to_i64(), + usage_in_bytes: self + .open_path("memory.current", false) + .and_then(read_u64_from) + .unwrap_or(0), + max_usage_in_bytes: 0, + move_charge_at_immigrate: 0, + numa_stat: NumaStat::default(), + oom_control: OomControl::default(), + soft_limit_in_bytes: set.low.unwrap().to_i64(), + stat: self + .open_path("memory.stat", false) + .and_then(read_string_from) + .and_then(parse_memory_stat) + .unwrap_or_default(), + swappiness: self + .open_path("memory.swap.current", false) + .and_then(read_u64_from) + .unwrap_or(0), + use_hierarchy: 0, } } @@ -439,6 +607,10 @@ impl MemController { /// See the individual fields for more explanation, and as always, remember to consult the /// kernel Documentation and/or sources. pub fn memory_stat(&self) -> Memory { + if self.v2 { + return self.memory_stat_v2(); + } + Memory { fail_cnt: self .open_path("memory.failcnt", false) @@ -446,7 +618,7 @@ impl MemController { .unwrap_or(0), limit_in_bytes: self .open_path("memory.limit_in_bytes", false) - .and_then(read_u64_from) + .and_then(read_i64_from) .unwrap_or(0), usage_in_bytes: self .open_path("memory.usage_in_bytes", false) @@ -464,21 +636,21 @@ impl MemController { .open_path("memory.numa_stat", false) .and_then(read_string_from) .and_then(parse_numa_stat) - .unwrap_or(NumaStat::default()), + .unwrap_or_default(), oom_control: self .open_path("memory.oom_control", false) .and_then(read_string_from) .and_then(parse_oom_control) - .unwrap_or(OomControl::default()), + .unwrap_or_default(), soft_limit_in_bytes: self .open_path("memory.soft_limit_in_bytes", false) - .and_then(read_u64_from) + .and_then(read_i64_from) .unwrap_or(0), stat: self .open_path("memory.stat", false) .and_then(read_string_from) .and_then(parse_memory_stat) - .unwrap_or(MemoryStat::default()), + .unwrap_or_default(), swappiness: self .open_path("memory.swappiness", false) .and_then(read_u64_from) @@ -499,8 +671,8 @@ impl MemController { .unwrap_or(0), limit_in_bytes: self .open_path("memory.kmem.limit_in_bytes", false) - .and_then(read_u64_from) - .unwrap_or(0), + .and_then(read_i64_from) + .unwrap_or(-1), usage_in_bytes: self .open_path("memory.kmem.usage_in_bytes", false) .and_then(read_u64_from) @@ -512,7 +684,7 @@ impl MemController { slabinfo: self .open_path("memory.kmem.slabinfo", false) .and_then(read_string_from) - .unwrap_or("".to_string()), + .unwrap_or_default(), } } @@ -526,7 +698,7 @@ impl MemController { .unwrap_or(0), limit_in_bytes: self .open_path("memory.kmem.tcp.limit_in_bytes", false) - .and_then(read_u64_from) + .and_then(read_i64_from) .unwrap_or(0), usage_in_bytes: self .open_path("memory.kmem.tcp.usage_in_bytes", false) @@ -539,9 +711,32 @@ impl MemController { } } + pub fn memswap_v2(&self) -> MemSwap { + MemSwap { + fail_cnt: self + .open_path("memory.swap.events", false) + .and_then(flat_keyed_to_hashmap) + .map(|x| *x.get("fail").unwrap_or(&0) as u64) + .unwrap(), + limit_in_bytes: self + .open_path("memory.swap.max", false) + .and_then(read_i64_from) + .unwrap_or(0), + usage_in_bytes: self + .open_path("memory.swap.current", false) + .and_then(read_u64_from) + .unwrap_or(0), + max_usage_in_bytes: 0, + } + } + /// Gathers information about the memory usage of the control group including the swap usage /// (if any). pub fn memswap(&self) -> MemSwap { + if self.v2 { + return self.memswap_v2(); + } + MemSwap { fail_cnt: self .open_path("memory.memsw.failcnt", false) @@ -549,7 +744,7 @@ impl MemController { .unwrap_or(0), limit_in_bytes: self .open_path("memory.memsw.limit_in_bytes", false) - .and_then(read_u64_from) + .and_then(read_i64_from) .unwrap_or(0), usage_in_bytes: self .open_path("memory.memsw.usage_in_bytes", false) @@ -564,15 +759,19 @@ impl MemController { /// Reset the fail counter pub fn reset_fail_count(&self) -> Result<()> { - self.open_path("memory.failcnt", true) - .and_then(|mut file| { - file.write_all("0".to_string().as_ref()) - .map_err(|e| Error::with_cause(WriteFailed, e)) - }) + self.open_path("memory.failcnt", true).and_then(|mut file| { + file.write_all("0".to_string().as_ref()) + .map_err(|e| Error::with_cause(WriteFailed, e)) + }) } /// Reset the kernel memory fail counter pub fn reset_kmem_fail_count(&self) -> Result<()> { + // Ignore kmem because there is no kmem in cgroup v2 + if self.v2 { + return Ok(()); + } + self.open_path("memory.kmem.failcnt", true) .and_then(|mut file| { file.write_all("0".to_string().as_ref()) @@ -582,6 +781,11 @@ impl MemController { /// Reset the TCP related fail counter pub fn reset_tcp_fail_count(&self) -> Result<()> { + // Ignore kmem because there is no kmem in cgroup v2 + if self.v2 { + return Ok(()); + } + self.open_path("memory.kmem.tcp.failcnt", true) .and_then(|mut file| { file.write_all("0".to_string().as_ref()) @@ -598,17 +802,34 @@ impl MemController { }) } - /// Set the memory usage limit of the control group, in bytes. - pub fn set_limit(&self, limit: u64) -> Result<()> { - self.open_path("memory.limit_in_bytes", true) + /// Reset the max memory usage recorded + pub fn reset_max_usage(&self) -> Result<()> { + self.open_path("memory.max_usage_in_bytes", true) .and_then(|mut file| { - file.write_all(limit.to_string().as_ref()) + file.write_all("0".to_string().as_ref()) .map_err(|e| Error::with_cause(WriteFailed, e)) }) } + /// Set the memory usage limit of the control group, in bytes. + pub fn set_limit(&self, limit: i64) -> Result<()> { + let mut file = "memory.limit_in_bytes"; + if self.v2 { + file = "memory.max"; + } + self.open_path(file, true).and_then(|mut file| { + file.write_all(limit.to_string().as_ref()) + .map_err(|e| Error::with_cause(WriteFailed, e)) + }) + } + /// Set the kernel memory limit of the control group, in bytes. - pub fn set_kmem_limit(&self, limit: u64) -> Result<()> { + pub fn set_kmem_limit(&self, limit: i64) -> Result<()> { + // Ignore kmem because there is no kmem in cgroup v2 + if self.v2 { + return Ok(()); + } + self.open_path("memory.kmem.limit_in_bytes", true) .and_then(|mut file| { file.write_all(limit.to_string().as_ref()) @@ -617,16 +838,24 @@ impl MemController { } /// Set the memory+swap limit of the control group, in bytes. - pub fn set_memswap_limit(&self, limit: u64) -> Result<()> { - self.open_path("memory.memsw.limit_in_bytes", true) - .and_then(|mut file| { - file.write_all(limit.to_string().as_ref()) - .map_err(|e| Error::with_cause(WriteFailed, e)) - }) + pub fn set_memswap_limit(&self, limit: i64) -> Result<()> { + let mut file = "memory.memsw.limit_in_bytes"; + if self.v2 { + file = "memory.swap.max"; + } + self.open_path(file, true).and_then(|mut file| { + file.write_all(limit.to_string().as_ref()) + .map_err(|e| Error::with_cause(WriteFailed, e)) + }) } /// Set how much kernel memory can be used for TCP-related buffers by the control group. - pub fn set_tcp_limit(&self, limit: u64) -> Result<()> { + pub fn set_tcp_limit(&self, limit: i64) -> Result<()> { + // Ignore kmem because there is no kmem in cgroup v2 + if self.v2 { + return Ok(()); + } + self.open_path("memory.kmem.tcp.limit_in_bytes", true) .and_then(|mut file| { file.write_all(limit.to_string().as_ref()) @@ -638,12 +867,15 @@ impl MemController { /// /// This limit is enforced when the system is nearing OOM conditions. Contrast this with the /// hard limit, which is _always_ enforced. - pub fn set_soft_limit(&self, limit: u64) -> Result<()> { - self.open_path("memory.soft_limit_in_bytes", true) - .and_then(|mut file| { - file.write_all(limit.to_string().as_ref()) - .map_err(|e| Error::with_cause(WriteFailed, e)) - }) + pub fn set_soft_limit(&self, limit: i64) -> Result<()> { + let mut file = "memory.soft_limit_in_bytes"; + if self.v2 { + file = "memory.low" + } + self.open_path(file, true).and_then(|mut file| { + file.write_all(limit.to_string().as_ref()) + .map_err(|e| Error::with_cause(WriteFailed, e)) + }) } /// Set how likely the kernel is to swap out parts of the address space used by the control @@ -651,12 +883,32 @@ impl MemController { /// /// Note that a value of zero does not imply that the process will not be swapped out. pub fn set_swappiness(&self, swp: u64) -> Result<()> { - self.open_path("memory.swappiness", true) + let mut file = "memory.swappiness"; + if self.v2 { + file = "memory.swap.max" + } + + self.open_path(file, true).and_then(|mut file| { + file.write_all(swp.to_string().as_ref()) + .map_err(|e| Error::with_cause(WriteFailed, e)) + }) + } + + pub fn disable_oom_killer(&self) -> Result<()> { + self.open_path("memory.oom_control", true) .and_then(|mut file| { - file.write_all(swp.to_string().as_ref()) + file.write_all("1".to_string().as_ref()) .map_err(|e| Error::with_cause(WriteFailed, e)) }) } + + pub fn register_oom_event(&self, key: &str) -> Result> { + if self.v2 { + events::notify_on_oom_v2(key, self.get_path()) + } else { + events::notify_on_oom_v1(key, self.get_path()) + } + } } impl ControllIdentifier for MemController { @@ -665,6 +917,8 @@ impl ControllIdentifier for MemController { } } +impl CustomizedAttribute for MemController {} + impl<'a> From<&'a Subsystem> for &'a MemController { fn from(sub: &'a Subsystem) -> &'a MemController { unsafe { @@ -672,29 +926,14 @@ impl<'a> From<&'a Subsystem> for &'a MemController { Subsystem::Mem(c) => c, _ => { assert_eq!(1, 0); - ::std::mem::uninitialized() + let v = std::mem::MaybeUninit::uninit(); + v.assume_init() } } } } } -fn read_u64_from(mut file: File) -> Result { - let mut string = String::new(); - match file.read_to_string(&mut string) { - Ok(_) => string.trim().parse().map_err(|e| Error::with_cause(ParseError, e)), - Err(e) => Err(Error::with_cause(ReadFailed, e)), - } -} - -fn read_string_from(mut file: File) -> Result { - let mut string = String::new(); - match file.read_to_string(&mut string) { - Ok(_) => Ok(string.trim().to_string()), - Err(e) => Err(Error::with_cause(ReadFailed, e)), - } -} - #[cfg(test)] mod tests { use crate::memory::{ @@ -712,7 +951,24 @@ hierarchical_anon=770402 N0=770402 N1=123 hierarchical_unevictable=20 N0=20 N1=123 "; - static GOOD_OOMCONTROL_VAL: &str = "\ + static GOOD_VALUE_NON_HIERARCHICAL: &str = "\ +total=51189 N0=51189 N1=123 +file=50175 N0=50175 N1=123 +anon=1014 N0=1014 N1=123 +unevictable=0 N0=0 N1=123 +"; + + static GOOD_OOMCONTROL_VAL_1: &str = "\ +oom_kill_disable 0 +oom_kill 1337 +"; + + static GOOD_OOMCONTROL_VAL_2: &str = "\ +oom_kill_disable 0 +under_oom 1 +"; + + static GOOD_OOMCONTROL_VAL_3: &str = "\ oom_kill_disable 0 under_oom 1 oom_kill 1337 @@ -782,11 +1038,61 @@ total_unevictable 81920 hierarchical_unevictable_pages_per_node: vec![20, 123], } ); + let ok = parse_numa_stat(GOOD_VALUE_NON_HIERARCHICAL.to_string()).unwrap(); + assert_eq!( + ok, + NumaStat { + total_pages: 51189, + total_pages_per_node: vec![51189, 123], + file_pages: 50175, + file_pages_per_node: vec![50175, 123], + anon_pages: 1014, + anon_pages_per_node: vec![1014, 123], + unevictable_pages: 0, + unevictable_pages_per_node: vec![0, 123], + + hierarchical_total_pages: 0, + hierarchical_total_pages_per_node: vec![], + hierarchical_file_pages: 0, + hierarchical_file_pages_per_node: vec![], + hierarchical_anon_pages: 0, + hierarchical_anon_pages_per_node: vec![], + hierarchical_unevictable_pages: 0, + hierarchical_unevictable_pages_per_node: vec![], + } + ); } #[test] fn test_parse_oom_control() { - let ok = parse_oom_control(GOOD_OOMCONTROL_VAL.to_string()).unwrap(); + let ok = parse_oom_control("".to_string()).unwrap(); + assert_eq!( + ok, + OomControl { + oom_kill_disable: false, + under_oom: false, + oom_kill: 0, + } + ); + let ok = parse_oom_control(GOOD_OOMCONTROL_VAL_1.to_string()).unwrap(); + assert_eq!( + ok, + OomControl { + oom_kill_disable: false, + under_oom: false, + oom_kill: 0, + } + ); + let ok = parse_oom_control(GOOD_OOMCONTROL_VAL_2.to_string()).unwrap(); + assert_eq!( + ok, + OomControl { + oom_kill_disable: false, + under_oom: true, + oom_kill: 0, + } + ); + let ok = parse_oom_control(GOOD_OOMCONTROL_VAL_3.to_string()).unwrap(); assert_eq!( ok, OomControl { @@ -800,6 +1106,7 @@ total_unevictable 81920 #[test] fn test_parse_memory_stat() { let ok = parse_memory_stat(GOOD_MEMORYSTAT_VAL.to_string()).unwrap(); + let raw = ok.raw.clone(); assert_eq!( ok, MemoryStat { @@ -839,6 +1146,7 @@ total_unevictable 81920 total_inactive_file: 1272135680, total_active_file: 2338816000, total_unevictable: 81920, + raw, } ); } diff --git a/src/net_cls.rs b/src/net_cls.rs index c3057f96..fd6d1332 100644 --- a/src/net_cls.rs +++ b/src/net_cls.rs @@ -1,17 +1,21 @@ +// Copyright (c) 2018 Levente Kurusa +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! This module contains the implementation of the `net_cls` cgroup subsystem. //! //! See the Kernel's documentation for more information about this subsystem, found at: //! [Documentation/cgroup-v1/net_cls.txt](https://www.kernel.org/doc/Documentation/cgroup-v1/net_cls.txt) -use std::fs::File; -use std::io::{Read, Write}; +use std::io::Write; use std::path::PathBuf; -use crate::error::*; use crate::error::ErrorKind::*; +use crate::error::*; +use crate::read_u64_from; use crate::{ - ControllIdentifier, ControllerInternal, Controllers, NetworkResources, Resources, - Subsystem, + ControllIdentifier, ControllerInternal, Controllers, NetworkResources, Resources, Subsystem, }; /// A controller that allows controlling the `net_cls` subsystem of a Cgroup. @@ -43,13 +47,9 @@ impl ControllerInternal for NetClsController { // get the resources that apply to this controller let res: &NetworkResources = &res.network; - if res.update_values { - let _ = self.set_class(res.class_id); - if self.get_class()? != res.class_id { - return Err(Error::new(Other)); - } - } - return Ok(()); + update_and_test!(self, set_class, res.class_id, get_class); + + Ok(()) } } @@ -66,26 +66,17 @@ impl<'a> From<&'a Subsystem> for &'a NetClsController { Subsystem::NetCls(c) => c, _ => { assert_eq!(1, 0); - ::std::mem::uninitialized() + let v = std::mem::MaybeUninit::uninit(); + v.assume_init() } } } } } -fn read_u64_from(mut file: File) -> Result { - let mut string = String::new(); - match file.read_to_string(&mut string) { - Ok(_) => string.trim().parse().map_err(|e| Error::with_cause(ParseError, e)), - Err(e) => Err(Error::with_cause(ReadFailed, e)), - } -} - impl NetClsController { - /// Constructs a new `NetClsController` with `oroot` serving as the root of the control group. - pub fn new(oroot: PathBuf) -> Self { - let mut root = oroot; - root.push(Self::controller_type().to_string()); + /// Constructs a new `NetClsController` with `root` serving as the root of the control group. + pub fn new(root: PathBuf) -> Self { Self { base: root.clone(), path: root, @@ -97,13 +88,14 @@ impl NetClsController { self.open_path("net_cls.classid", true) .and_then(|mut file| { let s = format!("{:#08X}", class); - file.write_all(s.as_ref()).map_err(|e| Error::with_cause(WriteFailed, e)) + file.write_all(s.as_ref()) + .map_err(|e| Error::with_cause(WriteFailed, e)) }) } /// Get the network class id of the outgoing packets of the control group's tasks. pub fn get_class(&self) -> Result { self.open_path("net_cls.classid", false) - .and_then(|file| read_u64_from(file)) + .and_then(read_u64_from) } } diff --git a/src/net_prio.rs b/src/net_prio.rs index 6c2ed7bc..4620eb35 100644 --- a/src/net_prio.rs +++ b/src/net_prio.rs @@ -1,18 +1,22 @@ +// Copyright (c) 2018 Levente Kurusa +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! This module contains the implementation of the `net_prio` cgroup subsystem. //! //! See the Kernel's documentation for more information about this subsystem, found at: //! [Documentation/cgroup-v1/net_prio.txt](https://www.kernel.org/doc/Documentation/cgroup-v1/net_prio.txt) use std::collections::HashMap; -use std::fs::File; -use std::io::{BufRead, BufReader, Read, Write}; +use std::io::{BufRead, BufReader, Write}; use std::path::PathBuf; -use crate::error::*; use crate::error::ErrorKind::*; +use crate::error::*; +use crate::read_u64_from; use crate::{ - ControllIdentifier, ControllerInternal, Controllers, NetworkResources, Resources, - Subsystem, + ControllIdentifier, ControllerInternal, Controllers, NetworkResources, Resources, Subsystem, }; /// A controller that allows controlling the `net_prio` subsystem of a Cgroup. @@ -44,10 +48,8 @@ impl ControllerInternal for NetPrioController { // get the resources that apply to this controller let res: &NetworkResources = &res.network; - if res.update_values { - for i in &res.priorities { - let _ = self.set_if_prio(&i.name, i.priority); - } + for i in &res.priorities { + let _ = self.set_if_prio(&i.name, i.priority); } Ok(()) @@ -67,26 +69,17 @@ impl<'a> From<&'a Subsystem> for &'a NetPrioController { Subsystem::NetPrio(c) => c, _ => { assert_eq!(1, 0); - ::std::mem::uninitialized() + let v = std::mem::MaybeUninit::uninit(); + v.assume_init() } } } } } -fn read_u64_from(mut file: File) -> Result { - let mut string = String::new(); - match file.read_to_string(&mut string) { - Ok(_) => string.trim().parse().map_err(|e| Error::with_cause(ParseError, e)), - Err(e) => Err(Error::with_cause(ReadFailed, e)), - } -} - impl NetPrioController { - /// Constructs a new `NetPrioController` with `oroot` serving as the root of the control group. - pub fn new(oroot: PathBuf) -> Self { - let mut root = oroot; - root.push(Self::controller_type().to_string()); + /// Constructs a new `NetPrioController` with `root` serving as the root of the control group. + pub fn new(root: PathBuf) -> Self { Self { base: root.clone(), path: root, @@ -101,6 +94,7 @@ impl NetPrioController { } /// A map of priorities for each network interface. + #[allow(clippy::iter_nth_zero, clippy::unnecessary_unwrap)] pub fn ifpriomap(&self) -> Result> { self.open_path("net_prio.ifpriomap", false) .and_then(|file| { @@ -112,6 +106,7 @@ impl NetPrioController { let mut acc = acc.unwrap(); let l = line.unwrap(); let mut sp = l.split_whitespace(); + let ifname = sp.nth(0); let ifprio = sp.nth(1); if ifname.is_none() || ifprio.is_none() { diff --git a/src/perf_event.rs b/src/perf_event.rs index 4f52b1be..f0e9240e 100644 --- a/src/perf_event.rs +++ b/src/perf_event.rs @@ -1,3 +1,8 @@ +// Copyright (c) 2018 Levente Kurusa +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! This module contains the implementation of the `perf_event` cgroup subsystem. //! //! See the Kernel's documentation for more information about this subsystem, found at: @@ -50,7 +55,8 @@ impl<'a> From<&'a Subsystem> for &'a PerfEventController { Subsystem::PerfEvent(c) => c, _ => { assert_eq!(1, 0); - ::std::mem::uninitialized() + let v = std::mem::MaybeUninit::uninit(); + v.assume_init() } } } @@ -58,10 +64,8 @@ impl<'a> From<&'a Subsystem> for &'a PerfEventController { } impl PerfEventController { - /// Constructs a new `PerfEventController` with `oroot` serving as the root of the control group. - pub fn new(oroot: PathBuf) -> Self { - let mut root = oroot; - root.push(Self::controller_type().to_string()); + /// Constructs a new `PerfEventController` with `root` serving as the root of the control group. + pub fn new(root: PathBuf) -> Self { Self { base: root.clone(), path: root, diff --git a/src/pid.rs b/src/pid.rs index d7f42d2c..b03e3dc4 100644 --- a/src/pid.rs +++ b/src/pid.rs @@ -1,16 +1,23 @@ +// Copyright (c) 2018 Levente Kurusa +// Copyright (c) 2020 Ant Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! This module contains the implementation of the `pids` cgroup subsystem. //! //! See the Kernel's documentation for more information about this subsystem, found at: //! [Documentation/cgroups-v1/pids.txt](https://www.kernel.org/doc/Documentation/cgroup-v1/pids.txt) -use std::fs::File; use std::io::{Read, Write}; use std::path::PathBuf; -use crate::error::*; use crate::error::ErrorKind::*; +use crate::error::*; +use crate::read_u64_from; use crate::{ - ControllIdentifier, ControllerInternal, Controllers, PidResources, Resources, Subsystem, + parse_max_value, ControllIdentifier, ControllerInternal, Controllers, MaxValue, PidResources, + Resources, Subsystem, }; /// A controller that allows controlling the `pids` subsystem of a Cgroup. @@ -18,22 +25,7 @@ use crate::{ pub struct PidController { base: PathBuf, path: PathBuf, -} - -/// The values found in the `pids.max` file in a Cgroup's `pids` subsystem. -#[derive(Eq, PartialEq, Copy, Clone, Debug)] -pub enum PidMax { - /// This value is returned when the text found `pids.max` is `"max"`. - Max, - /// When the value in `pids.max` is a numerical value, they are returned via this enum field. - Value(i64), -} - -impl Default for PidMax { - /// By default, (as per the kernel) `pids.max` should contain `"max"`. - fn default() -> Self { - PidMax::Max - } + v2: bool, } impl ControllerInternal for PidController { @@ -50,21 +42,21 @@ impl ControllerInternal for PidController { &self.base } + fn is_v2(&self) -> bool { + self.v2 + } + fn apply(&self, res: &Resources) -> Result<()> { // get the resources that apply to this controller let pidres: &PidResources = &res.pid; - if pidres.update_values { - // apply pid_max - let _ = self.set_pid_max(pidres.maximum_number_of_processes); - - // now, verify - if self.get_pid_max()? == pidres.maximum_number_of_processes { - return Ok(()); - } else { - return Err(Error::new(Other)); - } - } + // apply pid_max + update_and_test!( + self, + set_pid_max, + pidres.maximum_number_of_processes, + get_pid_max + ); Ok(()) } @@ -89,30 +81,22 @@ impl<'a> From<&'a Subsystem> for &'a PidController { Subsystem::Pid(c) => c, _ => { assert_eq!(1, 0); - ::std::mem::uninitialized() + let v = std::mem::MaybeUninit::uninit(); + v.assume_init() } } } } } -fn read_u64_from(mut file: File) -> Result { - let mut string = String::new(); - match file.read_to_string(&mut string) { - Ok(_) => string.trim().parse().map_err(|e| Error::with_cause(ParseError, e)), - Err(e) => Err(Error::with_cause(ReadFailed, e)), - } -} - impl PidController { - /// Constructors a new `PidController` instance, with `oroot` serving as the controller's root + /// Constructors a new `PidController` instance, with `root` serving as the controller's root /// directory. - pub fn new(oroot: PathBuf) -> Self { - let mut root = oroot; - root.push(Self::controller_type().to_string()); + pub fn new(root: PathBuf, v2: bool) -> Self { Self { base: root.clone(), path: root, + v2, } } @@ -140,19 +124,12 @@ impl PidController { } /// The maximum number of processes that can exist at one time in the control group. - pub fn get_pid_max(&self) -> Result { + pub fn get_pid_max(&self) -> Result { self.open_path("pids.max", false).and_then(|mut file| { let mut string = String::new(); let res = file.read_to_string(&mut string); match res { - Ok(_) => if string.trim() == "max" { - Ok(PidMax::Max) - } else { - match string.trim().parse() { - Ok(val) => Ok(PidMax::Value(val)), - Err(e) => Err(Error::with_cause(ParseError, e)), - } - }, + Ok(_) => parse_max_value(&string), Err(e) => Err(Error::with_cause(ReadFailed, e)), } }) @@ -163,12 +140,9 @@ impl PidController { /// Note that if `get_pid_current()` returns a higher number than what you /// are about to set (`max_pid`), then no processess will be killed. Additonally, attaching /// extra processes to a control group disregards the limit. - pub fn set_pid_max(&self, max_pid: PidMax) -> Result<()> { + pub fn set_pid_max(&self, max_pid: MaxValue) -> Result<()> { self.open_path("pids.max", true).and_then(|mut file| { - let string_to_write = match max_pid { - PidMax::Max => "max".to_string(), - PidMax::Value(num) => num.to_string(), - }; + let string_to_write = max_pid.to_string(); match file.write_all(string_to_write.as_ref()) { Ok(_) => Ok(()), Err(e) => Err(Error::with_cause(WriteFailed, e)), diff --git a/src/rdma.rs b/src/rdma.rs index fb9dcc61..1ee1fc87 100644 --- a/src/rdma.rs +++ b/src/rdma.rs @@ -1,14 +1,19 @@ +// Copyright (c) 2018 Levente Kurusa +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! This module contains the implementation of the `rdma` cgroup subsystem. //! //! See the Kernel's documentation for more information about this subsystem, found at: //! [Documentation/cgroup-v1/rdma.txt](https://www.kernel.org/doc/Documentation/cgroup-v1/rdma.txt) -use std::fs::File; -use std::io::{Read, Write}; +use std::io::Write; use std::path::PathBuf; -use crate::error::*; use crate::error::ErrorKind::*; +use crate::error::*; +use crate::read_string_from; use crate::{ControllIdentifier, ControllerInternal, Controllers, Resources, Subsystem}; /// A controller that allows controlling the `rdma` subsystem of a Cgroup. @@ -53,26 +58,17 @@ impl<'a> From<&'a Subsystem> for &'a RdmaController { Subsystem::Rdma(c) => c, _ => { assert_eq!(1, 0); - ::std::mem::uninitialized() + let v = std::mem::MaybeUninit::uninit(); + v.assume_init() } } } } } -fn read_string_from(mut file: File) -> Result { - let mut string = String::new(); - match file.read_to_string(&mut string) { - Ok(_) => Ok(string.trim().to_string()), - Err(e) => Err(Error::with_cause(ReadFailed, e)), - } -} - impl RdmaController { - /// Constructs a new `RdmaController` with `oroot` serving as the root of the control group. - pub fn new(oroot: PathBuf) -> Self { - let mut root = oroot; - root.push(Self::controller_type().to_string()); + /// Constructs a new `RdmaController` with `root` serving as the root of the control group. + pub fn new(root: PathBuf) -> Self { Self { base: root.clone(), path: root, diff --git a/src/systemd.rs b/src/systemd.rs new file mode 100644 index 00000000..8eb3db26 --- /dev/null +++ b/src/systemd.rs @@ -0,0 +1,72 @@ +// Copyright (c) 2020 Ant Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + +//! This module contains the implementation of the `systemd` cgroup subsystem. +//! +use std::path::PathBuf; + +use crate::error::*; + +use crate::{ControllIdentifier, ControllerInternal, Controllers, Resources, Subsystem}; + +/// A controller that allows controlling the `systemd` subsystem of a Cgroup. +/// +#[derive(Debug, Clone)] +pub struct SystemdController { + base: PathBuf, + path: PathBuf, + v2: bool, +} + +impl ControllerInternal for SystemdController { + fn control_type(&self) -> Controllers { + Controllers::Systemd + } + fn get_path(&self) -> &PathBuf { + &self.path + } + fn get_path_mut(&mut self) -> &mut PathBuf { + &mut self.path + } + fn get_base(&self) -> &PathBuf { + &self.base + } + + fn apply(&self, _res: &Resources) -> Result<()> { + Ok(()) + } +} + +impl ControllIdentifier for SystemdController { + fn controller_type() -> Controllers { + Controllers::Systemd + } +} + +impl<'a> From<&'a Subsystem> for &'a SystemdController { + fn from(sub: &'a Subsystem) -> &'a SystemdController { + unsafe { + match sub { + Subsystem::Systemd(c) => c, + _ => { + assert_eq!(1, 0); + let v = std::mem::MaybeUninit::uninit(); + v.assume_init() + } + } + } + } +} + +impl SystemdController { + /// Constructs a new `SystemdController` with `root` serving as the root of the control group. + pub fn new(root: PathBuf, v2: bool) -> Self { + Self { + base: root.clone(), + path: root, + v2, + } + } +} diff --git a/tests/builder.rs b/tests/builder.rs index 45c11128..7e0bed22 100644 --- a/tests/builder.rs +++ b/tests/builder.rs @@ -1,22 +1,28 @@ +// Copyright (c) 2018 Levente Kurusa +// Copyright (c) 2020 And Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! Some simple tests covering the builder pattern for control groups. -use cgroups::*; -use cgroups::cpu::*; -use cgroups::devices::*; -use cgroups::pid::*; -use cgroups::memory::*; -use cgroups::net_cls::*; -use cgroups::hugetlb::*; -use cgroups::blkio::*; -use cgroups::cgroup_builder::*; +use cgroups_rs::blkio::*; +use cgroups_rs::cgroup_builder::*; +use cgroups_rs::cpu::*; +use cgroups_rs::devices::*; +use cgroups_rs::hugetlb::*; +use cgroups_rs::memory::*; +use cgroups_rs::net_cls::*; +use cgroups_rs::pid::*; +use cgroups_rs::*; #[test] pub fn test_cpu_res_build() { - let v1 = crate::hierarchies::V1::new(); - let cg: Cgroup = CgroupBuilder::new("test_cpu_res_build", &v1) + let h = cgroups_rs::hierarchies::auto(); + let cg: Cgroup = CgroupBuilder::new("test_cpu_res_build") .cpu() - .shares(85) - .done() - .build(); + .shares(85) + .done() + .build(h); { let cpu: &CpuController = cg.controller_of().unwrap(); @@ -24,121 +30,135 @@ pub fn test_cpu_res_build() { assert_eq!(cpu.shares().unwrap(), 85); } - cg.delete(); + cg.delete().unwrap(); } #[test] pub fn test_memory_res_build() { - let v1 = crate::hierarchies::V1::new(); - let cg: Cgroup = CgroupBuilder::new("test_memory_res_build", &v1) + let h = cgroups_rs::hierarchies::auto(); + let cg: Cgroup = CgroupBuilder::new("test_memory_res_build") .memory() - .kernel_memory_limit(128 * 1024 * 1024) - .swappiness(70) - .memory_hard_limit(1024 * 1024 * 1024) - .done() - .build(); + .kernel_memory_limit(128 * 1024 * 1024) + .swappiness(70) + .memory_hard_limit(1024 * 1024 * 1024) + .done() + .build(h); { let c: &MemController = cg.controller_of().unwrap(); - assert_eq!(c.kmem_stat().limit_in_bytes, 128 * 1024 * 1024); - assert_eq!(c.memory_stat().swappiness, 70); + if !c.v2() { + assert_eq!(c.kmem_stat().limit_in_bytes, 128 * 1024 * 1024); + assert_eq!(c.memory_stat().swappiness, 70); + } assert_eq!(c.memory_stat().limit_in_bytes, 1024 * 1024 * 1024); } - cg.delete(); + cg.delete().unwrap(); } #[test] pub fn test_pid_res_build() { - let v1 = crate::hierarchies::V1::new(); - let cg: Cgroup = CgroupBuilder::new("test_pid_res_build", &v1) + let h = cgroups_rs::hierarchies::auto(); + let cg: Cgroup = CgroupBuilder::new("test_pid_res_build") .pid() - .maximum_number_of_processes(PidMax::Value(123)) - .done() - .build(); + .maximum_number_of_processes(MaxValue::Value(123)) + .done() + .build(h); { let c: &PidController = cg.controller_of().unwrap(); assert!(c.get_pid_max().is_ok()); - assert_eq!(c.get_pid_max().unwrap(), PidMax::Value(123)); + assert_eq!(c.get_pid_max().unwrap(), MaxValue::Value(123)); } - cg.delete(); + cg.delete().unwrap(); } #[test] #[ignore] // ignore this test for now, not sure why my kernel doesn't like it pub fn test_devices_res_build() { - let v1 = crate::hierarchies::V1::new(); - let cg: Cgroup = CgroupBuilder::new("test_devices_res_build", &v1) + let h = cgroups_rs::hierarchies::auto(); + let cg: Cgroup = CgroupBuilder::new("test_devices_res_build") .devices() - .device(1, 6, DeviceType::Char, true, - vec![DevicePermissions::Read]) - .done() - .build(); + .device(1, 6, DeviceType::Char, true, vec![DevicePermissions::Read]) + .done() + .build(h); { let c: &DevicesController = cg.controller_of().unwrap(); assert!(c.allowed_devices().is_ok()); - assert_eq!(c.allowed_devices().unwrap(), vec![ - DeviceResource { - allow: true, - devtype: DeviceType::Char, - major: 1, - minor: 6, - access: vec![DevicePermissions::Read], - } - ]); + assert_eq!( + c.allowed_devices().unwrap(), + vec![DeviceResource { + allow: true, + devtype: DeviceType::Char, + major: 1, + minor: 6, + access: vec![DevicePermissions::Read], + }] + ); } - cg.delete(); + cg.delete().unwrap(); } #[test] pub fn test_network_res_build() { - let v1 = crate::hierarchies::V1::new(); - let cg: Cgroup = CgroupBuilder::new("test_network_res_build", &v1) + let h = cgroups_rs::hierarchies::auto(); + if h.v2() { + // FIXME add cases for v2 + return; + } + let cg: Cgroup = CgroupBuilder::new("test_network_res_build") .network() - .class_id(1337) - .done() - .build(); + .class_id(1337) + .done() + .build(h); { let c: &NetClsController = cg.controller_of().unwrap(); assert!(c.get_class().is_ok()); assert_eq!(c.get_class().unwrap(), 1337); } - cg.delete(); + cg.delete().unwrap(); } #[test] pub fn test_hugepages_res_build() { - let v1 = crate::hierarchies::V1::new(); - let cg: Cgroup = CgroupBuilder::new("test_hugepages_res_build", &v1) + let h = cgroups_rs::hierarchies::auto(); + if h.v2() { + // FIXME add cases for v2 + return; + } + let cg: Cgroup = CgroupBuilder::new("test_hugepages_res_build") .hugepages() - .limit("2MB".to_string(), 4 * 2 * 1024 * 1024) - .done() - .build(); + .limit("2MB".to_string(), 4 * 2 * 1024 * 1024) + .done() + .build(h); { let c: &HugeTlbController = cg.controller_of().unwrap(); assert!(c.limit_in_bytes(&"2MB".to_string()).is_ok()); - assert_eq!(c.limit_in_bytes(&"2MB".to_string()).unwrap(), 4 * 2 * 1024 * 1024); + assert_eq!( + c.limit_in_bytes(&"2MB".to_string()).unwrap(), + 4 * 2 * 1024 * 1024 + ); } - cg.delete(); + cg.delete().unwrap(); } #[test] +#[ignore] // high version kernel not support `blkio.weight` pub fn test_blkio_res_build() { - let v1 = crate::hierarchies::V1::new(); - let cg: Cgroup = CgroupBuilder::new("test_blkio_res_build", &v1) + let h = cgroups_rs::hierarchies::auto(); + let cg: Cgroup = CgroupBuilder::new("test_blkio_res_build") .blkio() - .weight(100) - .done() - .build(); + .weight(100) + .done() + .build(h); { let c: &BlkIoController = cg.controller_of().unwrap(); assert_eq!(c.blkio().weight, 100); } - cg.delete(); + cg.delete().unwrap(); } diff --git a/tests/cgroup.rs b/tests/cgroup.rs index 8b15d9a4..563168b1 100644 --- a/tests/cgroup.rs +++ b/tests/cgroup.rs @@ -1,14 +1,26 @@ +// Copyright (c) 2018 Levente Kurusa +// Copyright (c) 2020 And Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! Simple unit tests about the control groups system. -use cgroups::{Cgroup, CgroupPid}; +use cgroups_rs::memory::MemController; +use cgroups_rs::Controller; +use cgroups_rs::{Cgroup, CgroupPid, Subsystem}; #[test] fn test_tasks_iterator() { - let hier = cgroups::hierarchies::V1::new(); + let h = cgroups_rs::hierarchies::auto(); let pid = libc::pid_t::from(nix::unistd::getpid()) as u64; - let cg = Cgroup::new(&hier, String::from("test_tasks_iterator")); + let cg = Cgroup::new(h, String::from("test_tasks_iterator")); { // Add a task to the control group. - cg.add_task(CgroupPid::from(pid)); + cg.add_task(CgroupPid::from(pid)).unwrap(); + + use std::{thread, time}; + thread::sleep(time::Duration::from_millis(100)); + let mut tasks = cg.tasks().into_iter(); // Verify that the task is indeed in the control group assert_eq!(tasks.next(), Some(CgroupPid::from(pid))); @@ -21,5 +33,73 @@ fn test_tasks_iterator() { // Verify that it was indeed removed. assert_eq!(tasks.next(), None); } - cg.delete(); + cg.delete().unwrap(); +} + +#[test] +fn test_cgroup_with_relative_paths() { + if cgroups_rs::hierarchies::is_cgroup2_unified_mode() { + return; + } + let h = cgroups_rs::hierarchies::auto(); + let cgroup_root = h.root(); + let cgroup_name = "test_cgroup_with_relative_paths"; + + let cg = Cgroup::load(h, String::from(cgroup_name)); + { + let subsystems = cg.subsystems(); + subsystems.iter().for_each(|sub| match sub { + Subsystem::Pid(c) => { + let cgroup_path = c.path().to_str().unwrap(); + let relative_path = "/pids/"; + // cgroup_path = cgroup_root + relative_path + cgroup_name + assert_eq!( + cgroup_path, + format!( + "{}{}{}", + cgroup_root.to_str().unwrap(), + relative_path, + cgroup_name + ) + ); + } + Subsystem::Mem(c) => { + let cgroup_path = c.path().to_str().unwrap(); + // cgroup_path = cgroup_root + relative_path + cgroup_name + assert_eq!( + cgroup_path, + format!("{}/memory/{}", cgroup_root.to_str().unwrap(), cgroup_name) + ); + } + _ => {} + }); + } + cg.delete().unwrap(); +} + +#[test] +fn test_cgroup_v2() { + if !cgroups_rs::hierarchies::is_cgroup2_unified_mode() { + return; + } + let h = cgroups_rs::hierarchies::auto(); + let cg = Cgroup::new(h, String::from("test_v2")); + + let mem_controller: &MemController = cg.controller_of().unwrap(); + let (mem, swp, rev) = (4 * 1024 * 1000, 2 * 1024 * 1000, 1024 * 1000); + + mem_controller.set_limit(mem).unwrap(); + mem_controller.set_memswap_limit(swp).unwrap(); + mem_controller.set_soft_limit(rev).unwrap(); + + let memory_stat = mem_controller.memory_stat(); + println!("memory_stat {:?}", memory_stat); + assert_eq!(mem, memory_stat.limit_in_bytes); + assert_eq!(rev, memory_stat.soft_limit_in_bytes); + + let memswap = mem_controller.memswap(); + println!("memswap {:?}", memswap); + assert_eq!(swp, memswap.limit_in_bytes); + + cg.delete().unwrap(); } diff --git a/tests/cpu.rs b/tests/cpu.rs new file mode 100644 index 00000000..959474e7 --- /dev/null +++ b/tests/cpu.rs @@ -0,0 +1,61 @@ +// Copyright (c) 2020 And Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + +//! Simple unit tests about the CPU control groups system. +use cgroups_rs::cpu::CpuController; +use cgroups_rs::Cgroup; + +#[test] +fn test_cfs_quota_and_periods() { + let h = cgroups_rs::hierarchies::auto(); + let cg = Cgroup::new(h, String::from("test_cfs_quota_and_periods")); + + let cpu_controller: &CpuController = cg.controller_of().unwrap(); + + let current_quota = cpu_controller.cfs_quota().unwrap(); + let current_peroid = cpu_controller.cfs_period().unwrap(); + + // verify default value + // The default is “max 100000”. + assert_eq!(-1, current_quota); + assert_eq!(100000, current_peroid); + + // case 1 set quota + let _ = cpu_controller.set_cfs_quota(2000); + + let current_quota = cpu_controller.cfs_quota().unwrap(); + let current_peroid = cpu_controller.cfs_period().unwrap(); + assert_eq!(2000, current_quota); + assert_eq!(100000, current_peroid); + + // case 2 set period + cpu_controller.set_cfs_period(1000000).unwrap(); + let current_quota = cpu_controller.cfs_quota().unwrap(); + let current_peroid = cpu_controller.cfs_period().unwrap(); + assert_eq!(2000, current_quota); + assert_eq!(1000000, current_peroid); + + // case 3 set both quota and period + cpu_controller + .set_cfs_quota_and_period(Some(5000), Some(100000)) + .unwrap(); + + let current_quota = cpu_controller.cfs_quota().unwrap(); + let current_peroid = cpu_controller.cfs_period().unwrap(); + assert_eq!(5000, current_quota); + assert_eq!(100000, current_peroid); + + // case 4 set both quota and period, set quota to -1 + cpu_controller + .set_cfs_quota_and_period(Some(-1), None) + .unwrap(); + + let current_quota = cpu_controller.cfs_quota().unwrap(); + let current_peroid = cpu_controller.cfs_period().unwrap(); + assert_eq!(-1, current_quota); + assert_eq!(100000, current_peroid); + + cg.delete().unwrap(); +} diff --git a/tests/cpuset.rs b/tests/cpuset.rs index c7a8fb0c..44e525ed 100644 --- a/tests/cpuset.rs +++ b/tests/cpuset.rs @@ -1,11 +1,19 @@ -use cgroups::cpuset::CpuSetController; -use cgroups::error::ErrorKind; -use cgroups::Cgroup; +// Copyright (c) 2018 Levente Kurusa +// Copyright (c) 2020 And Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + +use cgroups_rs::cpuset::CpuSetController; +use cgroups_rs::error::ErrorKind; +use cgroups_rs::{Cgroup, CgroupPid}; + +use std::fs; #[test] fn test_cpuset_memory_pressure_root_cg() { - let hier = cgroups::hierarchies::V1::new(); - let cg = Cgroup::new(&hier, String::from("test_cpuset_memory_pressure_root_cg")); + let h = cgroups_rs::hierarchies::auto(); + let cg = Cgroup::new(h, String::from("test_cpuset_memory_pressure_root_cg")); { let cpuset: &CpuSetController = cg.controller_of().unwrap(); @@ -13,5 +21,72 @@ fn test_cpuset_memory_pressure_root_cg() { let res = cpuset.set_enable_memory_pressure(true); assert_eq!(res.unwrap_err().kind(), &ErrorKind::InvalidOperation); } - cg.delete(); + cg.delete().unwrap(); +} + +#[test] +fn test_cpuset_set_cpus() { + let h = cgroups_rs::hierarchies::auto(); + let cg = Cgroup::new(h, String::from("test_cpuset_set_cpus")); + { + let cpuset: &CpuSetController = cg.controller_of().unwrap(); + + let set = cpuset.cpuset(); + if cg.v2() { + assert_eq!(0, set.cpus.len()); + } else { + // for cgroup v1, cpuset is copied from parent. + assert_eq!(true, !set.cpus.is_empty()); + } + + // 0 + let r = cpuset.set_cpus("0"); + assert_eq!(true, r.is_ok()); + + let set = cpuset.cpuset(); + assert_eq!(1, set.cpus.len()); + assert_eq!((0, 0), set.cpus[0]); + + // all cpus in system + let cpus = fs::read_to_string("/sys/fs/cgroup/cpuset.cpus.effective").unwrap_or_default(); + let cpus = cpus.trim(); + if !cpus.is_empty() { + let r = cpuset.set_cpus(&cpus); + assert_eq!(true, r.is_ok()); + let set = cpuset.cpuset(); + assert_eq!(1, set.cpus.len()); + assert_eq!(format!("{}-{}", set.cpus[0].0, set.cpus[0].1), cpus); + } + } + cg.delete().unwrap(); +} + +#[test] +fn test_cpuset_set_cpus_add_task() { + let h = cgroups_rs::hierarchies::auto(); + let cg = Cgroup::new(h, String::from("test_cpuset_set_cpus_add_task/sub-dir")); + + let cpuset: &CpuSetController = cg.controller_of().unwrap(); + let set = cpuset.cpuset(); + if cg.v2() { + assert_eq!(0, set.cpus.len()); + } else { + // for cgroup v1, cpuset is copied from parent. + assert_eq!(true, !set.cpus.is_empty()); + } + + // Add a task to the control group. + let pid_i = libc::pid_t::from(nix::unistd::getpid()) as u64; + let _ = cg.add_task(CgroupPid::from(pid_i)); + let tasks = cg.tasks(); + assert_eq!(true, !tasks.is_empty()); + println!("tasks after added: {:?}", tasks); + + // remove task + let _ = cg.remove_task(CgroupPid::from(pid_i)); + let tasks = cg.tasks(); + println!("tasks after deleted: {:?}", tasks); + assert_eq!(0, tasks.len()); + + cg.delete().unwrap(); } diff --git a/tests/devices.rs b/tests/devices.rs index 2fbb3058..90c1bef2 100644 --- a/tests/devices.rs +++ b/tests/devices.rs @@ -1,26 +1,39 @@ +// Copyright (c) 2018 Levente Kurusa +// Copyright (c) 2020 And Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! Integration tests about the devices subsystem -use cgroups::devices::{DevicePermissions, DeviceType, DevicesController}; -use cgroups::{Cgroup, DeviceResource}; +use cgroups_rs::devices::{DevicePermissions, DeviceType, DevicesController}; +use cgroups_rs::{Cgroup, DeviceResource}; #[test] fn test_devices_parsing() { - let hier = cgroups::hierarchies::V1::new(); - let cg = Cgroup::new(&hier, String::from("test_devices_parsing")); + // now only v2 + if cgroups_rs::hierarchies::is_cgroup2_unified_mode() { + return; + } + + let h = cgroups_rs::hierarchies::auto(); + let cg = Cgroup::new(h, String::from("test_devices_parsing")); { let devices: &DevicesController = cg.controller_of().unwrap(); // Deny access to all devices first - devices.deny_device( - DeviceType::All, - -1, - -1, - &vec![ - DevicePermissions::Read, - DevicePermissions::Write, - DevicePermissions::MkNod, - ], - ); + devices + .deny_device( + DeviceType::All, + -1, + -1, + &[ + DevicePermissions::Read, + DevicePermissions::Write, + DevicePermissions::MkNod, + ], + ) + .unwrap(); // Acquire the list of allowed devices after we denied all let allowed_devices = devices.allowed_devices(); // Verify that there are no devices that we can access. @@ -28,7 +41,9 @@ fn test_devices_parsing() { assert_eq!(allowed_devices.unwrap(), Vec::new()); // Now add mknod access to /dev/null device - devices.allow_device(DeviceType::Char, 1, 3, &vec![DevicePermissions::MkNod]); + devices + .allow_device(DeviceType::Char, 1, 3, &[DevicePermissions::MkNod]) + .unwrap(); let allowed_devices = devices.allowed_devices(); assert!(allowed_devices.is_ok()); let allowed_devices = allowed_devices.unwrap(); @@ -45,12 +60,14 @@ fn test_devices_parsing() { ); // Now deny, this device explicitly. - devices.deny_device(DeviceType::Char, 1, 3, &DevicePermissions::all()); + devices + .deny_device(DeviceType::Char, 1, 3, &DevicePermissions::all()) + .unwrap(); // Finally, check that. let allowed_devices = devices.allowed_devices(); // Verify that there are no devices that we can access. assert!(allowed_devices.is_ok()); assert_eq!(allowed_devices.unwrap(), Vec::new()); } - cg.delete(); + cg.delete().unwrap(); } diff --git a/tests/hugetlb.rs b/tests/hugetlb.rs new file mode 100644 index 00000000..cc33e286 --- /dev/null +++ b/tests/hugetlb.rs @@ -0,0 +1,44 @@ +// Copyright (c) 2020 And Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + +//! Integration tests about the hugetlb subsystem +use cgroups_rs::error::*; +use cgroups_rs::hugetlb::{self, HugeTlbController}; +use cgroups_rs::Cgroup; +use std::fs; + +#[test] +fn test_hugetlb_sizes() { + // now only v2 + if cgroups_rs::hierarchies::is_cgroup2_unified_mode() { + return; + } + + let h = cgroups_rs::hierarchies::auto(); + let cg = Cgroup::new(h, String::from("test_hugetlb_sizes")); + { + let hugetlb_controller: &HugeTlbController = cg.controller_of().unwrap(); + let _ = hugetlb_controller.get_sizes(); + + // test sizes count + let sizes = hugetlb_controller.get_sizes(); + let sizes_count = fs::read_dir(hugetlb::HUGEPAGESIZE_DIR).unwrap().count(); + assert_eq!(sizes.len(), sizes_count); + + for size in sizes { + let supported = hugetlb_controller.size_supported(&size); + assert_eq!(supported, true); + assert_no_error(hugetlb_controller.failcnt(&size)); + assert_no_error(hugetlb_controller.limit_in_bytes(&size)); + assert_no_error(hugetlb_controller.usage_in_bytes(&size)); + assert_no_error(hugetlb_controller.max_usage_in_bytes(&size)); + } + } + cg.delete().unwrap(); +} + +fn assert_no_error(r: Result) { + assert_eq!(!r.is_err(), true) +} diff --git a/tests/memory.rs b/tests/memory.rs new file mode 100644 index 00000000..9ab2e018 --- /dev/null +++ b/tests/memory.rs @@ -0,0 +1,91 @@ +// Copyright (c) 2020 And Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + +//! Integration tests about the hugetlb subsystem +use cgroups_rs::memory::{MemController, SetMemory}; +use cgroups_rs::Controller; +use cgroups_rs::{Cgroup, MaxValue}; + +#[test] +fn test_disable_oom_killer() { + let h = cgroups_rs::hierarchies::auto(); + let cg = Cgroup::new(h, String::from("test_disable_oom_killer")); + { + let mem_controller: &MemController = cg.controller_of().unwrap(); + + // before disable + let m = mem_controller.memory_stat(); + assert_eq!(m.oom_control.oom_kill_disable, false); + + // now only v1 + if !mem_controller.v2() { + // disable oom killer + let r = mem_controller.disable_oom_killer(); + assert_eq!(r.is_err(), false); + + // after disable + let m = mem_controller.memory_stat(); + assert_eq!(m.oom_control.oom_kill_disable, true); + } + } + cg.delete().unwrap(); +} + +#[test] +fn set_mem_v2() { + let h = cgroups_rs::hierarchies::auto(); + if !h.v2() { + return; + } + + let cg = Cgroup::new(h, String::from("set_mem_v2")); + { + let mem_controller: &MemController = cg.controller_of().unwrap(); + + // before disable + let m = mem_controller.get_mem().unwrap(); + // case 1: get default value + assert_eq!(m.low, Some(MaxValue::Value(0))); + assert_eq!(m.min, Some(MaxValue::Value(0))); + assert_eq!(m.high, Some(MaxValue::Max)); + assert_eq!(m.max, Some(MaxValue::Max)); + + // case 2: set parts + let m = SetMemory { + low: Some(MaxValue::Value(1024 * 1024 * 2)), + high: Some(MaxValue::Value(1024 * 1024 * 1024 * 2)), + min: Some(MaxValue::Value(1024 * 1024 * 3)), + max: None, + }; + let r = mem_controller.set_mem(m); + assert_eq!(true, r.is_ok()); + + let m = mem_controller.get_mem().unwrap(); + // get + assert_eq!(m.low, Some(MaxValue::Value(1024 * 1024 * 2))); + assert_eq!(m.min, Some(MaxValue::Value(1024 * 1024 * 3))); + assert_eq!(m.high, Some(MaxValue::Value(1024 * 1024 * 1024 * 2))); + assert_eq!(m.max, Some(MaxValue::Max)); + + // case 3: set parts + let m = SetMemory { + max: Some(MaxValue::Value(1024 * 1024 * 1024 * 2)), + min: Some(MaxValue::Value(1024 * 1024 * 4)), + high: Some(MaxValue::Max), + low: None, + }; + let r = mem_controller.set_mem(m); + assert_eq!(true, r.is_ok()); + + let m = mem_controller.get_mem().unwrap(); + // get + assert_eq!(m.low, Some(MaxValue::Value(1024 * 1024 * 2))); + assert_eq!(m.min, Some(MaxValue::Value(1024 * 1024 * 4))); + assert_eq!(m.max, Some(MaxValue::Value(1024 * 1024 * 1024 * 2))); + assert_eq!(m.high, Some(MaxValue::Max)); + } + + cg.delete().unwrap(); +} diff --git a/tests/pids.rs b/tests/pids.rs index 4c91e54b..27a2de79 100644 --- a/tests/pids.rs +++ b/tests/pids.rs @@ -1,73 +1,77 @@ +// Copyright (c) 2018 Levente Kurusa +// Copyright (c) 2020 And Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! Integration tests about the pids subsystem -use cgroups::pid::{PidController, PidMax}; -use cgroups::Controller; -use cgroups::{Cgroup, CgroupPid, PidResources, Resources}; +use cgroups_rs::pid::PidController; +use cgroups_rs::Controller; +use cgroups_rs::{Cgroup, MaxValue}; use nix::sys::wait::{waitpid, WaitStatus}; -use nix::unistd::{fork, ForkResult, Pid}; +use nix::unistd::{fork, ForkResult}; use libc::pid_t; -use std::thread; - #[test] fn create_and_delete_cgroup() { - let hier = cgroups::hierarchies::V1::new(); - let cg = Cgroup::new(&hier, String::from("create_and_delete_cgroup")); + let h = cgroups_rs::hierarchies::auto(); + let cg = Cgroup::new(h, String::from("create_and_delete_cgroup")); { let pidcontroller: &PidController = cg.controller_of().unwrap(); - pidcontroller.set_pid_max(PidMax::Value(1337)); + pidcontroller.set_pid_max(MaxValue::Value(1337)).unwrap(); let max = pidcontroller.get_pid_max(); assert!(max.is_ok()); - assert_eq!(max.unwrap(), PidMax::Value(1337)); + assert_eq!(max.unwrap(), MaxValue::Value(1337)); } - cg.delete(); + cg.delete().unwrap(); } #[test] fn test_pids_current_is_zero() { - let hier = cgroups::hierarchies::V1::new(); - let cg = Cgroup::new(&hier, String::from("test_pids_current_is_zero")); + let h = cgroups_rs::hierarchies::auto(); + let cg = Cgroup::new(h, String::from("test_pids_current_is_zero")); { let pidcontroller: &PidController = cg.controller_of().unwrap(); let current = pidcontroller.get_pid_current(); assert_eq!(current.unwrap(), 0); } - cg.delete(); + cg.delete().unwrap(); } #[test] fn test_pids_events_is_zero() { - let hier = cgroups::hierarchies::V1::new(); - let cg = Cgroup::new(&hier, String::from("test_pids_events_is_zero")); + let h = cgroups_rs::hierarchies::auto(); + let cg = Cgroup::new(h, String::from("test_pids_events_is_zero")); { let pidcontroller: &PidController = cg.controller_of().unwrap(); let events = pidcontroller.get_pid_events(); assert!(events.is_ok()); assert_eq!(events.unwrap(), 0); } - cg.delete(); + cg.delete().unwrap(); } #[test] fn test_pid_events_is_not_zero() { - let hier = cgroups::hierarchies::V1::new(); - let cg = Cgroup::new(&hier, String::from("test_pid_events_is_not_zero")); + let h = cgroups_rs::hierarchies::auto(); + let cg = Cgroup::new(h, String::from("test_pid_events_is_not_zero")); { let pids: &PidController = cg.controller_of().unwrap(); let before = pids.get_pid_events(); let before = before.unwrap(); - match fork() { + match unsafe { fork() } { Ok(ForkResult::Parent { child, .. }) => { // move the process into the control group - pids.add_task(&(pid_t::from(child) as u64).into()); + let _ = pids.add_task(&(pid_t::from(child) as u64).into()); println!("added task to cg: {:?}", child); // Set limit to one - pids.set_pid_max(PidMax::Value(1)); - println!("err = {:?}", pids.get_pid_max()); + let _ = pids.set_pid_max(MaxValue::Value(1)); + println!("current pid.max = {:?}", pids.get_pid_max()); // wait on the child let res = waitpid(child, None); @@ -84,8 +88,8 @@ fn test_pid_events_is_not_zero() { } Ok(ForkResult::Child) => loop { let pids_max = pids.get_pid_max(); - if pids_max.is_ok() && pids_max.unwrap() == PidMax::Value(1) { - if let Err(_) = fork() { + if pids_max.is_ok() && pids_max.unwrap() == MaxValue::Value(1) { + if unsafe { fork() }.is_err() { unsafe { libc::exit(0) }; } else { unsafe { libc::exit(1) }; @@ -95,5 +99,5 @@ fn test_pid_events_is_not_zero() { Err(_) => panic!("failed to fork"), } } - cg.delete(); + cg.delete().unwrap(); } diff --git a/tests/resources.rs b/tests/resources.rs index c612df12..056349ee 100644 --- a/tests/resources.rs +++ b/tests/resources.rs @@ -1,26 +1,31 @@ +// Copyright (c) 2018 Levente Kurusa +// Copyright (c) 2020 And Group +// +// SPDX-License-Identifier: Apache-2.0 or MIT +// + //! Integration test about setting resources using `apply()` -use cgroups::pid::{PidController, PidMax}; -use cgroups::{Cgroup, PidResources, Resources}; +use cgroups_rs::pid::PidController; +use cgroups_rs::{Cgroup, MaxValue, PidResources, Resources}; #[test] fn pid_resources() { - let hier = cgroups::hierarchies::V1::new(); - let cg = Cgroup::new(&hier, String::from("pid_resources")); + let h = cgroups_rs::hierarchies::auto(); + let cg = Cgroup::new(h, String::from("pid_resources")); { let res = Resources { pid: PidResources { - update_values: true, - maximum_number_of_processes: PidMax::Value(512), + maximum_number_of_processes: Some(MaxValue::Value(512)), }, ..Default::default() }; - cg.apply(&res); + cg.apply(&res).unwrap(); // verify let pidcontroller: &PidController = cg.controller_of().unwrap(); let pid_max = pidcontroller.get_pid_max(); assert_eq!(pid_max.is_ok(), true); - assert_eq!(pid_max.unwrap(), PidMax::Value(512)); + assert_eq!(pid_max.unwrap(), MaxValue::Value(512)); } - cg.delete(); + cg.delete().unwrap(); } diff --git a/tools/create_cgroup.sh b/tools/create_cgroup.sh index 49dbfa0c..96407ad3 100755 --- a/tools/create_cgroup.sh +++ b/tools/create_cgroup.sh @@ -1,4 +1,9 @@ #!/bin/sh +# +# Copyright (c) 2018 Levente Kurusa +# +# SPDX-License-Identifier: Apache-2.0 or MIT +# CONTROL_GROUPS=`cargo test -- --list 2>/dev/null | egrep 'test$' | egrep -v '^src' | cut -d':' -f1` diff --git a/tools/delete_cgroup.sh b/tools/delete_cgroup.sh index f3f83a3f..e728cb56 100755 --- a/tools/delete_cgroup.sh +++ b/tools/delete_cgroup.sh @@ -1,4 +1,9 @@ #!/bin/sh +# +# Copyright (c) 2018 Levente Kurusa +# +# SPDX-License-Identifier: Apache-2.0 or MIT +# CONTROL_GROUPS=`cargo test -- --list 2>/dev/null | egrep 'test$' | egrep -v '^src' | cut -d':' -f1`