sys-lib 0.5.0

System facts for Rust services: OS, CPU, memory, disks, network, host, this process, virtualization, power, and device identity, container-aware by default, plus build info, instance info, and a process guard. No subprocesses, zero third-party dependencies.
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
//! Memory on Linux: `/proc/meminfo` for the host, the cgroup's files for the
//! limit and the usage counted against it (E-1 to E-11), `/proc/vmstat` for
//! paging, and the huge-page files.

use core::ops::ControlFlow;

use super::{
    Root,
    cgroup::{Cgroups, Controller},
};
use crate::{
    Bytes, Error, Result,
    mem::{
        EventScope, HugePages, LimitSource, MemRaw, MemWant, MemoryEvents, MemoryLimit,
        MemorySource, PagingRaw, TransparentHugePages,
    },
    parse::procfs::{self, Meminfo, space_value as space_kv_value},
};

/// cgroup v1 writes "no limit" as the largest page-aligned `i64`
/// (`9223372036854771712` on 4 KiB pages, E-3); anything this large is none.
const V1_UNLIMITED: u64 = 1 << 62;

/// The cgroup memory files of this process, as paths, built once so a reading
/// formats nothing.
#[derive(Clone, Debug, PartialEq, Eq)]
pub(crate) struct MemPaths {
    v2: bool,
    /// v2: `memory.max` of the cgroup and every ancestor.
    max: Vec<String>,
    /// v2: `memory.swap.max` of the cgroup and every ancestor.
    swap_max: Vec<String>,
    /// v2 `memory.high`, v1 `memory.soft_limit_in_bytes`.
    soft: String,
    /// v2 `memory.current`, v1 `memory.usage_in_bytes`.
    current: String,
    /// `memory.stat` (v1: also `hierarchical_memory_limit`).
    stat: String,
    /// v1 `memory.memsw.limit_in_bytes`.
    memsw: String,
    /// v2 `memory.events`, v1 `memory.oom_control`.
    events: String,
}

impl MemPaths {
    pub(crate) fn new(cgroups: &Cgroups) -> Option<MemPaths> {
        let Controller { dir, v2 } = cgroups.memory.as_ref()?;
        let leaf = &dir.leaf;
        let each = |file: &str| dir.ancestors().map(|d| format!("{d}/{file}")).collect();
        Some(if *v2 {
            MemPaths {
                v2: true,
                max: each("memory.max"),
                swap_max: each("memory.swap.max"),
                soft: format!("{leaf}/memory.high"),
                current: format!("{leaf}/memory.current"),
                stat: format!("{leaf}/memory.stat"),
                memsw: String::new(),
                events: format!("{leaf}/memory.events"),
            }
        } else {
            MemPaths {
                v2: false,
                max: Vec::new(),
                swap_max: Vec::new(),
                soft: format!("{leaf}/memory.soft_limit_in_bytes"),
                current: format!("{leaf}/memory.usage_in_bytes"),
                stat: format!("{leaf}/memory.stat"),
                memsw: format!("{leaf}/memory.memsw.limit_in_bytes"),
                events: format!("{leaf}/memory.oom_control"),
            }
        })
    }
}

fn meminfo(root: &Root) -> Result<Meminfo> {
    root.with("/proc/meminfo", "memory", "/proc/meminfo", procfs::meminfo)
}

/// A cgroup value in bytes; `None` for `max`, a missing file, or junk.
fn bytes_file(root: &Root, path: &str) -> Option<u64> {
    root.with(
        path,
        "cgroup memory",
        "cgroup memory file",
        procfs::cgroup_bytes,
    )
    .ok()
    .flatten()
    .flatten()
}

/// The smallest finite value over `paths` (the tightest ancestor, E-2).
fn tightest(root: &Root, paths: &[String]) -> Option<u64> {
    paths.iter().filter_map(|p| bytes_file(root, p)).min()
}

/// A `key N` value of a cgroup `memory.stat` or `memory.events` file.
fn stat_value(root: &Root, path: &str, key: &[u8]) -> Option<u64> {
    root.with(path, "cgroup memory", "cgroup memory file", |d| {
        space_kv_value(d, key)
    })
    .ok()
    .flatten()
}

/// The memory reading. `paths` is `None` outside a memory cgroup; `lxcfs`
/// marks a `/proc/meminfo` that already shows the container (E-10).
pub(crate) fn read(
    root: &Root,
    paths: Option<&MemPaths>,
    lxcfs: bool,
    want: MemWant,
) -> Result<MemRaw> {
    let m = meminfo(root)?;
    let malformed = || Error::Malformed {
        what: "memory",
        source: "/proc/meminfo",
        line: None,
    };
    let host_total = m.total.ok_or_else(malformed)?;
    let host_available = m.available().ok_or_else(malformed)?;
    let mut raw = MemRaw {
        host_total,
        host_available,
        host_unused: m.free,
        cached: m.cache(),
        swap: if want.swap {
            Some((m.swap_total.unwrap_or(0), m.swap_free.unwrap_or(0)))
        } else {
            None
        },
        source: Some(if lxcfs {
            MemorySource::Lxcfs
        } else {
            MemorySource::Host
        }),
        ..MemRaw::default()
    };
    if lxcfs || !want.limit {
        return Ok(raw);
    }
    let Some(p) = paths else { return Ok(raw) };
    if p.v2 {
        let Some(hard) = tightest(root, &p.max) else {
            return Ok(raw);
        };
        // E-11: inactive_file read at a different instant may exceed current.
        let used = want
            .usage
            .then(|| bytes_file(root, &p.current))
            .flatten()
            .map(|c| c.saturating_sub(stat_value(root, &p.stat, b"inactive_file").unwrap_or(0)));
        raw.limit = Some(MemoryLimit {
            hard: Bytes::new(hard),
            soft: bytes_file(root, &p.soft).map(Bytes::new),
            swap: tightest(root, &p.swap_max).map(Bytes::new),
            source: LimitSource::Cgroup2,
        });
        raw.limited_used = used;
    } else {
        let hard =
            stat_value(root, &p.stat, b"hierarchical_memory_limit").filter(|&v| v < V1_UNLIMITED);
        let Some(hard) = hard else { return Ok(raw) };
        let used = want
            .usage
            .then(|| bytes_file(root, &p.current))
            .flatten()
            .map(|c| {
                c.saturating_sub(stat_value(root, &p.stat, b"total_inactive_file").unwrap_or(0))
            });
        let soft = bytes_file(root, &p.soft).filter(|&v| v < V1_UNLIMITED);
        // memory.memsw is memory plus swap; the swap share is the difference.
        let swap = bytes_file(root, &p.memsw)
            .filter(|&v| v < V1_UNLIMITED)
            .map(|memsw| memsw.saturating_sub(hard));
        raw.limit = Some(MemoryLimit {
            hard: Bytes::new(hard),
            soft: soft.map(Bytes::new),
            swap: swap.map(Bytes::new),
            source: LimitSource::Cgroup1,
        });
        raw.limited_used = used;
    }
    Ok(raw)
}

/// Out-of-memory events: the cgroup's, else the host's from `/proc/vmstat`.
pub(crate) fn events(root: &Root, paths: Option<&MemPaths>) -> Result<Option<MemoryEvents>> {
    if let Some(p) = paths {
        if p.v2 {
            if let Ok(e) = root.with(&p.events, "memory events", "cgroup memory.events", |d| {
                MemoryEvents {
                    oom_kills: space_kv_value(d, b"oom_kill").unwrap_or(0),
                    oom: space_kv_value(d, b"oom").unwrap_or(0),
                    max: space_kv_value(d, b"max").unwrap_or(0),
                    high: space_kv_value(d, b"high").unwrap_or(0),
                    scope: EventScope::Cgroup,
                }
            }) {
                return Ok(Some(e));
            }
        } else if let Some(kills) = stat_value(root, &p.events, b"oom_kill") {
            return Ok(Some(MemoryEvents {
                oom_kills: kills,
                scope: EventScope::Cgroup,
                ..MemoryEvents::default()
            }));
        }
    }
    // Kernel 4.13 and later count kills in /proc/vmstat.
    let kills = vmstat(root, b"oom_kill")?;
    Ok(kills.map(|oom_kills| MemoryEvents {
        oom_kills,
        scope: EventScope::Host,
        ..MemoryEvents::default()
    }))
}

/// One `/proc/vmstat` counter, read line by line through the stack.
fn vmstat(root: &Root, key: &[u8]) -> Result<Option<u64>> {
    let mut value = None;
    root.for_each_line(
        "/proc/vmstat",
        super::FILE_LIMIT,
        "paging",
        "/proc/vmstat",
        |line| match space_kv_value(line, key) {
            Some(v) => {
                value = Some(v);
                ControlFlow::Break(())
            }
            None => ControlFlow::Continue(()),
        },
    )?;
    Ok(value)
}

/// The paging counters in `/proc/vmstat` (pages, turned into bytes).
pub(crate) fn paging(root: &Root, page_size: u64) -> Result<PagingRaw> {
    let (mut swap_in, mut swap_out, mut major) = (None, None, None);
    root.for_each_line(
        "/proc/vmstat",
        super::FILE_LIMIT,
        "paging",
        "/proc/vmstat",
        |line| {
            if let Some(v) = space_kv_value(line, b"pswpin") {
                swap_in = Some(v);
            } else if let Some(v) = space_kv_value(line, b"pswpout") {
                swap_out = Some(v);
            } else if let Some(v) = space_kv_value(line, b"pgmajfault") {
                major = Some(v);
            }
            if swap_in.is_some() && swap_out.is_some() && major.is_some() {
                ControlFlow::Break(())
            } else {
                ControlFlow::Continue(())
            }
        },
    )?;
    let malformed = || Error::Malformed {
        what: "paging",
        source: "/proc/vmstat",
        line: None,
    };
    Ok(PagingRaw {
        swap_in: swap_in.ok_or_else(malformed)?.saturating_mul(page_size),
        swap_out: swap_out.ok_or_else(malformed)?.saturating_mul(page_size),
        major_faults: major,
    })
}

/// The explicit huge-page pool and the transparent-huge-page mode.
pub(crate) fn huge_pages(root: &Root) -> Result<Option<HugePages>> {
    let m = meminfo(root)?;
    let Some(size) = m.huge_size else {
        return Ok(None);
    };
    let transparent = root
        .with(
            "/sys/kernel/mm/transparent_hugepage/enabled",
            "huge pages",
            "/sys/kernel/mm/transparent_hugepage/enabled",
            |d| match procfs::bracketed(d) {
                Some(b"always") => Some(TransparentHugePages::Always),
                Some(b"madvise") => Some(TransparentHugePages::Madvise),
                Some(b"never") => Some(TransparentHugePages::Never),
                _ => None,
            },
        )
        .ok()
        .flatten();
    Ok(Some(HugePages {
        page_size: Bytes::new(size),
        total: m.huge_total.unwrap_or(0),
        free: m.huge_free.unwrap_or(0),
        reserved: m.huge_rsvd.unwrap_or(0),
        transparent,
        large_page_minimum: None,
    }))
}

#[cfg(test)]
mod tests {
    use super::{super::fixtures, *};
    use crate::mem::from_raw;

    fn snapshot(name: &str) -> crate::mem::Memory {
        let root = fixtures::root(name);
        let cg = fixtures::cgroups(name);
        let paths = MemPaths::new(&cg);
        from_raw(
            read(
                &root,
                paths.as_ref(),
                cg.lxcfs.meminfo,
                MemWant::ALL_FOR_TESTS,
            )
            .unwrap(),
        )
    }

    #[test]
    fn test_linux_mem_docker_cgroup_v2_limit_is_the_total() {
        let m = snapshot("docker-cgroup2-limited");
        assert_eq!(m.total, Bytes::from_mib(256));
        assert_eq!(m.source, MemorySource::Cgroup2);
        let limit = m.limit.unwrap();
        // Docker maps --memory-reservation to memory.low on cgroup v2, not to
        // the memory.high throttle point: no soft limit.
        assert_eq!((limit.hard, limit.soft), (Bytes::from_mib(256), None));
        assert!(m.used <= m.total && m.host_total > m.total);
        assert_eq!(m.used.saturating_add(m.free), m.total);
    }

    #[test]
    fn test_linux_mem_scope_limit_found_through_max_ancestors() {
        // E-2, E-5: a systemd scope with MemoryMax=256M under ancestors at max.
        let m = snapshot("wsl2-systemd-scope-limited");
        assert_eq!(m.total, Bytes::from_mib(256));
        assert_eq!(m.limit.unwrap().soft, Some(Bytes::from_mib(200)));
    }

    #[test]
    fn test_linux_mem_cgroup_v1_hierarchical_limit() {
        // E-4: the v1 memory controller.
        let m = snapshot("docker-cgroup1-limited");
        assert_eq!(m.total, Bytes::from_mib(256));
        assert_eq!(m.source, MemorySource::Cgroup1);
        assert!(m.used <= m.total);
    }

    #[test]
    fn test_linux_mem_without_a_limit_is_the_host() {
        // E-1: the WSL 2 init scope has memory.max = max.
        let m = snapshot("wsl2");
        assert_eq!(m.source, MemorySource::Host);
        assert_eq!(m.total, m.host_total);
        assert!(m.swap.total >= m.swap.used);
        let free_b: u64 = fixtures::tool("wsl2", "free -b")
            .unwrap()
            .lines()
            .find(|l| l.starts_with("Mem:"))
            .and_then(|l| l.split_whitespace().nth(1)?.parse().ok())
            .unwrap();
        assert_eq!(m.host_total.raw(), free_b);
    }

    #[test]
    fn test_linux_mem_lxcfs_is_not_limited_twice() {
        let root = fixtures::root("docker-cgroup2-limited");
        let paths = MemPaths::new(&fixtures::cgroups("docker-cgroup2-limited"));
        let raw = read(&root, paths.as_ref(), true, MemWant::ALL_FOR_TESTS).unwrap();
        assert_eq!((raw.limit, raw.source), (None, Some(MemorySource::Lxcfs)));
    }

    #[test]
    fn test_linux_mem_events_huge_pages_and_paging_from_fixtures() {
        let name = "wsl2-systemd-scope-limited";
        let root = fixtures::root(name);
        let paths = MemPaths::new(&fixtures::cgroups(name));
        let e = events(&root, paths.as_ref()).unwrap().unwrap();
        assert_eq!(e.scope, EventScope::Cgroup);
        let h = huge_pages(&root).unwrap().unwrap();
        assert_eq!(h.page_size, Bytes::from_mib(2));
        assert!(h.free <= h.total);
        let p = paging(&root, 4096).unwrap();
        assert!(p.major_faults.is_some());
        // No cgroup: the host's kill counter.
        let host = events(&root, None).unwrap().unwrap();
        assert_eq!(host.scope, EventScope::Host);
    }

    #[test]
    fn test_linux_mem_missing_meminfo_is_not_found() {
        let err = read(
            &fixtures::root("no-such-fixture"),
            None,
            false,
            MemWant::ALL_FOR_TESTS,
        )
        .unwrap_err();
        assert!(err.is_unavailable(), "{err}");
    }
}