blackship 0.2.1

A FreeBSD jail manager with state machine control
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
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
//! RCTL (Resource Control) for per-jail resource limits
//!
//! Wraps FreeBSD's rctl(4) syscalls to apply resource limits to jails.
//! Rule format: `jail:<name>:<resource>:<action>=<amount>`

// FFI module: rctl(2) syscalls.
#![allow(unsafe_code)]

use crate::error::{Error, Result};
use serde::{Deserialize, Serialize};
use std::fmt::Write as _;

// Syscall numbers from sys/syscall.h
const SYS_RCTL_ADD_RULE: libc::c_int = 528;
const SYS_RCTL_REMOVE_RULE: libc::c_int = 529;
const SYS_RCTL_GET_RACCT: libc::c_int = 525;

/// Resource limit configuration for a jail
#[derive(Debug, Clone, Default, Deserialize, Serialize)]
pub struct ResourceConfig {
    /// Memory limit (e.g., "1g", "512m")
    pub memory: Option<String>,
    /// Virtual memory limit
    pub vmemory: Option<String>,
    /// CPU percentage limit (e.g., "100" = one core)
    pub cpu: Option<String>,
    /// Maximum processes
    pub maxproc: Option<u32>,
    /// Maximum threads
    pub maxthreads: Option<u32>,
    /// Maximum open files
    pub openfiles: Option<u32>,
    /// Maximum swap usage
    pub swap: Option<String>,
    /// Pin the jail to this many cores, chosen least-loaded across the
    /// other jails' pins
    pub cores: Option<u16>,
    /// Pin the jail to an explicit CPU list (e.g., "0,1,2"); wins over cores
    pub cpuset: Option<String>,
    /// Custom RCTL rules (advanced, raw format)
    #[serde(default)]
    pub rules: Vec<String>,
}

impl ResourceConfig {
    pub const fn is_empty(&self) -> bool {
        self.memory.is_none()
            && self.vmemory.is_none()
            && self.cpu.is_none()
            && self.maxproc.is_none()
            && self.maxthreads.is_none()
            && self.openfiles.is_none()
            && self.swap.is_none()
            && self.cores.is_none()
            && self.cpuset.is_none()
            && self.rules.is_empty()
    }

    /// True when only cpuset pinning is configured (no rctl rules needed)
    pub const fn has_rctl_rules(&self) -> bool {
        self.memory.is_some()
            || self.vmemory.is_some()
            || self.cpu.is_some()
            || self.maxproc.is_some()
            || self.maxthreads.is_some()
            || self.openfiles.is_some()
            || self.swap.is_some()
            || !self.rules.is_empty()
    }

    /// Convert to a list of RCTL rule strings for a given jail name
    pub(crate) fn to_rules(&self, jail_name: &str) -> Vec<String> {
        let mut rules = Vec::new();

        if let Some(ref mem) = self.memory {
            rules.push(format!("jail:{jail_name}:memoryuse:deny={mem}"));
        }
        if let Some(ref vmem) = self.vmemory {
            rules.push(format!("jail:{jail_name}:vmemoryuse:deny={vmem}"));
        }
        if let Some(ref cpu) = self.cpu {
            rules.push(format!("jail:{jail_name}:pcpu:deny={cpu}"));
        }
        if let Some(maxproc) = self.maxproc {
            rules.push(format!("jail:{jail_name}:maxproc:deny={maxproc}"));
        }
        if let Some(maxthreads) = self.maxthreads {
            rules.push(format!("jail:{jail_name}:nthr:deny={maxthreads}"));
        }
        if let Some(openfiles) = self.openfiles {
            rules.push(format!("jail:{jail_name}:openfiles:deny={openfiles}"));
        }
        if let Some(ref swap) = self.swap {
            rules.push(format!("jail:{jail_name}:swapuse:deny={swap}"));
        }

        let own_prefix = format!("jail:{jail_name}:");
        for rule in &self.rules {
            if rule.starts_with("jail:") {
                if !rule.starts_with(&own_prefix) {
                    eprintln!("Warning: rejecting RCTL rule '{rule}' -- targets a different jail");
                    continue;
                }
                rules.push(rule.clone());
            } else {
                // Prefix with jail subject if not already
                rules.push(format!("jail:{jail_name}:{rule}"));
            }
        }

        rules
    }
}

/// What moves a jail from its startup profile to its steady profile.
#[derive(Debug, Clone, Copy, PartialEq, Eq, Default, Deserialize, Serialize)]
#[serde(rename_all = "lowercase")]
pub enum QosTransition {
    /// First passing health check. Needs `blackship supervise`.
    #[default]
    Healthy,
    /// `max_startup_secs` elapsing.
    Timer,
    /// Only `blackship qos <jail> --steady`.
    Manual,
}

/// Two resource profiles and the rule that switches between them.
///
/// A jail needs headroom while it builds caches and forks workers, and a
/// tighter cap once it settles. One static profile has to be the looser of the
/// two, which leaves the steady state unprotected.
#[derive(Debug, Clone, Default, Deserialize, Serialize)]
pub struct QosConfig {
    /// Applied at start, in place of `resources`.
    pub startup: ResourceConfig,
    /// Applied on transition. Absent means the startup profile is permanent.
    #[serde(default)]
    pub steady: Option<ResourceConfig>,
    #[serde(default)]
    pub transition: QosTransition,
    /// Fallback deadline for the `healthy` transition, in seconds.
    #[serde(default)]
    pub max_startup_secs: Option<u64>,
}

impl QosConfig {
    pub fn profile(&self, phase: crate::scope::QosPhase) -> &ResourceConfig {
        match phase {
            crate::scope::QosPhase::Steady => self.steady.as_ref().unwrap_or(&self.startup),
            _ => &self.startup,
        }
    }
}

/// Resources a profile owns. Anything else in `rules` belongs to the user.
const MANAGED_RESOURCES: [&str; 7] = [
    "memoryuse",
    "vmemoryuse",
    "pcpu",
    "maxproc",
    "nthr",
    "openfiles",
    "swapuse",
];

/// The resources a profile owns, for callers outside this module.
#[cfg_attr(
    not(feature = "mcp"),
    expect(
        dead_code,
        reason = "the MCP host_facts tool is this accessor's only caller"
    )
)]
pub const fn managed_resources() -> &'static [&'static str] {
    &MANAGED_RESOURCES
}

/// Split a profile's rules into the managed `(resource, rule)` pairs and the
/// user's own rules, in one pass over `to_rules`.
fn split_rules(config: &ResourceConfig, jail_name: &str) -> (Vec<(String, String)>, Vec<String>) {
    let prefix = format!("jail:{jail_name}:");
    let mut managed = Vec::new();
    let mut custom = Vec::new();

    for rule in config.to_rules(jail_name) {
        let resource = rule
            .strip_prefix(&prefix)
            .and_then(|rest| rest.split(':').next())
            .filter(|resource| MANAGED_RESOURCES.contains(resource))
            .map(str::to_string);
        match resource {
            Some(resource) => managed.push((resource, rule)),
            None => custom.push(rule),
        }
    }

    (managed, custom)
}

/// Rules to remove and rules to add for a phase switch.
///
/// `remove_limits` uses the blanket `jail:<name>` filter, which would drop the
/// user's custom `rules` along with the profile's. A switch therefore removes
/// per-resource, and only for resources the profiles actually manage.
pub fn phase_diff(
    jail_name: &str,
    from: &ResourceConfig,
    to: &ResourceConfig,
) -> (Vec<String>, Vec<String>) {
    let (from_managed, from_custom) = split_rules(from, jail_name);
    let (to_managed, to_custom) = split_rules(to, jail_name);

    let mut removals = Vec::new();
    let mut additions = Vec::new();

    for (resource, rule) in &from_managed {
        match to_managed.iter().find(|(r, _)| r == resource) {
            Some((_, new_rule)) if new_rule == rule => {}
            _ => removals.push(format!("jail:{jail_name}:{resource}")),
        }
    }

    for (resource, rule) in &to_managed {
        match from_managed.iter().find(|(r, _)| r == resource) {
            Some((_, old_rule)) if old_rule == rule => {}
            _ => additions.push(rule.clone()),
        }
    }

    for rule in &from_custom {
        if !to_custom.contains(rule) {
            removals.push(rule.clone());
        }
    }
    for rule in &to_custom {
        if !from_custom.contains(rule) {
            additions.push(rule.clone());
        }
    }

    (removals, additions)
}

/// Switch a live jail from one profile to another.
pub fn apply_phase(jail_name: &str, from: &ResourceConfig, to: &ResourceConfig) -> Result<()> {
    let (removals, additions) = phase_diff(jail_name, from, to);
    if removals.is_empty() && additions.is_empty() {
        return Ok(());
    }

    if !is_enabled() {
        return Err(Error::Rctl(
            "RCTL not enabled. Set kern.racct.enable=1 in /boot/loader.conf and reboot.".into(),
        ));
    }

    for filter in &removals {
        if let Err(e) = rctl_remove_rule(filter) {
            eprintln!("Warning: {e}");
        }
    }
    for rule in &additions {
        rctl_add_rule(rule)?;
    }

    Ok(())
}

/// Resolve the CPU list a jail should be pinned to.
///
/// Explicit `cpuset` wins. `cores = N` picks the N logical CPUs with the
/// fewest explicit claims across the other jails' configs (least-loaded
/// spreading), so co-hosted jails end up on different cores by default.
pub fn resolve_cpu_list<'a>(
    config: &ResourceConfig,
    other_pins: impl Iterator<Item = &'a str>,
) -> Option<String> {
    if let Some(list) = &config.cpuset {
        return Some(list.clone());
    }
    let want = config.cores? as usize;
    // SAFETY: plain integer arg.
    let ncpu = unsafe { libc::sysconf(libc::_SC_NPROCESSORS_ONLN) };
    let ncpu = usize::try_from(ncpu).ok().filter(|&n| n > 0).unwrap_or(1);

    let mut load = vec![0usize; ncpu];
    for pin in other_pins {
        for part in pin.split(',') {
            if let Ok(core) = part.trim().parse::<usize>()
                && core < ncpu
            {
                load[core] += 1;
            }
        }
    }

    let mut cores: Vec<usize> = (0..ncpu).collect();
    cores.sort_by_key(|&c| (load[c], c));
    cores.truncate(want.min(ncpu));
    cores.sort_unstable();
    let mut list = String::with_capacity(cores.len() * 3);
    for core in cores {
        if !list.is_empty() {
            list.push(',');
        }
        // Writing to a String cannot fail.
        let _ = write!(list, "{core}");
    }
    Some(list)
}

/// Pin a running jail to a CPU list via cpuset(1)
pub fn apply_cpuset(jid: i32, cpu_list: &str) -> Result<()> {
    let status = std::process::Command::new("/usr/bin/cpuset")
        .args(["-l", cpu_list, "-j", &jid.to_string()])
        .status()
        .map_err(|e| Error::Rctl(format!("Failed to run cpuset: {e}")))?;
    if !status.success() {
        return Err(Error::Rctl(format!("cpuset -l {cpu_list} -j {jid} failed")));
    }
    Ok(())
}

/// Check if RCTL is enabled in the kernel
pub fn is_enabled() -> bool {
    // Check kern.racct.enable sysctl
    let output = std::process::Command::new("/sbin/sysctl")
        .args(["-n", "kern.racct.enable"])
        .output();

    match output {
        Ok(out) if out.status.success() => String::from_utf8_lossy(&out.stdout).trim() == "1",
        _ => false,
    }
}

/// Apply resource limits to a jail
pub fn apply_limits(jail_name: &str, config: &ResourceConfig) -> Result<()> {
    if !config.has_rctl_rules() {
        // Pure cpuset pinning needs no rctl (and no RACCT).
        return Ok(());
    }

    if !is_enabled() {
        return Err(Error::Rctl(
            "RCTL not enabled. Set kern.racct.enable=1 in /boot/loader.conf and reboot.".into(),
        ));
    }

    let rules = config.to_rules(jail_name);

    for rule in &rules {
        rctl_add_rule(rule)?;
    }

    if !rules.is_empty() {
        println!("  Applied {} resource limit(s)", rules.len());
    }

    Ok(())
}

/// Remove all resource limits for a jail
pub fn remove_limits(jail_name: &str) {
    // Remove all rules matching this jail
    let filter = format!("jail:{jail_name}");
    // ESRCH (no rules) and ENOSYS (no RACCT) are expected.
    let _ = rctl_remove_rule(&filter);
}

/// Get current resource usage for a jail
pub fn get_usage(jail_name: &str) -> Result<std::collections::HashMap<String, u64>> {
    let filter = format!("jail:{jail_name}");
    rctl_get_racct(&filter)
}

/// Invoke an RCTL syscall that takes a rule/filter string and no output buffer
fn rctl_string_syscall(
    syscall_num: libc::c_int,
    arg: &str,
    fail_msg: impl FnOnce(std::io::Error) -> String,
) -> Result<()> {
    let arg_bytes = arg.as_bytes();

    // SAFETY: live input with its length; NULL output with length 0.
    let ret = unsafe {
        libc::syscall(
            syscall_num,
            arg_bytes.as_ptr().cast::<libc::c_void>(),
            arg_bytes.len(),
            std::ptr::null_mut::<libc::c_void>(),
            0usize,
        )
    };

    if ret != 0 {
        return Err(Error::Rctl(fail_msg(std::io::Error::last_os_error())));
    }

    Ok(())
}

/// Add an RCTL rule via syscall
fn rctl_add_rule(rule: &str) -> Result<()> {
    rctl_string_syscall(SYS_RCTL_ADD_RULE, rule, |e| {
        format!("Failed to add rule '{rule}': {e}")
    })
}

/// Remove RCTL rules matching a filter
fn rctl_remove_rule(filter: &str) -> Result<()> {
    rctl_string_syscall(SYS_RCTL_REMOVE_RULE, filter, |e| {
        format!("Failed to remove rules for '{filter}': {e}")
    })
}

/// Get resource accounting data for a subject
fn rctl_get_racct(filter: &str) -> Result<std::collections::HashMap<String, u64>> {
    let filter_bytes = filter.as_bytes();
    let mut outbuf = [0u8; 4096];

    // SAFETY: live buffers, each passed with its own length.
    let ret = unsafe {
        libc::syscall(
            SYS_RCTL_GET_RACCT,
            filter_bytes.as_ptr().cast::<libc::c_void>(),
            filter_bytes.len(),
            outbuf.as_mut_ptr().cast::<libc::c_void>(),
            outbuf.len(),
        )
    };

    if ret != 0 {
        return Err(Error::Rctl(format!(
            "Failed to get resource usage for '{}': {}",
            filter,
            std::io::Error::last_os_error()
        )));
    }

    // Parse output: "resource=value,resource=value,..."
    let output = String::from_utf8_lossy(&outbuf);
    let output = output.trim_end_matches('\0');

    let mut usage = std::collections::HashMap::new();
    for pair in output.split(',') {
        if let Some((key, val)) = pair.split_once('=')
            && let Ok(v) = val.parse::<u64>()
        {
            usage.insert(key.to_string(), v);
        }
    }

    Ok(usage)
}

#[cfg(test)]
mod tests;