Merge pull request #131 from YJDoc2/comment-capabilities

Document capabilities rs and refactor its drop_privileges function

docs/doc-draft.md

@@ -92,3 +92,7 @@ This also provides implementation for Linux syscalls for the trait.
 
 [oci runtime specification]: https://github.com/opencontainers/runtime-spec/blob/master/runtime.md
 [runc man pages]: (https://github.com/opencontainers/runc/blob/master/man/runc.8.md)
+
+## Capabilities
+
+- [Simple explanation of capabilities](https://blog.container-solutions.com/linux-capabilities-in-practice)

src/capabilities.rs

@@ -1,9 +1,11 @@
+//! Handles Management of Capabilities
 use crate::command::Syscall;
 use caps::*;
 
 use anyhow::Result;
 use oci_spec::{LinuxCapabilities, LinuxCapabilityType};
 
+/// Converts a list of capability types to capabilities has set
 fn to_set(caps: &[LinuxCapabilityType]) -> CapsHashSet {
     let mut capabilities = CapsHashSet::new();
     for c in caps {
@@ -12,29 +14,25 @@ fn to_set(caps: &[LinuxCapabilityType]) -> CapsHashSet {
     capabilities
 }
 
+/// reset capabilities of process calling this to effective capabilities
+/// effective capability set is set of capabilities used by kernel to perform checks
+/// see https://man7.org/linux/man-pages/man7/capabilities.7.html for more information
 pub fn reset_effective(syscall: &impl Syscall) -> Result<()> {
     log::debug!("reset all caps");
     syscall.set_capability(CapSet::Effective, &caps::all())?;
     Ok(())
 }
 
+/// Drop any extra granted capabilities, and reset to defaults which are in oci specification
 pub fn drop_privileges(cs: &LinuxCapabilities, syscall: &impl Syscall) -> Result<()> {
-    let all = caps::all();
     log::debug!("dropping bounding capabilities to {:?}", cs.bounding);
-    for c in all.difference(&to_set(&cs.bounding)) {
-        match c {
-            Capability::CAP_PERFMON | Capability::CAP_CHECKPOINT_RESTORE | Capability::CAP_BPF => {
-                log::warn!("{:?} doesn't support.", c);
-                continue;
-            }
-            _ => caps::drop(None, CapSet::Bounding, *c)?,
-        }
-    }
+    syscall.set_capability(CapSet::Bounding, &to_set(&cs.bounding))?;
 
     syscall.set_capability(CapSet::Effective, &to_set(&cs.effective))?;
     syscall.set_capability(CapSet::Permitted, &to_set(&cs.permitted))?;
     syscall.set_capability(CapSet::Inheritable, &to_set(&cs.inheritable))?;
 
+    // check specifically for ambient, as those might not always be available
     if let Err(e) = syscall.set_capability(CapSet::Ambient, &to_set(&cs.ambient)) {
         log::error!("failed to set ambient capabilities: {}", e);
     }

src/command/linux.rs

@@ -5,7 +5,7 @@ use std::sync::Arc;
 use std::{any::Any, mem, path::Path, ptr};
 
 use anyhow::{bail, Result};
-use caps::{errors::CapsError, CapSet, CapsHashSet};
+use caps::{errors::CapsError, CapSet, Capability, CapsHashSet};
 use libc::{c_char, uid_t};
 use nix::{
     errno::Errno,
@@ -114,7 +114,31 @@ impl Syscall for LinuxSyscall {
 
     /// Set capabilities for container process
     fn set_capability(&self, cset: CapSet, value: &CapsHashSet) -> Result<(), CapsError> {
-        caps::set(None, cset, value)
+        match cset {
+            // caps::set cannot set capabilities in bounding set,
+            // so we do it differently
+            CapSet::Bounding => {
+                // get all capabilities
+                let all = caps::all();
+                // the difference will give capabilities
+                // which are to be unset
+                // for each such =, drop that capability
+                // after this, only those which are to be set will remain set
+                for c in all.difference(value) {
+                    match c {
+                        Capability::CAP_PERFMON
+                        | Capability::CAP_CHECKPOINT_RESTORE
+                        | Capability::CAP_BPF => {
+                            log::warn!("{:?} is not supported.", c);
+                            continue;
+                        }
+                        _ => caps::drop(None, CapSet::Bounding, *c)?,
+                    }
+                }
+                Ok(())
+            }
+            _ => caps::set(None, cset, value),
+        }
     }
 
     /// Sets hostname for process