kernel_cmdline: Add canonicalized() methods

This seems generally useful, mostly I could see always canonicalizing
the cmdline generated by bootc to remove a source of potential
drift/nondeterminism and better enable reproducible builds.

Signed-off-by: John Eckersberg <[email protected]>

Author: jeckersb

crates/kernel_cmdline/src/bytes.rs

@@ -340,6 +340,31 @@ impl<'a> Cmdline<'a> {
         removed
     }
 
+    /// Returns the canonicalized version of the `Cmdline`.
+    ///
+    /// This:
+    ///
+    /// 1. Sorts the parameter list
+    /// 2. Canonicalizes each `Parameter`
+    /// 3. Joins each parameter together with a single space ' '
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use bootc_kernel_cmdline::bytes::Cmdline;
+    ///
+    /// let cmdline = Cmdline::from(b"z a=\"b c\"");
+    /// assert_eq!(&cmdline.canonicalized(), b"\"a=b c\" z");
+    /// ```
+    pub fn canonicalized(&self) -> Vec<u8> {
+        let mut params = self.iter().collect::<Vec<_>>();
+        params.sort();
+
+        let canonicalized_params: Vec<Vec<u8>> = params.iter().map(|p| p.canonicalized()).collect();
+
+        canonicalized_params.join(&b' ')
+    }
+
     #[cfg(test)]
     pub(crate) fn is_owned(&self) -> bool {
         matches!(self.0, Cow::Owned(_))
@@ -426,6 +451,21 @@ impl ParameterKey<'_> {
             .iter()
             .map(|&c: &u8| if c == b'-' { b'_' } else { c })
     }
+
+    /// Returns the canonicalized version of the key.  This replaces
+    /// all dashes '-' with underscores '_'.
+    ///
+    /// # Example
+    ///
+    /// ```
+    /// use bootc_kernel_cmdline::bytes::ParameterKey;
+    ///
+    /// assert_eq!(&ParameterKey::from("key-with-dashes").canonicalized(),
+    ///            "key_with_dashes".as_bytes());
+    /// ```
+    pub fn canonicalized(&self) -> Vec<u8> {
+        self.iter().collect()
+    }
 }
 
 impl PartialEq for ParameterKey<'_> {
@@ -528,6 +568,58 @@ impl<'a> Parameter<'a> {
     pub fn value(&self) -> Option<&'a [u8]> {
         self.value
     }
+
+    /// Returns the canonical representation of the parameter.
+    ///
+    /// The canonical representation:
+    ///
+    /// 1.  Will use the canonicalized form of the key via
+    /// `ParameterKey::canonicalized`
+    ///
+    /// 2.  Will be "externally" quoted if either the key or
+    /// (optional) value contains ascii whitespace.
+    ///
+    /// 3.  Unnecessary quoting will be removed.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use bootc_kernel_cmdline::bytes::Parameter;
+    ///
+    /// // key is canonicalized
+    /// assert_eq!(Parameter::parse("a-dashed-key").unwrap().canonicalized(),
+    ///            "a_dashed_key".as_bytes());
+    ///
+    /// // quotes are externally added if needed
+    /// assert_eq!(Parameter::parse("foo=\"has some spaces\"").unwrap().canonicalized(),
+    ///            "\"foo=has some spaces\"".as_bytes());
+    ///
+    /// // unnecessary quotes are removed
+    /// assert_eq!(Parameter::parse("foo=\"bar\"").unwrap().canonicalized(),
+    ///            "foo=bar".as_bytes());
+    /// ```
+    pub fn canonicalized(&self) -> Vec<u8> {
+        let key = self.key.canonicalized();
+        let spaces = key.iter().any(|b| b.is_ascii_whitespace())
+            || self
+                .value
+                .map_or(false, |val| val.iter().any(|b| b.is_ascii_whitespace()));
+
+        let mut res = if spaces { vec![b'"'] } else { vec![] };
+
+        res.extend(&key);
+
+        if let Some(val) = self.value {
+            res.push(b'=');
+            res.extend(val);
+        }
+
+        if spaces {
+            res.push(b'"');
+        }
+
+        res
+    }
 }
 
 impl PartialEq for Parameter<'_> {

crates/kernel_cmdline/src/utf8.rs

@@ -203,6 +203,29 @@ impl<'a> Cmdline<'a> {
         self.0.remove_exact(&param.0)
     }
 
+    /// Returns the canonicalized version of the `Cmdline`.
+    ///
+    /// This:
+    ///
+    /// 1. Sorts the parameter list
+    /// 2. Canonicalizes each `Parameter`
+    /// 3. Joins each parameter together with a single space ' '
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use bootc_kernel_cmdline::utf8::Cmdline;
+    ///
+    /// let cmdline = Cmdline::from("z a=\"b c\"");
+    /// assert_eq!(&cmdline.canonicalized(), "\"a=b c\" z");
+    /// ```
+    pub fn canonicalized(&self) -> String {
+        self.0
+            .canonicalized()
+            .try_into()
+            .expect("We only construct the underlying bytes from valid UTF-8")
+    }
+
     #[cfg(test)]
     pub(crate) fn is_owned(&self) -> bool {
         self.0.is_owned()
@@ -298,6 +321,24 @@ impl<'a> ParameterKey<'a> {
     fn from_bytes(input: bytes::ParameterKey<'a>) -> Self {
         Self(input)
     }
+
+    /// Returns the canonicalized version of the key.  This replaces
+    /// all dashes '-' with underscores '_'.
+    ///
+    /// # Example
+    ///
+    /// ```
+    /// use bootc_kernel_cmdline::utf8::ParameterKey;
+    ///
+    /// assert_eq!(ParameterKey::from("key-with-dashes").canonicalized(),
+    ///            "key_with_dashes".to_string());
+    /// ```
+    pub fn canonicalized(&self) -> String {
+        self.0
+            .canonicalized()
+            .try_into()
+            .expect("We only construct the underlying bytes from valid UTF-8")
+    }
 }
 
 impl<'a, T: AsRef<str> + ?Sized> From<&'a T> for ParameterKey<'a> {
@@ -358,6 +399,42 @@ impl<'a> Parameter<'a> {
             str::from_utf8(p).expect("We only construct the underlying bytes from valid UTF-8")
         })
     }
+
+    /// Returns the canonical representation of the parameter.
+    ///
+    /// The canonical representation:
+    ///
+    /// 1.  Will use the canonicalized form of the key via
+    /// `ParameterKey::canonicalized`
+    ///
+    /// 2.  Will be "externally" quoted if either the key or
+    /// (optional) value contains ascii whitespace.
+    ///
+    /// 3.  Unnecessary quoting will be removed.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use bootc_kernel_cmdline::utf8::Parameter;
+    ///
+    /// // key is canonicalized
+    /// assert_eq!(Parameter::parse("a-dashed-key").unwrap().canonicalized(),
+    ///            "a_dashed_key".to_string());
+    ///
+    /// // quotes are externally added if needed
+    /// assert_eq!(Parameter::parse("foo=\"has some spaces\"").unwrap().canonicalized(),
+    ///            "\"foo=has some spaces\"".to_string());
+    ///
+    /// // unnecessary quotes are removed
+    /// assert_eq!(Parameter::parse("foo=\"bar\"").unwrap().canonicalized(),
+    ///            "foo=bar".to_string());
+    /// ```
+    pub fn canonicalized(&self) -> String {
+        self.0
+            .canonicalized()
+            .try_into()
+            .expect("We only construct the underlying bytes from valid UTF-8")
+    }
 }
 
 impl<'a> TryFrom<bytes::Parameter<'a>> for Parameter<'a> {