Read short import files

crates/examples/src/objdump.rs

@@ -1,6 +1,7 @@
+use object::import::{ImportName, ShortImportFile};
 use object::read::archive::ArchiveFile;
 use object::read::macho::{DyldCache, FatArch, FatHeader};
-use object::{Endianness, Object, ObjectComdat, ObjectSection, ObjectSymbol};
+use object::{Endianness, FileKind, Object, ObjectComdat, ObjectSection, ObjectSymbol};
 use std::io::{Result, Write};
 
 pub fn print<W: Write, E: Write>(
@@ -21,7 +22,11 @@ pub fn print<W: Write, E: Write>(
                         writeln!(w)?;
                         writeln!(w, "{}:", String::from_utf8_lossy(member.name()))?;
                         if let Ok(data) = member.data(file) {
-                            dump_object(w, e, data)?;
+                            if FileKind::parse(data) == Ok(FileKind::ShortImport) {
+                                dump_import(w, e, data)?;
+                            } else {
+                                dump_object(w, e, data)?;
+                            }
                         }
                     }
                 }
@@ -245,3 +250,24 @@ fn dump_parsed_object<W: Write, E: Write>(w: &mut W, e: &mut E, file: &object::F
 
     Ok(())
 }
+
+fn dump_import<W: Write, E: Write>(w: &mut W, e: &mut E, data: &[u8]) -> Result<()> {
+    let file = match ShortImportFile::parse(data) {
+        Ok(import) => import,
+        Err(err) => {
+            writeln!(e, "Failed to parse short import: {}", err)?;
+            return Ok(());
+        }
+    };
+
+    writeln!(w, "Format: Short Import File")?;
+    writeln!(w, "Architecture: {:?}", file.architecture())?;
+    writeln!(w, "DLL: {:?}", String::from_utf8_lossy(file.dll()))?;
+    writeln!(w, "Symbol: {:?}", String::from_utf8_lossy(file.symbol()))?;
+    write!(w, "Import: ")?;
+    match file.import() {
+        ImportName::Ordinal(n) => writeln!(w, "Ordinal({n})")?,
+        ImportName::Name(name) => writeln!(w, "Name({:?})", String::from_utf8_lossy(name))?,
+    }
+    Ok(())
+}

src/read/import.rs

@@ -0,0 +1,207 @@
+//! Support for reading short import files.
+//!
+//! These are used by some Windows linkers as a more compact way to describe
+//! dynamically imported symbols.
+
+use crate::read::{Architecture, Error, ReadError, ReadRef, Result};
+use crate::{pe, ByteString, Bytes, LittleEndian as LE};
+
+/// Short form description of a symbol to import.
+/// Used in Windows import libraries.
+#[derive(Debug, Clone)]
+pub struct ShortImportFile<'data> {
+    header: &'data pe::ImportObjectHeader,
+    dll: ByteString<'data>,
+    symbol: ByteString<'data>,
+    kind: ImportType,
+    import: Option<ByteString<'data>>,
+}
+impl<'data> ShortImportFile<'data> {
+    /// Parse it.
+    pub fn parse<R: ReadRef<'data>>(data: R) -> Result<Self> {
+        let mut offset = 0;
+        let header = pe::ImportObjectHeader::parse(data, &mut offset)?;
+
+        let mut strings = Bytes(
+            data.read_bytes(&mut offset, header.data_size() as u64)
+                .read_error("Not enough data")?,
+        );
+        let symbol = strings
+            .read_string()
+            .read_error("Could not read symbol name")?;
+        let dll = strings
+            .read_string()
+            .read_error("Could not read DLL name")?;
+
+        // Unmangles a name by removing a `?`, `@` or `_` prefix.
+        fn strip_prefix(s: &[u8]) -> &[u8] {
+            match s.split_first() {
+                Some((b'?' | b'@' | b'_', rest)) => rest,
+                _ => s,
+            }
+        }
+        Ok(Self {
+            header,
+            dll: ByteString(dll),
+            symbol: ByteString(symbol),
+            kind: header.import_type()?,
+            import: match header.name_type()? {
+                ImportNameType::Ordinal => None,
+                ImportNameType::Name => Some(symbol),
+                ImportNameType::NameNoPrefix => Some(strip_prefix(symbol)),
+                ImportNameType::NameUndecorate => {
+                    Some(strip_prefix(symbol).split(|&b| b == b'@').next().unwrap())
+                }
+                ImportNameType::NameExportAs => Some(
+                    strings
+                        .read_string()
+                        .read_error("Could not read export name")?,
+                ),
+            }
+            .map(ByteString),
+        })
+    }
+
+    /// Get the machine type.
+    pub fn architecture(&self) -> Architecture {
+        self.header.architecture()
+    }
+
+    /// The name of the DLL to import the symbol from.
+    pub fn dll(&self) -> &'data [u8] {
+        self.dll.0
+    }
+
+    /// The name exported from the DLL.
+    pub fn import(&self) -> ImportName<'data> {
+        match self.import {
+            Some(name) => ImportName::Name(name.0),
+            None => ImportName::Ordinal(self.header.ordinal()),
+        }
+    }
+
+    /// The type of import. Usually either a function or data.
+    pub fn import_type(&self) -> ImportType {
+        self.kind
+    }
+
+    /// The public symbol name
+    pub fn symbol(&self) -> &'data [u8] {
+        self.symbol.0
+    }
+}
+
+/// The name or ordinal to import.
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum ImportName<'data> {
+    /// Import by ordinal. Ordinarily this is a 1-based index.
+    Ordinal(u16),
+    /// Import by name.
+    Name(&'data [u8]),
+}
+
+/// The kind of import.
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
+pub enum ImportType {
+    /// Executable code
+    Code,
+    /// Some data
+    Data,
+    /// A constant
+    Const,
+}
+impl ImportType {
+    /// `n` must be a two bit number.
+    fn from_u2(n: u16) -> Result<Self> {
+        match n {
+            pe::IMPORT_OBJECT_CODE => Ok(Self::Code),
+            pe::IMPORT_OBJECT_DATA => Ok(Self::Data),
+            pe::IMPORT_OBJECT_CONST => Ok(Self::Const),
+            0b11 => Err(Error("Invalid import type")),
+            _ => unreachable!("ImportType must be a two bit number"),
+        }
+    }
+}
+
+/// Determines how the symbol is imported.
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
+pub enum ImportNameType {
+    /// Import by an offset into the DLL's export table.
+    Ordinal,
+    /// Import using the same name as the symbol name.
+    Name,
+    /// Strip any leading `?`, `@` or `_` from the symbol name to get the import name.
+    NameNoPrefix,
+    /// The same as [`NameNoPrefix`] but also truncates at the next `@`.
+    NameUndecorate,
+    /// The import name is given after the DLL name.
+    NameExportAs,
+}
+impl ImportNameType {
+    /// `n` must be a three bit number.
+    fn from_u3(n: u16) -> Result<Self> {
+        match n {
+            pe::IMPORT_OBJECT_ORDINAL => Ok(ImportNameType::Ordinal),
+            pe::IMPORT_OBJECT_NAME => Ok(ImportNameType::Name),
+            pe::IMPORT_OBJECT_NAME_NO_PREFIX => Ok(ImportNameType::NameNoPrefix),
+            pe::IMPORT_OBJECT_NAME_UNDECORATE => Ok(ImportNameType::NameUndecorate),
+            pe::IMPORT_OBJECT_NAME_EXPORTAS => Ok(ImportNameType::NameExportAs),
+            5..=7 => Err(Error("Unknown name type")),
+            _ => unreachable!("ImportNameType must be a three bit number"),
+        }
+    }
+}
+
+impl pe::ImportObjectHeader {
+    /// Read the short import header.
+    ///
+    /// Also checks that the signature and version are valid.
+    /// Directly following this header will be the string data.
+    pub fn parse<'data, R: ReadRef<'data>>(data: R, offset: &mut u64) -> Result<&'data Self> {
+        let header = data
+            .read::<crate::pe::ImportObjectHeader>(offset)
+            .read_error("Invalid import header size")?;
+        if header.sig1.get(LE) != 0 || header.sig2.get(LE) != pe::IMPORT_OBJECT_HDR_SIG2 {
+            Err(Error("Invalid import header"))
+        } else if header.version.get(LE) != 0 {
+            Err(Error("Unknown import header version"))
+        } else {
+            Ok(header)
+        }
+    }
+
+    /// Get the machine type.
+    pub fn architecture(&self) -> Architecture {
+        match self.machine.get(LE) {
+            pe::IMAGE_FILE_MACHINE_ARMNT => Architecture::Arm,
+            pe::IMAGE_FILE_MACHINE_ARM64 => Architecture::Aarch64,
+            pe::IMAGE_FILE_MACHINE_I386 => Architecture::I386,
+            pe::IMAGE_FILE_MACHINE_AMD64 => Architecture::X86_64,
+            _ => Architecture::Unknown,
+        }
+    }
+
+    /// The type of import can be code, data or const.
+    pub fn import_type(&self) -> Result<ImportType> {
+        ImportType::from_u2(self.name_type.get(LE) & 0b11)
+    }
+
+    /// Describes how to derive the import name.
+    ///
+    /// Returns an error if the type is not recognised.
+    pub fn name_type(&self) -> Result<ImportNameType> {
+        ImportNameType::from_u3((self.name_type.get(LE) >> 2) & 0b111)
+    }
+
+    /// The index into the DLL's export table where the symbol may be found.
+    ///
+    /// If `name_type` is not `Ordinal` then this is only a hint and may be ignored.
+    pub fn ordinal(&self) -> u16 {
+        self.ordinal_or_hint.get(LE)
+    }
+
+    /// The combined size of the null-terminated strings that follow the header.
+    pub fn data_size(&self) -> u32 {
+        self.size_of_data.get(LE)
+    }
+}

src/read/mod.rs

@@ -57,6 +57,9 @@ pub mod wasm;
 #[cfg(feature = "xcoff")]
 pub mod xcoff;
 
+#[cfg(feature = "archive")]
+pub mod import;
+
 mod traits;
 pub use traits::*;
 
@@ -165,6 +168,9 @@ pub enum FileKind {
     /// A 64-bit ELF file.
     #[cfg(feature = "elf")]
     Elf64,
+    /// A Windows short import file.
+    #[cfg(feature = "archive")]
+    ShortImport,
     /// A 32-bit Mach-O file.
     #[cfg(feature = "macho")]
     MachO32,
@@ -263,6 +269,8 @@ impl FileKind {
             [0x01, 0xDF, ..] => FileKind::Xcoff32,
             #[cfg(feature = "xcoff")]
             [0x01, 0xF7, ..] => FileKind::Xcoff64,
+            #[cfg(feature = "archive")]
+            [0, 0, 0xFF, 0xFF, ..] => FileKind::ShortImport,
             _ => return Err(Error("Unknown file magic")),
         };
         Ok(kind)