elf: add ELFCOMPRESS_ZSTD (#532)

Also add readobj support for compression headers.

Cargo.toml

@@ -18,6 +18,7 @@ indexmap = { version = "1.6", optional = true }
 wasmparser = { version = "0.102.0", optional = true }
 memchr = { version = "2.4.1", default-features = false }
 hashbrown = { version = "0.13.1", features = ["ahash"], default-features = false, optional = true }
+ruzstd = { version = "0.3.1", optional = true }
 
 # Internal feature, only used when building as part of libstd, not part of the
 # stable interface of this crate.
@@ -48,7 +49,7 @@ write = ["write_std", "coff", "elf", "macho", "pe", "xcoff"]
 std = ["memchr/std"]
 # Enable decompression of compressed sections.
 # This feature is not required if you want to do your own decompression.
-compression = ["flate2", "std"]
+compression = ["flate2", "ruzstd", "std"]
 # Treat all types as unaligned.
 # Normally types use the alignment required by the specifications, but
 # sometimes files do not strictly follow the specifications.

crates/examples/src/readobj/elf.rs

@@ -262,6 +262,15 @@ fn print_section_headers<Elf: FileHeader>(
             p.field_hex("AddressAlign", section.sh_addralign(endian).into());
             p.field_hex("EntrySize", section.sh_entsize(endian).into());
 
+            if let Some(Some((compression, _, _))) = section.compression(endian, data).print_err(p)
+            {
+                p.group("CompressionHeader", |p| {
+                    p.field_enum("Type", compression.ch_type(endian), FLAGS_ELFCOMPRESS);
+                    p.field_hex("Size", compression.ch_size(endian).into());
+                    p.field_hex("AddressAlign", compression.ch_addralign(endian).into());
+                });
+            }
+
             match section.sh_type(endian) {
                 SHT_SYMTAB | SHT_DYNSYM => {
                     print_section_symbols(p, endian, data, elf, sections, index, section)
@@ -1315,6 +1324,7 @@ static FLAGS_SHF_PARISC: &[Flag<u32>] = &flags!(SHF_PARISC_SHORT, SHF_PARISC_HUG
 static FLAGS_SHF_ALPHA: &[Flag<u32>] = &flags!(SHF_ALPHA_GPREL);
 static FLAGS_SHF_ARM: &[Flag<u32>] = &flags!(SHF_ARM_ENTRYSECT, SHF_ARM_COMDEF);
 static FLAGS_SHF_IA_64: &[Flag<u32>] = &flags!(SHF_IA_64_SHORT, SHF_IA_64_NORECOV);
+static FLAGS_ELFCOMPRESS: &[Flag<u32>] = &flags!(ELFCOMPRESS_ZLIB, ELFCOMPRESS_ZSTD);
 static FLAGS_STT: &[Flag<u8>] = &flags!(
     STT_NOTYPE,
     STT_OBJECT,

crates/examples/testfiles/elf/base-debug-zlib.o.readobj.compression

@@ -0,0 +1,19 @@
+SectionHeader {
+    Index: 6
+    Name: ".debug_info" (0x3E)
+    Type: SHT_PROGBITS (0x1)
+    Flags: 0x800
+        SHF_COMPRESSED (0x800)
+    Address: 0x0
+    Offset: 0x70
+    Size: 0x74
+    Link: 0
+    Info: 0
+    AddressAlign: 0x8
+    EntrySize: 0x0
+    CompressionHeader {
+        Type: ELFCOMPRESS_ZLIB (0x1)
+        Size: 0xAF
+        AddressAlign: 0x1
+    }
+}

crates/examples/testfiles/elf/base-debug-zstd.o.readobj.compression

@@ -0,0 +1,19 @@
+SectionHeader {
+    Index: 6
+    Name: ".debug_info" (0x3E)
+    Type: SHT_PROGBITS (0x1)
+    Flags: 0x800
+        SHF_COMPRESSED (0x800)
+    Address: 0x0
+    Offset: 0x70
+    Size: 0x84
+    Link: 0
+    Info: 0
+    AddressAlign: 0x8
+    EntrySize: 0x0
+    CompressionHeader {
+        Type: ELFCOMPRESS_ZSTD (0x2)
+        Size: 0xAF
+        AddressAlign: 0x1
+    }
+}

src/elf.rs

@@ -819,6 +819,8 @@ pub struct CompressionHeader64<E: Endian> {
 
 /// ZLIB/DEFLATE algorithm.
 pub const ELFCOMPRESS_ZLIB: u32 = 1;
+/// Zstandard algorithm.
+pub const ELFCOMPRESS_ZSTD: u32 = 2;
 /// Start of OS-specific compression types.
 pub const ELFCOMPRESS_LOOS: u32 = 0x6000_0000;
 /// End of OS-specific compression types.

src/read/elf/section.rs

@@ -381,32 +381,24 @@ impl<'data, 'file, Elf: FileHeader, R: ReadRef<'data>> ElfSection<'data, 'file,
 
     fn maybe_compressed(&self) -> read::Result<Option<CompressedFileRange>> {
         let endian = self.file.endian;
-        if (self.section.sh_flags(endian).into() & u64::from(elf::SHF_COMPRESSED)) == 0 {
-            return Ok(None);
-        }
-        let (section_offset, section_size) = self
-            .section
-            .file_range(endian)
-            .read_error("Invalid ELF compressed section type")?;
-        let mut offset = section_offset;
-        let header = self
-            .file
-            .data
-            .read::<Elf::CompressionHeader>(&mut offset)
-            .read_error("Invalid ELF compressed section offset")?;
-        if header.ch_type(endian) != elf::ELFCOMPRESS_ZLIB {
-            return Err(Error("Unsupported ELF compression type"));
+        if let Some((header, offset, compressed_size)) =
+            self.section.compression(endian, self.file.data)?
+        {
+            let format = match header.ch_type(endian) {
+                elf::ELFCOMPRESS_ZLIB => CompressionFormat::Zlib,
+                elf::ELFCOMPRESS_ZSTD => CompressionFormat::Zstandard,
+                _ => return Err(Error("Unsupported ELF compression type")),
+            };
+            let uncompressed_size = header.ch_size(endian).into();
+            Ok(Some(CompressedFileRange {
+                format,
+                offset,
+                compressed_size,
+                uncompressed_size,
+            }))
+        } else {
+            Ok(None)
         }
-        let uncompressed_size = header.ch_size(endian).into();
-        let compressed_size = section_size
-            .checked_sub(offset - section_offset)
-            .read_error("Invalid ELF compressed section size")?;
-        Ok(Some(CompressedFileRange {
-            format: CompressionFormat::Zlib,
-            offset,
-            compressed_size,
-            uncompressed_size,
-        }))
     }
 
     /// Try GNU-style "ZLIB" header decompression.
@@ -1006,6 +998,40 @@ pub trait SectionHeader: Debug + Pod {
         let data = self.data(endian, data)?;
         AttributesSection::new(endian, data)
     }
+
+    /// Parse the compression header if present.
+    ///
+    /// Returns the header, and the offset and size of the compressed section data
+    /// in the file.
+    ///
+    /// Returns `Ok(None)` if the section flags do not have `SHF_COMPRESSED`.
+    /// Returns `Err` for invalid values.
+    fn compression<'data, R: ReadRef<'data>>(
+        &self,
+        endian: Self::Endian,
+        data: R,
+    ) -> read::Result<
+        Option<(
+            &'data <Self::Elf as FileHeader>::CompressionHeader,
+            u64,
+            u64,
+        )>,
+    > {
+        if (self.sh_flags(endian).into() & u64::from(elf::SHF_COMPRESSED)) == 0 {
+            return Ok(None);
+        }
+        let (section_offset, section_size) = self
+            .file_range(endian)
+            .read_error("Invalid ELF compressed section type")?;
+        let mut offset = section_offset;
+        let header = data
+            .read::<<Self::Elf as FileHeader>::CompressionHeader>(&mut offset)
+            .read_error("Invalid ELF compressed section offset")?;
+        let compressed_size = section_size
+            .checked_sub(offset - section_offset)
+            .read_error("Invalid ELF compressed section size")?;
+        Ok(Some((header, offset, compressed_size)))
+    }
 }
 
 impl<Endian: endian::Endian> SectionHeader for elf::SectionHeader32<Endian> {

src/read/mod.rs

@@ -632,6 +632,10 @@ pub enum CompressionFormat {
     ///
     /// Used for ELF compression and GNU compressed debug information.
     Zlib,
+    /// Zstandard.
+    ///
+    /// Used for ELF compression.
+    Zstandard,
 }
 
 /// A range in a file that may be compressed.
@@ -731,6 +735,25 @@ impl<'data> CompressedData<'data> {
                     .read_error("Invalid zlib compressed data")?;
                 Ok(Cow::Owned(decompressed))
             }
+            #[cfg(feature = "compression")]
+            CompressionFormat::Zstandard => {
+                use core::convert::TryInto;
+                use std::io::Read;
+                let size = self
+                    .uncompressed_size
+                    .try_into()
+                    .ok()
+                    .read_error("Uncompressed data size is too large.")?;
+                let mut decompressed = Vec::with_capacity(size);
+                let mut decoder = ruzstd::StreamingDecoder::new(self.data)
+                    .ok()
+                    .read_error("Invalid zstd compressed data")?;
+                decoder
+                    .read_to_end(&mut decompressed)
+                    .ok()
+                    .read_error("Invalid zstd compressed data")?;
+                Ok(Cow::Owned(decompressed))
+            }
             _ => Err(Error("Unsupported compressed data.")),
         }
     }