Parquet: Verify 32-bit CRC checksum when decoding pages

parquet/Cargo.toml

@@ -96,7 +96,7 @@ zstd-sys = { version = ">=2.0.0, <2.0.14", default-features = false }
 all-features = true
 
 [features]
-default = ["arrow", "snap", "brotli", "flate2", "lz4", "zstd", "base64"]
+default = ["arrow", "snap", "brotli", "flate2", "lz4", "zstd", "base64", "crc"]
 # Enable lz4
 lz4 = ["lz4_flex"]
 # Enable arrow reader/writer APIs
@@ -117,6 +117,8 @@ object_store = ["dep:object_store", "async"]
 zstd = ["dep:zstd", "zstd-sys"]
 # Display memory in example/write_parquet.rs
 sysinfo = ["dep:sysinfo"]
+# Verify 32-bit CRC checksum when decoding parquet pages
+crc = ["dep:crc32fast"]
 
 [[example]]
 name = "read_parquet"
@@ -215,3 +217,4 @@ harness = false
 
 [lib]
 bench = false
+

parquet/src/errors.rs

@@ -43,6 +43,7 @@ pub enum ParquetError {
     /// Returned when reading into arrow or writing from arrow.
     ArrowError(String),
     IndexOutOfBound(usize, usize),
+    Crc32Mismatch,
     /// An external error variant
     External(Box<dyn Error + Send + Sync>),
 }
@@ -60,6 +61,9 @@ impl std::fmt::Display for ParquetError {
             ParquetError::IndexOutOfBound(index, ref bound) => {
                 write!(fmt, "Index {index} out of bound: {bound}")
             }
+            ParquetError::Crc32Mismatch => {
+                write!(fmt, "Page CRC checksum mismatch")
+            }
             ParquetError::External(e) => write!(fmt, "External: {e}"),
         }
     }

parquet/src/file/serialized_reader.rs

@@ -391,6 +391,15 @@ pub(crate) fn decode_page(
     physical_type: Type,
     decompressor: Option<&mut Box<dyn Codec>>,
 ) -> Result<Page> {
+    // Verify the 32-bit CRC checksum of the page
+    #[cfg(feature = "crc")]
+    if let Some(expected_crc) = page_header.crc {
+        let crc = crc32fast::hash(&buffer);
+        if crc != expected_crc as u32 {
+            return Err(ParquetError::Crc32Mismatch);
+        }
+    }
+
     // When processing data page v2, depending on enabled compression for the
     // page, we should account for uncompressed data ('offset') of
     // repetition and definition levels.