Fixes out-of-bounds access for small files in unzip

cpp/src/io/comp/uncomp.cpp

@@ -193,16 +193,18 @@ bool OpenZipArchive(zip_archive_s *dst, const uint8_t *raw, size_t len)
   memset(dst, 0, sizeof(zip_archive_s));
   // Find the end of central directory
   if (len >= sizeof(zip_eocd_s) + 2) {
-    for (size_t i = len - sizeof(zip_eocd_s) - 2; i + sizeof(zip_eocd_s) + 2 + 0xffff >= len; i--) {
+    for (ptrdiff_t i = len - sizeof(zip_eocd_s) - 2;
+         i + sizeof(zip_eocd_s) + 2 + 0xffff >= len && i >= 0;
+         i--) {
       const zip_eocd_s *eocd = reinterpret_cast<zip_eocd_s const *>(raw + i);
       if (eocd->sig == 0x06054b50 &&
           eocd->disk_id == eocd->start_disk  // multi-file archives not supported
           && eocd->num_entries == eocd->total_entries &&
           eocd->cdir_size >= sizeof(zip_cdfh_s) * eocd->num_entries && eocd->cdir_offset < len &&
-          i + *reinterpret_cast<const uint16_t *>(eocd + 1) <= len) {
+          i + *reinterpret_cast<const uint16_t *>(eocd + 1) <= static_cast<ptrdiff_t>(len)) {
         const zip_cdfh_s *cdfh = reinterpret_cast<const zip_cdfh_s *>(raw + eocd->cdir_offset);
         dst->eocd              = eocd;
-        if (i >= sizeof(zip64_eocdl)) {
+        if (i >= static_cast<ptrdiff_t>(sizeof(zip64_eocdl))) {
           const zip64_eocdl *eocdl =
             reinterpret_cast<const zip64_eocdl *>(raw + i - sizeof(zip64_eocdl));
           if (eocdl->sig == 0x07064b50) { dst->eocdl = eocdl; }

cpp/tests/io/csv_test.cpp

@@ -1081,6 +1081,48 @@ TEST_F(CsvReaderTest, ArrowFileSource)
   expect_column_data_equal(std::vector<int8_t>{9, 8, 7, 6, 5, 4, 3, 2}, view.column(0));
 }
 
+TEST_F(CsvReaderTest, ZippedCSV)
+{
+  // input is the following CSV zip'ed:
+  // "year,name,description\n1997,Ford,"Super, luxurious truck"\n1997,Ford,"Super, luxurious truck"
+  constexpr uint8_t compressed[] = {
+    0x50, 0x4b, 0x03, 0x04, 0x14, 0x00, 0x00, 0x00, 0x08, 0x00, 0x4e, 0x5b, 0xcf, 0x52, 0xaf, 0x6f,
+    0xfa, 0xe1, 0x3c, 0x00, 0x00, 0x00, 0x5c, 0x00, 0x00, 0x00, 0x0e, 0x00, 0x1c, 0x00, 0x7a, 0x69,
+    0x70, 0x65, 0x78, 0x61, 0x6d, 0x70, 0x6c, 0x65, 0x2e, 0x63, 0x73, 0x76, 0x55, 0x54, 0x09, 0x00,
+    0x03, 0x44, 0x72, 0xc8, 0x60, 0x45, 0x72, 0xc8, 0x60, 0x75, 0x78, 0x0b, 0x00, 0x01, 0x04, 0xf7,
+    0x01, 0x00, 0x00, 0x04, 0x14, 0x00, 0x00, 0x00, 0xab, 0x4c, 0x4d, 0x2c, 0xd2, 0xc9, 0x4b, 0xcc,
+    0x4d, 0xd5, 0x49, 0x49, 0x2d, 0x4e, 0x2e, 0xca, 0x2c, 0x28, 0xc9, 0xcc, 0xcf, 0xe3, 0x32, 0xb4,
+    0xb4, 0x34, 0xd7, 0x71, 0xcb, 0x2f, 0x4a, 0xd1, 0x51, 0x0a, 0x2e, 0x2d, 0x48, 0x2d, 0xd2, 0x51,
+    0xc8, 0x29, 0xad, 0x28, 0x2d, 0xca, 0xcc, 0x2f, 0x2d, 0x56, 0x28, 0x29, 0x2a, 0x4d, 0xce, 0x56,
+    0x22, 0x46, 0x09, 0x00, 0x50, 0x4b, 0x01, 0x02, 0x1e, 0x03, 0x14, 0x00, 0x00, 0x00, 0x08, 0x00,
+    0x4e, 0x5b, 0xcf, 0x52, 0xaf, 0x6f, 0xfa, 0xe1, 0x3c, 0x00, 0x00, 0x00, 0x5c, 0x00, 0x00, 0x00,
+    0x0e, 0x00, 0x18, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0xa4, 0x81, 0x00, 0x00,
+    0x00, 0x00, 0x7a, 0x69, 0x70, 0x65, 0x78, 0x61, 0x6d, 0x70, 0x6c, 0x65, 0x2e, 0x63, 0x73, 0x76,
+    0x55, 0x54, 0x05, 0x00, 0x03, 0x44, 0x72, 0xc8, 0x60, 0x75, 0x78, 0x0b, 0x00, 0x01, 0x04, 0xf7,
+    0x01, 0x00, 0x00, 0x04, 0x14, 0x00, 0x00, 0x00, 0x50, 0x4b, 0x05, 0x06, 0x00, 0x00, 0x00, 0x00,
+    0x01, 0x00, 0x01, 0x00, 0x54, 0x00, 0x00, 0x00, 0x84, 0x00, 0x00, 0x00, 0x00, 0x00,
+  };
+
+  const auto filepath = temp_env->get_temp_dir() + "zipped_example.zip";
+  {
+    std::ofstream outfile(filepath, std::ofstream::out | std::ofstream::binary);
+    outfile.write(reinterpret_cast<const char*>(compressed), sizeof(compressed));
+  }
+
+  cudf_io::csv_reader_options in_opts =
+    cudf_io::csv_reader_options::builder(cudf_io::source_info{filepath})
+      .compression(cudf::io::compression_type::ZIP);
+  const auto result      = cudf_io::read_csv(in_opts);
+  const auto result_view = result.tbl->view();
+
+  auto year_col  = std::vector<int64_t>{1997L, 1997L};
+  auto name_col  = std::vector<std::string>{"Ford", "Ford"};
+  auto descr_col = std::vector<std::string>{"Super, luxurious truck", "Super, luxurious truck"};
+  expect_column_data_equal(year_col, result_view.column(0));
+  expect_column_data_equal(name_col, result_view.column(1));
+  expect_column_data_equal(descr_col, result_view.column(2));
+}
+
 TEST_F(CsvReaderTest, InvalidFloatingPoint)
 {
   const auto filepath = temp_env->get_temp_dir() + "InvalidFloatingPoint.csv";

python/cudf/cudf/tests/test_csv.py

@@ -979,6 +979,30 @@ def test_csv_reader_filepath_or_buffer(tmpdir, path_or_buf, src):
     assert_eq(expect, got)
 
 
+def test_small_zip(tmpdir):
+    df = pd.DataFrame(
+        {
+            "a": [1997] * 2,
+            "b": ["Ford"] * 2,
+            "c": ["Super, luxurious truck"] * 2,
+        }
+    )
+
+    fname = tmpdir.join("small_zip_file.zip")
+    df.to_csv(fname, index=False)
+
+    df = pd.DataFrame(
+        {
+            "a": [1997] * 2,
+            "b": ["Ford"] * 2,
+            "c": ["Superd, luxurious truck"] * 2,
+        }
+    )
+
+    got = cudf.read_csv(fname)
+    assert_eq(df, got)
+
+
 def test_csv_reader_carriage_return(tmpdir):
     rows = 1000
     names = ["int_row", "int_double_row"]