Abort multi part download if the object is modified during download (#3248)

Abort multi part download if the object is modified during download

Author: richardwang1124

gems/aws-sdk-s3/CHANGELOG.md

@@ -1,5 +1,6 @@
 Unreleased Changes
 ------------------
+* Issue - Abort multipart download if object is modified during download.
 
 1.186.0 (2025-05-12)
 ------------------

gems/aws-sdk-s3/lib/aws-sdk-s3/file_downloader.rb

@@ -43,7 +43,7 @@ def download(destination, options = {})
           when 'get_range'
             if @chunk_size
               resp = @client.head_object(@params)
-              multithreaded_get_by_ranges(resp.content_length)
+              multithreaded_get_by_ranges(resp.content_length, resp.etag)
             else
               msg = 'In :get_range mode, :chunk_size must be provided'
               raise ArgumentError, msg
@@ -71,26 +71,26 @@ def multipart_download
           if resp.content_length <= MIN_CHUNK_SIZE
             single_request
           else
-            multithreaded_get_by_ranges(resp.content_length)
+            multithreaded_get_by_ranges(resp.content_length, resp.etag)
           end
         else
           # partNumber is an option
           resp = @client.head_object(@params)
           if resp.content_length <= MIN_CHUNK_SIZE
             single_request
           else
-            compute_mode(resp.content_length, count)
+            compute_mode(resp.content_length, count, resp.etag)
           end
         end
       end
 
-      def compute_mode(file_size, count)
+      def compute_mode(file_size, count, etag)
         chunk_size = compute_chunk(file_size)
         part_size = (file_size.to_f / count.to_f).ceil
         if chunk_size < part_size
-          multithreaded_get_by_ranges(file_size)
+          multithreaded_get_by_ranges(file_size, etag)
         else
-          multithreaded_get_by_parts(count, file_size)
+          multithreaded_get_by_parts(count, file_size, etag)
         end
       end
 
@@ -122,7 +122,7 @@ def batches(chunks, mode)
         chunks.each_slice(@thread_count).to_a
       end
 
-      def multithreaded_get_by_ranges(file_size)
+      def multithreaded_get_by_ranges(file_size, etag)
         offset = 0
         default_chunk_size = compute_chunk(file_size)
         chunks = []
@@ -134,17 +134,17 @@ def multithreaded_get_by_ranges(file_size)
           chunks << Part.new(
             part_number: part_number,
             size: (progress-offset),
-            params: @params.merge(range: range)
+            params: @params.merge(range: range, if_match: etag)
           )
           part_number += 1
           offset = progress
         end
         download_in_threads(PartList.new(chunks), file_size)
       end
 
-      def multithreaded_get_by_parts(n_parts, total_size)
+      def multithreaded_get_by_parts(n_parts, total_size, etag)
         parts = (1..n_parts).map do |part|
-          Part.new(part_number: part, params: @params.merge(part_number: part))
+          Part.new(part_number: part, params: @params.merge(part_number: part, if_match: etag))
         end
         download_in_threads(PartList.new(parts), total_size)
       end

gems/aws-sdk-s3/spec/object/download_file_spec.rb

@@ -221,6 +221,58 @@ module S3
           end.to raise_error(Aws::Errors::ChecksumError)
         end
 
+        it 'does not download object when ETAG does not match during multipart get by ranges' do
+          allow(client).to receive(:head_object).with({
+            bucket: 'bucket',
+            key: 'single',
+            part_number: 1,
+          }).and_return(
+            client.stub_data(
+              :head_object,
+              content_length: 15 * one_meg,
+              parts_count: nil,
+              etag: 'test-etag'
+            )
+          )
+
+          client.stub_responses(:get_object, -> (ctx) {
+            expect(ctx.params[:if_match]).to eq('test-etag')
+            'PreconditionFailed'
+          })
+
+          thread = double(value: nil)
+          expect(Thread).to receive(:new).and_yield.and_return(thread)
+
+          expect do
+            single_obj.download_file(path)
+          end.to raise_error(Aws::S3::Errors::PreconditionFailed)
+        end
+
+        it 'does not download object when ETAG does not match during multipart get by parts' do
+          allow(client).to receive(:head_object).with({
+            bucket: 'bucket',
+            key: 'large'
+          }).and_return(
+            client.stub_data(
+              :head_object,
+              content_length: 20 * one_meg,
+              etag: 'test-etag'
+            )
+          )
+
+          client.stub_responses(:get_object, -> (ctx) {
+            expect(ctx.params[:if_match]).to eq('test-etag')
+            'PreconditionFailed'
+          })
+
+          thread = double(value: nil)
+          expect(Thread).to receive(:new).and_yield.and_return(thread)
+
+          expect do
+            large_obj.download_file(path)
+          end.to raise_error(Aws::S3::Errors::PreconditionFailed)
+        end
+
         it 'calls on_checksum_validated on single part' do
           callback_data = {called: 0}
           mutex = Mutex.new