diff --git a/crates/ecstore/src/bucket/quota/reservation.rs b/crates/ecstore/src/bucket/quota/reservation.rs
index 33aa5c876..d4a35c23e 100644
--- a/crates/ecstore/src/bucket/quota/reservation.rs
+++ b/crates/ecstore/src/bucket/quota/reservation.rs
@@ -38,6 +38,7 @@ const MAX_ORPHANS_REAPED_PER_WRITE: usize = 64;
 const MAX_ORPHAN_PROBES_PER_WRITE: usize = 128;
 const ORPHAN_PROBE_CONCURRENCY: usize = 32;
 const EVENT_QUOTA_LEDGER_SETTLEMENT: &str = "quota_ledger_settlement";
+const EVENT_QUOTA_ADMISSION: &str = "quota_admission";
 const LOG_COMPONENT_ECSTORE: &str = "ecstore";
 const LOG_SUBSYSTEM_QUOTA: &str = "quota";
 
@@ -513,6 +514,7 @@ pub(crate) async fn begin(
         .as_ref()
         .is_some_and(|quota| quota.has_unsupported_reservation_protocol())
     {
+        log_admission_rejected(bucket, object, "unsupported_reservation_protocol");
         return Err(StorageError::PartMissingOrCorrupt);
     }
     let durable_quota = quota.as_ref().filter(|quota| quota.uses_durable_reservations());
@@ -529,7 +531,16 @@ pub(crate) async fn begin(
         Some(quota) => match (quota.quota, snapshot_admission) {
             (Some(limit), Some(admission)) if admission.quota_limit() == limit => Some(admission),
             (Some(_), None) if data_movement => None,
-            (Some(_), _) => return Err(StorageError::PartMissingOrCorrupt),
+            (Some(_), _) => {
+                // A snapshot-protocol quota requires the request handler's
+                // admission on every write. Missing or mismatched admission
+                // means a caller rebuilt `ObjectOptions` without carrying it
+                // over (rustfs/rustfs#7674 lost it on CopyObject); fail closed
+                // but leave a diagnosable trace, because the storage error is
+                // the generic `PartMissingOrCorrupt`.
+                log_admission_rejected(bucket, object, "snapshot_quota_admission_missing");
+                return Err(StorageError::PartMissingOrCorrupt);
+            }
             (None, _) => None,
         },
         None => None,
@@ -904,6 +915,18 @@ pub fn fail_next_quota_ledger_save_for_test() {
     FAIL_NEXT_LEDGER_SAVE.store(true, std::sync::atomic::Ordering::SeqCst);
 }
 
+fn log_admission_rejected(bucket: &str, object: &str, state: &'static str) {
+    warn!(
+        event = EVENT_QUOTA_ADMISSION,
+        component = LOG_COMPONENT_ECSTORE,
+        subsystem = LOG_SUBSYSTEM_QUOTA,
+        state,
+        bucket = %bucket,
+        object = %object,
+        "quota admission rejected the write before commit"
+    );
+}
+
 fn log_deferred_settlement(data: &LedgerReservationData, state: &'static str, err: &StorageError) {
     warn!(
         event = EVENT_QUOTA_LEDGER_SETTLEMENT,
diff --git a/crates/ecstore/src/core/sets.rs b/crates/ecstore/src/core/sets.rs
index 487d85a77..a83c4d699 100644
--- a/crates/ecstore/src/core/sets.rs
+++ b/crates/ecstore/src/core/sets.rs
@@ -790,6 +790,7 @@ impl crate::storage_api_contracts::object::ObjectOperations for Sets {
             version_id: dst_opts.version_id.clone(),
             mod_time: dst_opts.mod_time,
             http_preconditions: dst_opts.http_preconditions.clone(),
+            quota_admission: dst_opts.quota_admission,
             ..Default::default()
         };
 
diff --git a/crates/ecstore/src/store/init.rs b/crates/ecstore/src/store/init.rs
index baee8e787..115121e15 100644
--- a/crates/ecstore/src/store/init.rs
+++ b/crates/ecstore/src/store/init.rs
@@ -3149,6 +3149,120 @@ mod tests {
         shutdown.cancel();
     }
 
+    /// rustfs/rustfs#7674: a bucket carrying a legacy snapshot-protocol quota
+    /// (written before durable reservations existed, so `quota.json` has no
+    /// `reservation_protocol`) must accept a cross-key CopyObject whose
+    /// destination options carry the handler's quota admission. The storage
+    /// layer fails closed with `PartMissingOrCorrupt` when a quota-enforced
+    /// bucket sees a write without admission, so dropping the admission while
+    /// rebuilding the destination options turns every copy into a
+    /// deterministic "part missing or corrupt" failure even though the source
+    /// object is perfectly readable.
+    #[cfg(feature = "test-util")]
+    #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+    #[serial_test::serial(storage_class_env)]
+    async fn copy_object_forwards_quota_admission_on_legacy_snapshot_quota_bucket() {
+        let temp_dir = tempfile::tempdir().expect("create legacy quota copy store dir");
+        let (ctx, store, shutdown) =
+            without_storage_class_env(build_isolated_test_store(temp_dir.path(), "legacy-quota-copy", &[1])).await;
+        crate::bucket::metadata_sys::init_bucket_metadata_sys(Arc::clone(&store), Vec::new()).await;
+
+        let bucket = format!("legacy-quota-copy-{}", Uuid::new_v4());
+        let source_object = "docker/registry/v2/repositories/example/_uploads/upload-id/data";
+        let target_object = "docker/registry/v2/blobs/sha256/a5/digest/data";
+        let payload = vec![0x5A; 8178];
+        let quota_limit = 1u64 << 30;
+
+        store
+            .make_bucket(&bucket, &MakeBucketOptions::default())
+            .await
+            .expect("create bucket for legacy quota copy");
+        // Legacy quota shape: no `reservation_protocol`, so the storage layer
+        // requires the handler-supplied snapshot admission on every write.
+        let legacy_quota = format!(r#"{{"quota":{quota_limit},"quota_type":"Hard"}}"#);
+        crate::bucket::metadata_sys::update_in(
+            &ctx,
+            &bucket,
+            crate::bucket::metadata::BUCKET_QUOTA_CONFIG_FILE,
+            legacy_quota.into_bytes(),
+        )
+        .await
+        .expect("persist legacy snapshot quota");
+        let (quota, _, _) = crate::bucket::metadata_sys::get_quota_config_and_incarnation_from_disk_in(&ctx, &bucket)
+            .await
+            .expect("legacy quota should load");
+        let quota = quota.expect("legacy quota must be persisted");
+        assert_eq!(quota.quota, Some(quota_limit));
+        assert!(!quota.uses_durable_reservations(), "fixture must stay on the snapshot protocol");
+
+        let mut write_opts = ObjectOptions::default();
+        assert!(write_opts.set_quota_admission(0, quota_limit));
+
+        let upload = store
+            .new_multipart_upload(&bucket, source_object, &write_opts)
+            .await
+            .expect("create source multipart upload");
+        let mut part_reader = PutObjReader::from_vec(payload.clone());
+        let part = store
+            .put_object_part(&bucket, source_object, &upload.upload_id, 1, &mut part_reader, &write_opts)
+            .await
+            .expect("stage multipart source part");
+        store
+            .clone()
+            .complete_multipart_upload(
+                &bucket,
+                source_object,
+                &upload.upload_id,
+                vec![crate::storage_api_contracts::multipart::CompletePart {
+                    part_num: part.part_num,
+                    etag: part.etag,
+                    ..Default::default()
+                }],
+                &write_opts,
+            )
+            .await
+            .expect("complete the multipart source under the legacy quota");
+
+        let source_reader = store
+            .get_object_reader(&bucket, source_object, None, HeaderMap::new(), &ObjectOptions::default())
+            .await
+            .expect("completed multipart source should be readable");
+        let mut copy_info = source_reader.object_info.clone();
+        let actual_size = copy_info.get_actual_size().expect("copy source logical size should resolve");
+        assert_eq!(actual_size, payload.len() as i64);
+        let copy_reader = rustfs_rio::HashReader::from_stream(source_reader.stream, actual_size, actual_size, None, None, false)
+            .expect("copy source hash reader should build");
+        copy_info.put_object_reader = Some(PutObjReader::new(copy_reader));
+
+        let mut dst_opts = ObjectOptions::default();
+        assert!(dst_opts.set_quota_admission(payload.len() as u64, quota_limit));
+        store
+            .copy_object(
+                &bucket,
+                source_object,
+                &bucket,
+                target_object,
+                &mut copy_info,
+                &ObjectOptions::default(),
+                &dst_opts,
+            )
+            .await
+            .expect("CopyObject must forward the handler quota admission to the destination write");
+
+        let mut target_reader = store
+            .get_object_reader(&bucket, target_object, None, HeaderMap::new(), &ObjectOptions::default())
+            .await
+            .expect("copied target should be readable");
+        let mut target_body = Vec::new();
+        target_reader
+            .stream
+            .read_to_end(&mut target_body)
+            .await
+            .expect("target body should stream");
+        assert_eq!(target_body, payload);
+        shutdown.cancel();
+    }
+
     #[cfg(feature = "test-util")]
     #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
     #[serial_test::serial(storage_class_env)]
diff --git a/crates/ecstore/src/store/object.rs b/crates/ecstore/src/store/object.rs
index c5fbe2d26..a801ffea1 100644
--- a/crates/ecstore/src/store/object.rs
+++ b/crates/ecstore/src/store/object.rs
@@ -4564,6 +4564,7 @@ impl ECStore {
                     namespace_lock_fence: dst_opts.namespace_lock_fence.clone(),
                     bucket_lifecycle_lock_fence: dst_opts.bucket_lifecycle_lock_fence.clone(),
                     object_lock_config_snapshot: dst_opts.object_lock_config_snapshot.clone(),
+                    quota_admission: dst_opts.quota_admission,
                     ..Default::default()
                 };
                 if !self.single_pool() {
@@ -4606,6 +4607,7 @@ impl ECStore {
                         namespace_lock_fence: dst_opts.namespace_lock_fence.clone(),
                         bucket_lifecycle_lock_fence: dst_opts.bucket_lifecycle_lock_fence.clone(),
                         object_lock_config_snapshot: dst_opts.object_lock_config_snapshot.clone(),
+                        quota_admission: dst_opts.quota_admission,
                         ..Default::default()
                     };
                     if !self.single_pool() {
@@ -4658,6 +4660,7 @@ impl ECStore {
             namespace_lock_fence: dst_opts.namespace_lock_fence.clone(),
             bucket_lifecycle_lock_fence: dst_opts.bucket_lifecycle_lock_fence.clone(),
             object_lock_config_snapshot: dst_opts.object_lock_config_snapshot.clone(),
+            quota_admission: dst_opts.quota_admission,
             ..Default::default()
         };
         if !self.single_pool() {