fix: sandbox SQL local filesystem access (#8708)

* fix: sandbox SQL local filesystem access

Signed-off-by: jeremyhi <fengjiachun@gmail.com>

* fix: address local file sandbox review findings

Signed-off-by: jeremyhi <fengjiachun@gmail.com>

* fix: support Windows local copy paths

Signed-off-by: jeremyhi <fengjiachun@gmail.com>

* fix: improve sandbox path errors

Signed-off-by: jeremyhi <fengjiachun@gmail.com>

* refactor: simplify local path error context

Signed-off-by: jeremyhi <fengjiachun@gmail.com>

* perf: stream secure filesystem listings

Signed-off-by: jeremyhi <fengjiachun@gmail.com>

* style: derive local file access default

Signed-off-by: jeremyhi <fengjiachun@gmail.com>

* fix: improve local file access errors

Signed-off-by: jeremyhi <fengjiachun@gmail.com>

* fix: address local file access review findings

Signed-off-by: jeremyhi <fengjiachun@gmail.com>

* test: simplify local file access coverage

Signed-off-by: jeremyhi <fengjiachun@gmail.com>

* fix: harden sandboxed local file backends

Signed-off-by: jeremyhi <fengjiachun@gmail.com>

* fix: reject directory copy targets before creation

Signed-off-by: jeremyhi <fengjiachun@gmail.com>

* fix: avoid implicit string clone in file table listing

Signed-off-by: jeremyhi <fengjiachun@gmail.com>

---------

Signed-off-by: jeremyhi <fengjiachun@gmail.com>

src/datanode/Cargo.toml

@@ -20,6 +20,7 @@ client.workspace = true
 common-base.workspace = true
 common-catalog.workspace = true
 common-config.workspace = true
+common-datasource.workspace = true
 common-error.workspace = true
 common-function.workspace = true
 common-greptimedb-telemetry.workspace = true

src/datanode/src/config.rs

@@ -39,6 +39,11 @@ use servers::http::HttpOptions;
 pub struct StorageConfig {
     /// The working directory of database
     pub data_home: String,
+    /// Root directory for standalone SQL access to local files.
+    ///
+    /// Defaults to `<data_home>/copy` when `data_home` is a local path.
+    /// Distributed deployments always disable SQL access to local files.
+    pub copy_root: Option<String>,
     #[serde(flatten)]
     pub store: ObjectStoreConfig,
     /// Object storage providers
@@ -56,6 +61,7 @@ impl Default for StorageConfig {
     fn default() -> Self {
         Self {
             data_home: DEFAULT_DATA_HOME.to_string(),
+            copy_root: None,
             store: ObjectStoreConfig::default(),
             providers: vec![],
         }

src/datanode/src/datanode.rs

@@ -19,6 +19,7 @@ use std::sync::Arc;
 use std::time::{Duration, Instant};
 
 use common_base::Plugins;
+use common_datasource::object_store::LocalFileAccess;
 use common_error::ext::BoxedError;
 use common_greptimedb_telemetry::GreptimeDBTelemetryTask;
 use common_meta::cache::{LayeredCacheRegistry, SchemaCacheRef, TableSchemaCacheRef};
@@ -164,6 +165,7 @@ pub struct DatanodeBuilder {
     cache_registry: Option<Arc<LayeredCacheRegistry>>,
     topic_stats_reporter: Option<Box<dyn TopicStatsReporter>>,
     open_regions_writable_override: Option<bool>,
+    local_file_access: LocalFileAccess,
     #[cfg(feature = "enterprise")]
     extension_range_provider_factory: Option<mito2::extension::BoxedExtensionRangeProviderFactory>,
 }
@@ -178,6 +180,7 @@ impl DatanodeBuilder {
             kv_backend,
             cache_registry: None,
             open_regions_writable_override: None,
+            local_file_access: LocalFileAccess::Disabled,
             #[cfg(feature = "enterprise")]
             extension_range_provider_factory: None,
             topic_stats_reporter: None,
@@ -198,6 +201,11 @@ impl DatanodeBuilder {
         self
     }
 
+    pub fn with_local_file_access(&mut self, local_file_access: LocalFileAccess) -> &mut Self {
+        self.local_file_access = local_file_access;
+        self
+    }
+
     pub fn kv_backend(&self) -> &KvBackendRef {
         &self.kv_backend
     }
@@ -528,6 +536,7 @@ impl DatanodeBuilder {
         let file_engine = FileRegionEngine::new(
             file_engine_config,
             object_store_manager.default_object_store().clone(), // TODO: implement custom storage for file engine
+            self.local_file_access.clone(),
         );
 
         Ok(vec![