SMB performance optimization: pread/pwrite, tokio::sync::Mutex, direct response, fast-path

- VfsFile trait: add read_at()/write_at() with seek+read default impl
- LocalFs: override with real pread/pwrite (FileExt::read_at/write_at) — 1 syscall vs 2
- smb_server_backend: use read_at/write_at + tokio::sync::Mutex (non-blocking async)
- read handler: build response directly, avoid Bytes→Vec<u8> copy + intermediate struct
- oplock break: fast-path skip when ≤1 open entry (single-user scenario)

markbase-core/src/vfs/local_fs.rs

@@ -3,7 +3,7 @@ use super::util;
 use super::{VfsAce, VfsAceFlag, VfsAceMask, VfsAceType, VfsAcl, VfsBackend, VfsDirEntry, VfsError, VfsFile, VfsPreviousVersion, VfsQuota, VfsQuotaUsage, VfsSnapshotInfo, VfsStat};
 use std::fs::{self, File, OpenOptions};
 use std::io::{Read, Seek, SeekFrom, Write};
-use std::os::unix::fs::{MetadataExt, PermissionsExt};
+use std::os::unix::fs::{FileExt, MetadataExt, PermissionsExt};
 use std::path::{Path, PathBuf};
 use std::time::SystemTime;
 
@@ -42,6 +42,14 @@ impl VfsFile for LocalFile {
         self.file.seek(pos).map_err(|e| VfsError::Io(e.to_string()))
     }
 
+    fn read_at(&mut self, buf: &mut [u8], offset: u64) -> Result<usize, VfsError> {
+        self.file.read_at(buf, offset).map_err(|e| VfsError::Io(e.to_string()))
+    }
+
+    fn write_at(&mut self, buf: &[u8], offset: u64) -> Result<usize, VfsError> {
+        self.file.write_at(buf, offset).map_err(|e| VfsError::Io(e.to_string()))
+    }
+
     fn flush(&mut self) -> Result<(), VfsError> {
         self.file.flush().map_err(|e| VfsError::Io(e.to_string()))
     }

markbase-core/src/vfs/mod.rs

@@ -103,6 +103,20 @@ pub trait VfsFile: Send {
     fn stat(&mut self) -> Result<VfsStat, VfsError>;
     fn set_len(&mut self, size: u64) -> Result<(), VfsError>;
 
+    /// Read at `offset` without changing the seek position (like pread).
+    /// Default implementation does seek + read.
+    fn read_at(&mut self, buf: &mut [u8], offset: u64) -> Result<usize, VfsError> {
+        self.seek(std::io::SeekFrom::Start(offset))?;
+        self.read(buf)
+    }
+
+    /// Write at `offset` without changing the seek position (like pwrite).
+    /// Default implementation does seek + write.
+    fn write_at(&mut self, buf: &[u8], offset: u64) -> Result<usize, VfsError> {
+        self.seek(std::io::SeekFrom::Start(offset))?;
+        self.write(buf)
+    }
+
     /// Write all bytes (convenience, default loops write() until done)
     fn write_all(&mut self, mut buf: &[u8]) -> Result<(), VfsError> {
         while !buf.is_empty() {

markbase-core/src/vfs/smb_server_backend.rs

@@ -1,7 +1,7 @@
 use std::path::{Path, PathBuf};
 use std::sync::Arc;
-use std::sync::Mutex;
 use std::time::SystemTime;
+use tokio::sync::Mutex;
 
 use async_trait::async_trait;
 use bytes::Bytes;
@@ -211,11 +211,9 @@ impl Handle for VfsHandle {
     async fn read(&self, offset: u64, len: u32) -> Result<Bytes, SmbError> {
         match self {
             Self::File { file, .. } => {
-                let mut file = file.lock().unwrap();
+                let mut file = file.lock().await;
-                file.seek(std::io::SeekFrom::Start(offset))
-                    .map_err(vfs_error_to_io)?;
                 let mut buf = vec![0u8; len as usize];
-                let n = file.read(&mut buf).map_err(map_error)?;
+                let n = file.read_at(&mut buf, offset).map_err(map_error)?;
                 buf.truncate(n);
                 Ok(Bytes::from(buf))
             }
@@ -226,10 +224,8 @@ impl Handle for VfsHandle {
     async fn write(&self, offset: u64, data: &[u8]) -> Result<u32, SmbError> {
         match self {
             Self::File { file, .. } => {
-                let mut file = file.lock().unwrap();
+                let mut file = file.lock().await;
-                file.seek(std::io::SeekFrom::Start(offset))
+                let n = file.write_at(data, offset).map_err(map_error)?;
-                    .map_err(vfs_error_to_io)?;
-                let n = file.write(data).map_err(map_error)?;
                 Ok(n as u32)
             }
             Self::Directory { .. } => Err(SmbError::NotSupported),
@@ -239,7 +235,7 @@ impl Handle for VfsHandle {
     async fn flush(&self) -> Result<(), SmbError> {
         match self {
             Self::File { file, .. } => {
-                let mut file = file.lock().unwrap();
+                let mut file = file.lock().await;
                 file.flush().map_err(map_error)
             }
             Self::Directory { .. } => Ok(()),
@@ -249,7 +245,7 @@ impl Handle for VfsHandle {
     async fn stat(&self) -> Result<FileInfo, SmbError> {
         match self {
             Self::File { file, path, .. } => {
-                let mut f = file.lock().unwrap();
+                let mut f = file.lock().await;
                 let vfs_stat = f.stat().map_err(map_error)?;
                 Ok(vfs_stat_to_file_info(&vfs_stat, "", path))
             }
@@ -278,7 +274,7 @@ impl Handle for VfsHandle {
     async fn truncate(&self, len: u64) -> Result<(), SmbError> {
         match self {
             Self::File { file, .. } => {
-                let mut file = file.lock().unwrap();
+                let mut file = file.lock().await;
                 file.set_len(len).map_err(map_error)
             }
             Self::Directory { .. } => Err(SmbError::NotSupported),

vendor/smb-server/src/backend.rs

@@ -207,8 +207,8 @@ pub trait Handle: Send + Sync {
     /// Write `data` at `offset`. Returns bytes written.
     async fn write(&self, offset: u64, data: &[u8]) -> SmbResult<u32>;
 
-    /// Write owned `data` at `offset`. Backends that need ownership across a
+    /// Write owned `data` at `offset`. Backends needing ownership across a
-    /// blocking boundary can override this to avoid an extra copy.
+    /// blocking boundary should override to avoid an extra copy.
     async fn write_owned(&self, offset: u64, data: Vec<u8>) -> SmbResult<u32> {
         self.write(offset, &data).await
     }

vendor/smb-server/src/handlers/read.rs

@@ -91,16 +91,15 @@ pub async fn handle(
     if bytes.is_empty() && req.length > 0 {
         return HandlerResponse::err(ntstatus::STATUS_END_OF_FILE);
     }
-    let resp = ReadResponse {
+    // Build response directly to avoid Bytes→Vec<u8> copy and intermediate struct
-        structure_size: 17,
+    let data_len = bytes.len() as u32;
-        data_offset: ReadResponse::STANDARD_DATA_OFFSET,
+    let mut buf = Vec::with_capacity(16 + bytes.len());
-        reserved: 0,
+    buf.extend_from_slice(&17u16.to_le_bytes());      // structure_size
-        data_length: bytes.len() as u32,
+    buf.push(ReadResponse::STANDARD_DATA_OFFSET);      // data_offset
-        data_remaining: 0,
+    buf.push(0);                                       // reserved
-        flags: 0,
+    buf.extend_from_slice(&data_len.to_le_bytes());    // data_length
-        data: bytes.to_vec(),
+    buf.extend_from_slice(&0u32.to_le_bytes());        // data_remaining
-    };
+    buf.extend_from_slice(&0u32.to_le_bytes());        // flags
-    let mut buf = Vec::new();
+    buf.extend_from_slice(&bytes);                     // data
-    resp.write_to(&mut buf).expect("encode");
     HandlerResponse::ok(buf)
 }

vendor/smb-server/src/oplock.rs

@@ -112,17 +112,23 @@ impl OplockManager {
         new_share_access: u32,
         new_granted_access: Access,
     ) -> Vec<OplockBreakNotification> {
+        // Fast-path: no entries or single entry can't conflict with itself
+        let entry_count = {
+            let file_opens = self.file_opens.read().await;
+            file_opens.get(path).map_or(0, |e| e.len())
+        };
+        if entry_count <= 1 {
+            return Vec::new();
+        }
+
         let mut notifications = Vec::new();
         let mut file_opens = self.file_opens.write().await;
 
         if let Some(entries) = file_opens.get_mut(path) {
             for entry in entries.iter_mut() {
-                // Check if new open conflicts with existing oplock
                 if !share_access_compatible(entry.share_access, new_share_access) {
-                    // Need to break the oplock
+                    let new_level = OplockLevel::Ii as u8;
-                    let new_level = OplockLevel::Ii as u8; // Downgrade to Level II
 
-                    // Build notification (MS-SMB2 §2.2.23.1)
                     notifications.push(OplockBreakNotification {
                         structure_size: 24,
                         oplock_level: new_level,
@@ -131,7 +137,6 @@ impl OplockManager {
                         file_id: entry.file_id,
                     });
 
-                    // Update entry's oplock level
                     entry.oplock_level = new_level;
                 }
             }
@@ -266,6 +271,11 @@ impl LeaseManager {
     
     /// Break lease when conflicting access occurs (MS-SMB2 §3.3.5.10).
     pub async fn break_lease(&self, requested_state: u32) -> Vec<LeaseBreakNotification> {
+        // Fast-path: no leases to break
+        if self.leases.read().await.is_empty() {
+            return Vec::new();
+        }
+
         let mut leases = self.leases.write().await;
         let mut notifications = Vec::new();