ksmbd: add support for read compound

MacOS sends a compound request including read to the server
(e.g. open-read-close). So far, ksmbd has not handled read as
a compound request. For compatibility between ksmbd and an OS that
supports SMB, This patch provides compound support for read requests.
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
Signed-off-by: Steve French <stfrench@microsoft.com>

fs/smb/server/auth.c

@@ -1029,11 +1029,15 @@ static struct scatterlist *ksmbd_init_sg(struct kvec *iov, unsigned int nvec,
 {
 	struct scatterlist *sg;
 	unsigned int assoc_data_len = sizeof(struct smb2_transform_hdr) - 20;
-	int i, nr_entries[3] = {0}, total_entries = 0, sg_idx = 0;
+	int i, *nr_entries, total_entries = 0, sg_idx = 0;
 
 	if (!nvec)
 		return NULL;
 
+	nr_entries = kcalloc(nvec, sizeof(int), GFP_KERNEL);
+	if (!nr_entries)
+		return NULL;
+
 	for (i = 0; i < nvec - 1; i++) {
 		unsigned long kaddr = (unsigned long)iov[i + 1].iov_base;
 
@@ -1051,8 +1055,10 @@ static struct scatterlist *ksmbd_init_sg(struct kvec *iov, unsigned int nvec,
 	total_entries += 2;
 
 	sg = kmalloc_array(total_entries, sizeof(struct scatterlist), GFP_KERNEL);
-	if (!sg)
+	if (!sg) {
+		kfree(nr_entries);
 		return NULL;
+	}
 
 	sg_init_table(sg, total_entries);
 	smb2_sg_set_buf(&sg[sg_idx++], iov[0].iov_base + 24, assoc_data_len);
@@ -1086,6 +1092,7 @@ static struct scatterlist *ksmbd_init_sg(struct kvec *iov, unsigned int nvec,
 		}
 	}
 	smb2_sg_set_buf(&sg[sg_idx], sign, SMB2_SIGNATURE_SIZE);
+	kfree(nr_entries);
 	return sg;
 }
 

fs/smb/server/connection.c

@@ -123,28 +123,22 @@ void ksmbd_conn_enqueue_request(struct ksmbd_work *work)
 	}
 }
 
-int ksmbd_conn_try_dequeue_request(struct ksmbd_work *work)
+void ksmbd_conn_try_dequeue_request(struct ksmbd_work *work)
 {
 	struct ksmbd_conn *conn = work->conn;
-	int ret = 1;
 
 	if (list_empty(&work->request_entry) &&
 	    list_empty(&work->async_request_entry))
-		return 0;
+		return;
 
-	if (!work->multiRsp)
 	atomic_dec(&conn->req_running);
-	if (!work->multiRsp) {
 	spin_lock(&conn->request_lock);
 	list_del_init(&work->request_entry);
 	spin_unlock(&conn->request_lock);
 	if (work->asynchronous)
 		release_async_work(work);
-		ret = 0;
-	}
 
 	wake_up_all(&conn->req_running_q);
-	return ret;
 }
 
 void ksmbd_conn_lock(struct ksmbd_conn *conn)
@@ -193,39 +187,20 @@ void ksmbd_conn_wait_idle(struct ksmbd_conn *conn, u64 sess_id)
 int ksmbd_conn_write(struct ksmbd_work *work)
 {
 	struct ksmbd_conn *conn = work->conn;
-	size_t len = 0;
 	int sent;
-	struct kvec iov[3];
-	int iov_idx = 0;
 
 	if (!work->response_buf) {
 		pr_err("NULL response header\n");
 		return -EINVAL;
 	}
 
-	if (work->tr_buf) {
-		iov[iov_idx] = (struct kvec) { work->tr_buf,
-				sizeof(struct smb2_transform_hdr) + 4 };
-		len += iov[iov_idx++].iov_len;
-	}
-
-	if (work->aux_payload_sz) {
-		iov[iov_idx] = (struct kvec) { work->response_buf, work->resp_hdr_sz };
-		len += iov[iov_idx++].iov_len;
-		iov[iov_idx] = (struct kvec) { work->aux_payload_buf, work->aux_payload_sz };
-		len += iov[iov_idx++].iov_len;
-	} else {
-		if (work->tr_buf)
-			iov[iov_idx].iov_len = work->resp_hdr_sz;
-		else
-			iov[iov_idx].iov_len = get_rfc1002_len(work->response_buf) + 4;
-		iov[iov_idx].iov_base = work->response_buf;
-		len += iov[iov_idx++].iov_len;
-	}
+	if (work->send_no_response)
+		return 0;
 
 	ksmbd_conn_lock(conn);
-	sent = conn->transport->ops->writev(conn->transport, &iov[0],
-					iov_idx, len,
+	sent = conn->transport->ops->writev(conn->transport, work->iov,
+			work->iov_cnt,
+			get_rfc1002_len(work->iov[0].iov_base) + 4,
 			work->need_invalidate_rkey,
 			work->remote_key);
 	ksmbd_conn_unlock(conn);

fs/smb/server/connection.h

@@ -158,7 +158,7 @@ int ksmbd_conn_rdma_write(struct ksmbd_conn *conn,
 			  struct smb2_buffer_desc_v1 *desc,
 			  unsigned int desc_len);
 void ksmbd_conn_enqueue_request(struct ksmbd_work *work);
-int ksmbd_conn_try_dequeue_request(struct ksmbd_work *work);
+void ksmbd_conn_try_dequeue_request(struct ksmbd_work *work);
 void ksmbd_conn_init_server_callbacks(struct ksmbd_conn_ops *ops);
 int ksmbd_conn_handler_loop(void *p);
 int ksmbd_conn_transport_init(void);

fs/smb/server/ksmbd_work.c

@@ -27,18 +27,35 @@ struct ksmbd_work *ksmbd_alloc_work_struct(void)
 		INIT_LIST_HEAD(&work->async_request_entry);
 		INIT_LIST_HEAD(&work->fp_entry);
 		INIT_LIST_HEAD(&work->interim_entry);
+		INIT_LIST_HEAD(&work->aux_read_list);
+		work->iov_alloc_cnt = 4;
+		work->iov = kcalloc(work->iov_alloc_cnt, sizeof(struct kvec),
+				    GFP_KERNEL);
+		if (!work->iov) {
+			kmem_cache_free(work_cache, work);
+			work = NULL;
+		}
 	}
 	return work;
 }
 
 void ksmbd_free_work_struct(struct ksmbd_work *work)
 {
+	struct aux_read *ar, *tmp;
+
 	WARN_ON(work->saved_cred != NULL);
 
 	kvfree(work->response_buf);
-	kvfree(work->aux_payload_buf);
+
+	list_for_each_entry_safe(ar, tmp, &work->aux_read_list, entry) {
+		kvfree(ar->buf);
+		list_del(&ar->entry);
+		kfree(ar);
+	}
+
 	kfree(work->tr_buf);
 	kvfree(work->request_buf);
+	kfree(work->iov);
 	if (work->async_id)
 		ksmbd_release_id(&work->conn->async_ida, work->async_id);
 	kmem_cache_free(work_cache, work);
@@ -77,3 +94,75 @@ bool ksmbd_queue_work(struct ksmbd_work *work)
 {
 	return queue_work(ksmbd_wq, &work->work);
 }
+
+static int ksmbd_realloc_iov_pin(struct ksmbd_work *work, void *ib,
+				 unsigned int ib_len)
+{
+
+	if (work->iov_alloc_cnt <= work->iov_cnt) {
+		struct kvec *new;
+
+		work->iov_alloc_cnt += 4;
+		new = krealloc(work->iov,
+			       sizeof(struct kvec) * work->iov_alloc_cnt,
+			       GFP_KERNEL | __GFP_ZERO);
+		if (!new)
+			return -ENOMEM;
+		work->iov = new;
+	}
+
+	work->iov[++work->iov_idx].iov_base = ib;
+	work->iov[work->iov_idx].iov_len = ib_len;
+	work->iov_cnt++;
+
+	return 0;
+}
+
+static int __ksmbd_iov_pin_rsp(struct ksmbd_work *work, void *ib, int len,
+			       void *aux_buf, unsigned int aux_size)
+{
+	/* Plus rfc_length size on first iov */
+	if (!work->iov_idx) {
+		work->iov[work->iov_idx].iov_base = work->response_buf;
+		*(__be32 *)work->iov[0].iov_base = 0;
+		work->iov[work->iov_idx].iov_len = 4;
+		work->iov_cnt++;
+	}
+
+	ksmbd_realloc_iov_pin(work, ib, len);
+	inc_rfc1001_len(work->iov[0].iov_base, len);
+
+	if (aux_size) {
+		struct aux_read *ar;
+
+		ksmbd_realloc_iov_pin(work, aux_buf, aux_size);
+		inc_rfc1001_len(work->iov[0].iov_base, aux_size);
+
+		ar = kmalloc(sizeof(struct aux_read), GFP_KERNEL);
+		if (!ar)
+			return -ENOMEM;
+
+		ar->buf = aux_buf;
+		list_add(&ar->entry, &work->aux_read_list);
+	}
+
+	return 0;
+}
+
+int ksmbd_iov_pin_rsp(struct ksmbd_work *work, void *ib, int len)
+{
+	return __ksmbd_iov_pin_rsp(work, ib, len, NULL, 0);
+}
+
+int ksmbd_iov_pin_rsp_read(struct ksmbd_work *work, void *ib, int len,
+			   void *aux_buf, unsigned int aux_size)
+{
+	return __ksmbd_iov_pin_rsp(work, ib, len, aux_buf, aux_size);
+}
+
+void ksmbd_iov_reset(struct ksmbd_work *work)
+{
+	work->iov_idx = 0;
+	work->iov_cnt = 0;
+	*(__be32 *)work->iov[0].iov_base = 0;
+}

fs/smb/server/ksmbd_work.h

@@ -19,6 +19,11 @@ enum {
 	KSMBD_WORK_CLOSED,
 };
 
+struct aux_read {
+	void *buf;
+	struct list_head entry;
+};
+
 /* one of these for every pending CIFS request at the connection */
 struct ksmbd_work {
 	/* Server corresponding to this mid */
@@ -31,13 +36,19 @@ struct ksmbd_work {
 	/* Response buffer */
 	void                            *response_buf;
 
-	/* Read data buffer */
-	void                            *aux_payload_buf;
+	struct list_head		aux_read_list;
+
+	struct kvec			*iov;
+	int				iov_alloc_cnt;
+	int				iov_cnt;
+	int				iov_idx;
 
 	/* Next cmd hdr in compound req buf*/
 	int                             next_smb2_rcv_hdr_off;
 	/* Next cmd hdr in compound rsp buf*/
 	int                             next_smb2_rsp_hdr_off;
+	/* Current cmd hdr in compound rsp buf*/
+	int                             curr_smb2_rsp_hdr_off;
 
 	/*
 	 * Current Local FID assigned compound response if SMB2 CREATE
@@ -53,16 +64,11 @@ struct ksmbd_work {
 	unsigned int			credits_granted;
 
 	/* response smb header size */
-	unsigned int                    resp_hdr_sz;
 	unsigned int                    response_sz;
-	/* Read data count */
-	unsigned int                    aux_payload_sz;
 
 	void				*tr_buf;
 
 	unsigned char			state;
-	/* Multiple responses for one request e.g. SMB ECHO */
-	bool                            multiRsp:1;
 	/* No response for cancelled request */
 	bool                            send_no_response:1;
 	/* Request is encrypted */
@@ -95,6 +101,15 @@ static inline void *ksmbd_resp_buf_next(struct ksmbd_work *work)
 	return work->response_buf + work->next_smb2_rsp_hdr_off + 4;
 }
 
+/**
+ * ksmbd_resp_buf_curr - Get current buffer on compound response.
+ * @work: smb work containing response buffer
+ */
+static inline void *ksmbd_resp_buf_curr(struct ksmbd_work *work)
+{
+	return work->response_buf + work->curr_smb2_rsp_hdr_off + 4;
+}
+
 /**
  * ksmbd_req_buf_next - Get next buffer on compound request.
  * @work: smb work containing response buffer
@@ -113,5 +128,8 @@ int ksmbd_work_pool_init(void);
 int ksmbd_workqueue_init(void);
 void ksmbd_workqueue_destroy(void);
 bool ksmbd_queue_work(struct ksmbd_work *work);
-
+int ksmbd_iov_pin_rsp_read(struct ksmbd_work *work, void *ib, int len,
+			   void *aux_buf, unsigned int aux_size);
+int ksmbd_iov_pin_rsp(struct ksmbd_work *work, void *ib, int len);
+void ksmbd_iov_reset(struct ksmbd_work *work);
 #endif /* __KSMBD_WORK_H__ */

fs/smb/server/oplock.c

@@ -639,7 +639,6 @@ static void __smb2_oplock_break_noti(struct work_struct *wk)
 {
 	struct smb2_oplock_break *rsp = NULL;
 	struct ksmbd_work *work = container_of(wk, struct ksmbd_work, work);
-	struct ksmbd_conn *conn = work->conn;
 	struct oplock_break_info *br_info = work->request_buf;
 	struct smb2_hdr *rsp_hdr;
 	struct ksmbd_file *fp;
@@ -656,8 +655,6 @@ static void __smb2_oplock_break_noti(struct work_struct *wk)
 
 	rsp_hdr = smb2_get_msg(work->response_buf);
 	memset(rsp_hdr, 0, sizeof(struct smb2_hdr) + 2);
-	*(__be32 *)work->response_buf =
-		cpu_to_be32(conn->vals->header_size);
 	rsp_hdr->ProtocolId = SMB2_PROTO_NUMBER;
 	rsp_hdr->StructureSize = SMB2_HEADER_STRUCTURE_SIZE;
 	rsp_hdr->CreditRequest = cpu_to_le16(0);
@@ -684,13 +681,15 @@ static void __smb2_oplock_break_noti(struct work_struct *wk)
 	rsp->PersistentFid = fp->persistent_id;
 	rsp->VolatileFid = fp->volatile_id;
 
-	inc_rfc1001_len(work->response_buf, 24);
+	ksmbd_fd_put(work, fp);
+	if (ksmbd_iov_pin_rsp(work, (void *)rsp,
+			      sizeof(struct smb2_oplock_break)))
+		goto out;
 
 	ksmbd_debug(OPLOCK,
 		    "sending oplock break v_id %llu p_id = %llu lock level = %d\n",
 		    rsp->VolatileFid, rsp->PersistentFid, rsp->OplockLevel);
 
-	ksmbd_fd_put(work, fp);
 	ksmbd_conn_write(work);
 
 out:
@@ -751,7 +750,6 @@ static void __smb2_lease_break_noti(struct work_struct *wk)
 	struct smb2_lease_break *rsp = NULL;
 	struct ksmbd_work *work = container_of(wk, struct ksmbd_work, work);
 	struct lease_break_info *br_info = work->request_buf;
-	struct ksmbd_conn *conn = work->conn;
 	struct smb2_hdr *rsp_hdr;
 
 	if (allocate_oplock_break_buf(work)) {
@@ -761,8 +759,6 @@ static void __smb2_lease_break_noti(struct work_struct *wk)
 
 	rsp_hdr = smb2_get_msg(work->response_buf);
 	memset(rsp_hdr, 0, sizeof(struct smb2_hdr) + 2);
-	*(__be32 *)work->response_buf =
-		cpu_to_be32(conn->vals->header_size);
 	rsp_hdr->ProtocolId = SMB2_PROTO_NUMBER;
 	rsp_hdr->StructureSize = SMB2_HEADER_STRUCTURE_SIZE;
 	rsp_hdr->CreditRequest = cpu_to_le16(0);
@@ -791,7 +787,9 @@ static void __smb2_lease_break_noti(struct work_struct *wk)
 	rsp->AccessMaskHint = 0;
 	rsp->ShareMaskHint = 0;
 
-	inc_rfc1001_len(work->response_buf, 44);
+	if (ksmbd_iov_pin_rsp(work, (void *)rsp,
+			      sizeof(struct smb2_lease_break)))
+		goto out;
 
 	ksmbd_conn_write(work);
 
@@ -845,6 +843,7 @@ static int smb2_lease_break_noti(struct oplock_info *opinfo)
 			setup_async_work(in_work, NULL, NULL);
 			smb2_send_interim_resp(in_work, STATUS_PENDING);
 			list_del(&in_work->interim_entry);
+			ksmbd_iov_reset(in_work);
 		}
 		INIT_WORK(&work->work, __smb2_lease_break_noti);
 		ksmbd_queue_work(work);

fs/smb/server/server.c

@@ -163,6 +163,7 @@ static void __handle_ksmbd_work(struct ksmbd_work *work,
 {
 	u16 command = 0;
 	int rc;
+	bool is_chained = false;
 
 	if (conn->ops->allocate_rsp_buf(work))
 		return;
@@ -229,14 +230,13 @@ static void __handle_ksmbd_work(struct ksmbd_work *work,
 			}
 		}
 
+		is_chained = is_chained_smb2_message(work);
+
 		if (work->sess &&
 		    (work->sess->sign || smb3_11_final_sess_setup_resp(work) ||
 		     conn->ops->is_sign_req(work, command)))
 			conn->ops->set_sign_rsp(work);
-	} while (is_chained_smb2_message(work));
-
-	if (work->send_no_response)
-		return;
+	} while (is_chained == true);
 
 send:
 	smb3_preauth_hash_rsp(work);

fs/smb/server/smb_common.c

@@ -319,12 +319,6 @@ static int init_smb1_rsp_hdr(struct ksmbd_work *work)
 	struct smb_hdr *rsp_hdr = (struct smb_hdr *)work->response_buf;
 	struct smb_hdr *rcv_hdr = (struct smb_hdr *)work->request_buf;
 
-	/*
-	 * Remove 4 byte direct TCP header.
-	 */
-	*(__be32 *)work->response_buf =
-		cpu_to_be32(sizeof(struct smb_hdr) - 4);
-
 	rsp_hdr->Command = SMB_COM_NEGOTIATE;
 	*(__le32 *)rsp_hdr->Protocol = SMB1_PROTO_NUMBER;
 	rsp_hdr->Flags = SMBFLG_RESPONSE;
@@ -560,10 +554,11 @@ static int smb_handle_negotiate(struct ksmbd_work *work)
 
 	ksmbd_debug(SMB, "Unsupported SMB1 protocol\n");
 
-	/* Add 2 byte bcc and 2 byte DialectIndex. */
-	inc_rfc1001_len(work->response_buf, 4);
-	neg_rsp->hdr.Status.CifsError = STATUS_SUCCESS;
+	if (ksmbd_iov_pin_rsp(work, (void *)neg_rsp,
+			      sizeof(struct smb_negotiate_rsp) - 4))
+		return -ENOMEM;
 
+	neg_rsp->hdr.Status.CifsError = STATUS_SUCCESS;
 	neg_rsp->hdr.WordCount = 1;
 	neg_rsp->DialectIndex = cpu_to_le16(work->conn->dialect);
 	neg_rsp->ByteCount = 0;

fs/smb/server/transport_rdma.c

@@ -1241,14 +1241,12 @@ static int smb_direct_writev(struct ksmbd_transport *t,
 
 	//FIXME: skip RFC1002 header..
 	buflen -= 4;
-	iov[0].iov_base += 4;
-	iov[0].iov_len -= 4;
 
 	remaining_data_length = buflen;
 	ksmbd_debug(RDMA, "Sending smb (RDMA): smb_len=%u\n", buflen);
 
 	smb_direct_send_ctx_init(st, &send_ctx, need_invalidate, remote_key);
-	start = i = 0;
+	start = i = 1;
 	buflen = 0;
 	while (true) {
 		buflen += iov[i].iov_len;

fs/smb/server/vfs.c

@@ -367,15 +367,15 @@ static int check_lock_range(struct file *filp, loff_t start, loff_t end,
  * @fid:	file id of open file
  * @count:	read byte count
  * @pos:	file pos
+ * @rbuf:	read data buffer
  *
  * Return:	number of read bytes on success, otherwise error
  */
 int ksmbd_vfs_read(struct ksmbd_work *work, struct ksmbd_file *fp, size_t count,
-		   loff_t *pos)
+		   loff_t *pos, char *rbuf)
 {
 	struct file *filp = fp->filp;
 	ssize_t nbytes = 0;
-	char *rbuf = work->aux_payload_buf;
 	struct inode *inode = file_inode(filp);
 
 	if (S_ISDIR(inode->i_mode))

fs/smb/server/vfs.h

@@ -76,8 +76,8 @@ void ksmbd_vfs_query_maximal_access(struct mnt_idmap *idmap,
 				   struct dentry *dentry, __le32 *daccess);
 int ksmbd_vfs_create(struct ksmbd_work *work, const char *name, umode_t mode);
 int ksmbd_vfs_mkdir(struct ksmbd_work *work, const char *name, umode_t mode);
-int ksmbd_vfs_read(struct ksmbd_work *work, struct ksmbd_file *fp,
-		   size_t count, loff_t *pos);
+int ksmbd_vfs_read(struct ksmbd_work *work, struct ksmbd_file *fp, size_t count,
+		   loff_t *pos, char *rbuf);
 int ksmbd_vfs_write(struct ksmbd_work *work, struct ksmbd_file *fp,
 		    char *buf, size_t count, loff_t *pos, bool sync,
 		    ssize_t *written);