On Thu, Aug 05, 2021 at 10:06:03AM -0700, Bart Van Assche wrote:

On 8/5/21 12:00 AM, longli@linuxonhyperv.com wrote:

quoted

diff --git a/include/uapi/misc/hv_azure_blob.h b/include/uapi/misc/hv_azure_blob.h
new file mode 100644
index 0000000..87a3f77
--- /dev/null
+++ b/include/uapi/misc/hv_azure_blob.h

@@ -0,0 +1,35 @@
+/* SPDX-License-Identifier: GPL-2.0-only WITH Linux-syscall-note */
+/* Copyright (c) 2021 Microsoft Corporation. */
+
+#ifndef _AZ_BLOB_H
+#define _AZ_BLOB_H
+
+#include <linux/ioctl.h>
+#include <linux/uuid.h>
+#include <linux/types.h>
+
+/* user-mode sync request sent through ioctl */
+struct az_blob_request_sync_response {
+	__u32 status;
+	__u32 response_len;
+};
+
+struct az_blob_request_sync {
+	guid_t guid;
+	__u32 timeout;
+	__u32 request_len;
+	__u32 response_len;
+	__u32 data_len;
+	__u32 data_valid;
+	__aligned_u64 request_buffer;
+	__aligned_u64 response_buffer;
+	__aligned_u64 data_buffer;
+	struct az_blob_request_sync_response response;
+};
+
+#define AZ_BLOB_MAGIC_NUMBER	'R'
+#define IOCTL_AZ_BLOB_DRIVER_USER_REQUEST \
+		_IOWR(AZ_BLOB_MAGIC_NUMBER, 0xf0, \
+			struct az_blob_request_sync)
+
+#endif /* define _AZ_BLOB_H */

So this driver only supports synchronous requests? Is it likely that users
will ask for support of an API that supports having multiple requests
outstanding at the same time without having to create multiple user space
threads?

They seem to want to go around the block layer and reinvent it all from
scratch, which is crazy, so let them have their slow i/o :(