view tools/blktap/README @ 21067:b4a1832a916f

Update Xen version to 4.0.0-rc6
author Keir Fraser <>
date Tue Mar 09 18:18:05 2010 +0000 (2010-03-09)
parents 2937703f0ed0
line source
1 Blktap Userspace Tools + Library
2 ================================
4 Andrew Warfield and Julian Chesterfield
5 16th June 2006
7 {firstname.lastname}
9 The blktap userspace toolkit provides a user-level disk I/O
10 interface. The blktap mechanism involves a kernel driver that acts
11 similarly to the existing Xen/Linux blkback driver, and a set of
12 associated user-level libraries. Using these tools, blktap allows
13 virtual block devices presented to VMs to be implemented in userspace
14 and to be backed by raw partitions, files, network, etc.
16 The key benefit of blktap is that it makes it easy and fast to write
17 arbitrary block backends, and that these user-level backends actually
18 perform very well. Specifically:
20 - Metadata disk formats such as Copy-on-Write, encrypted disks, sparse
21 formats and other compression features can be easily implemented.
23 - Accessing file-based images from userspace avoids problems related
24 to flushing dirty pages which are present in the Linux loopback
25 driver. (Specifically, doing a large number of writes to an
26 NFS-backed image don't result in the OOM killer going berserk.)
28 - Per-disk handler processes enable easier userspace policing of block
29 resources, and process-granularity QoS techniques (disk scheduling
30 and related tools) may be trivially applied to block devices.
32 - It's very easy to take advantage of userspace facilities such as
33 networking libraries, compression utilities, peer-to-peer
34 file-sharing systems and so on to build more complex block backends.
36 - Crashes are contained -- incremental development/debugging is very
37 fast.
39 How it works (in one paragraph):
41 Working in conjunction with the kernel blktap driver, all disk I/O
42 requests from VMs are passed to the userspace deamon (using a shared
43 memory interface) through a character device. Each active disk is
44 mapped to an individual device node, allowing per-disk processes to
45 implement individual block devices where desired. The userspace
46 drivers are implemented using asynchronous (Linux libaio),
47 O_DIRECT-based calls to preserve the unbuffered, batched and
48 asynchronous request dispatch achieved with the existing blkback
49 code. We provide a simple, asynchronous virtual disk interface that
50 makes it quite easy to add new disk implementations.
52 As of June 2006 the current supported disk formats are:
54 - Raw Images (both on partitions and in image files)
55 - File-backed Qcow disks
56 - Standalone sparse Qcow disks
57 - Fast shareable RAM disk between VMs (requires some form of cluster-based
58 filesystem support e.g. OCFS2 in the guest kernel)
59 - Some VMDK images - your mileage may vary
61 Raw and QCow images have asynchronous backends and so should perform
62 fairly well. VMDK is based directly on the qemu vmdk driver, which is
63 synchronous (a.k.a. slow).
65 Build and Installation Instructions
66 ===================================
68 Make to configure the blktap backend driver in your dom0 kernel. It
69 will cooperate fine with the existing backend driver, so you can
70 experiment with tap disks without breaking existing VM configs.
72 To build the tools separately, "make && make install" in
73 tools/blktap.
76 Using the Tools
77 ===============
79 Prepare the image for booting. For qcow files use the qcow utilities
80 installed earlier. e.g. qcow-create generates a blank standalone image
81 or a file-backed CoW image. img2qcow takes an existing image or
82 partition and creates a sparse, standalone qcow-based file.
84 The userspace disk agent is configured to start automatically via xend
85 (alternatively you can start it manually => 'blktapctrl')
87 Customise the VM config file to use the 'tap' handler, followed by the
88 driver type. e.g. for a raw image such as a file or partition:
90 disk = ['tap:aio:<FILENAME>,sda1,w']
92 e.g. for a qcow image:
94 disk = ['tap:qcow:<FILENAME>,sda1,w']
97 Mounting images in Dom0 using the blktap driver
98 ===============================================
99 Tap (and blkback) disks are also mountable in Dom0 without requiring an
100 active VM to attach. You will need to build a xenlinux Dom0 kernel that
101 includes the blkfront driver (e.g. the default 'make world' or
102 'make kernels' build. Simply use the xm command-line tool to activate
103 the backend disks, and blkfront will generate a virtual block device that
104 can be accessed in the same way as a loop device or partition:
106 e.g. for a raw image file <FILENAME> that would normally be mounted using
107 the loopback driver (such as 'mount -o loop <FILENAME> /mnt/disk'), do the
108 following:
110 xm block-attach 0 tap:aio:<FILENAME> /dev/xvda1 w 0
111 mount /dev/xvda1 /mnt/disk <--- don't use loop driver
113 In this way, you can use any of the userspace device-type drivers built
114 with the blktap userspace toolkit to open and mount disks such as qcow
115 or vmdk images:
117 xm block-attach 0 tap:qcow:<FILENAME> /dev/xvda1 w 0
118 mount /dev/xvda1 /mnt/disk