vhd creation¶
There is currently a hard dependency on the patched version of
vhd-utils
to create Xenserver VHD compatible images. The main
target for these images is the Rackspace Public Cloud.
This note should explain the current issues.
The patch to include convert
as an option to vhd-utils
,
avaialble at https://github.com/citrix-openstack/xenserver-utils/, is
built by infra into a PPA available at
https://launchpad.net/~openstack-ci-core/+archive/ubuntu/vhd-util.
Upstream Xen has actually removed all the code related to blktap2
with
https://xenbits.xen.org/gitweb/?p=xen.git;a=commit;h=5c883cf036cf5ab8b1b79390549e2475f7a568dd.
The tools have been removed from Debuntu with
https://bugs.debian.org/917907. So this patch is even less trivial to
apply; the project has seemingly been split into
https://github.com/xapi-project/blktap/ but support status is unclear.
It is tempting to use qemu-img
image conversion routines. Indeed
this supports converting RAW images to VPC output. However, VPC is a
fairly flexible standard across a number of different hypervisors.
The main standard is available from Microsoft at
https://www.microsoft.com/en-us/download/confirmation.aspx?id=23850
and is what qemu-img
impelements.
However, trying to boot an image created with this results in a “resize” error. This is described in https://bugs.launchpad.net/nova/+bug/862653 but comes down to the following in https://github.com/xapi-project/blktap/blob/master/vhd/lib/vhd-util-resize.c#L1055:
if (!vhd_creator_tapdisk(&vhd)) {
printf("%s not created by xen; resize not supported\n", name);
err = -EINVAL;
}
It seems that Xen will refuse to grow a image without a creator
tap\0
. It is tempting to patch qemu-img
to do this (patch
follows for those interested). However, if you look at the output
with vhd-util read -p -n ./<file>.vhd
you will notice it complains
about an invalid batmap
. The batmap
appears to be a Xen
extension to the VPC standard. Checking vhd_has_batmap
at
https://github.com/xen-project/xen/blob/365aabb6e5023cee476adf81106729efd49c644f/tools/blktap2/vhd/lib/libvhd.c#L1193
it seems we can fool Xen by setting the vhd version tags low enough
that it thinks the image doesn’t have a batmap
(i.e. presumably
created with tools before batmap
’s existed).
Side note: there is a bunch of qemu work to implement overlay VHD
disks with
https://patchwork.ozlabs.org/project/qemu-devel/list/?submitter=64750
which has never merged. This also seems to include a batmap
implementation.
With the full patch applied below, we get images that appear fairly
similar from (patched) vhd-util
and qemu-img
; generated from
the same .raw
file.
qemu-img output:
VHD Footer Summary:
-------------------
Cookie : conectix
Features : (0x00000002) <RESV>
File format version : Major: 1, Minor: 0
Data offset : 512
Timestamp : Wed Oct 30 06:25:41 2019
Creator Application : 'tap'
Creator version : Major: 1, Minor: 0
Creator OS : Unknown!
Original disk size : 16667 MB (17477591040 Bytes)
Current disk size : 16667 MB (17477591040 Bytes)
Geometry : Cyl: 33865, Hds: 16, Sctrs: 63
: = 16667 MB (17477591040 Bytes)
Disk type : Dynamic hard disk
Checksum : 0xffffec8c|0xffffec8c (Good!)
UUID : 1bfda481-dd4d-43aa-8f3a-84689b5ab3d7
Saved state : No
Hidden : 0
VHD Header Summary:
-------------------
Cookie : cxsparse
Data offset (unusd) : 18446744073709
Table offset : 1536
Header version : 0x00010000
Max BAT size : 8334
Block size : 2097152 (2 MB)
Parent name :
Parent UUID : 00000000-0000-0000-0000-000000000000
Parent timestamp : Sat Jan 1 00:00:00 2000
Checksum : 0xfffff3c9|0xfffff3c9 (Good!)
vhd-utils output:
VHD Footer Summary:
-------------------
Cookie : conectix
Features : (0x00000002) <RESV>
File format version : Major: 1, Minor: 0
Data offset : 512
Timestamp : Tue Oct 8 09:18:54 2019
Creator Application : 'tap'
Creator version : Major: 1, Minor: 3
Creator OS : Unknown!
Original disk size : 16668 MB (17477664768 Bytes)
Current disk size : 16668 MB (17477664768 Bytes)
Geometry : Cyl: 33865, Hds: 16, Sctrs: 63
: = 16667 MB (17477591040 Bytes)
Disk type : Dynamic hard disk
Checksum : 0xffffeebb|0xffffeebb (Good!)
UUID : 14f2710a-b9b8-48f7-94bb-005ba6d566b2
Saved state : No
Hidden : 0
VHD Header Summary:
-------------------
Cookie : cxsparse
Data offset (unusd) : 18446744073709
Table offset : 1536
Header version : 0x00010000
Max BAT size : 8334
Block size : 2097152 (2 MB)
Parent name :
Parent UUID : 00000000-0000-0000-0000-000000000000
Parent timestamp : Sat Jan 1 00:00:00 2000
Checksum : 0xfffff3c9|0xfffff3c9 (Good!)
VHD Batmap Summary:
-------------------
Batmap offset : 35840
Batmap size (secs) : 3
Batmap version : 0x00010002
Checksum : 0xfffbf214|0xfffbf214 (Good!)
This does upload and boot in RAX, but the root disk does not appear to grow correctly.
Emperically, it also seems that fixed sized disks (generated with
subformat=fixed
) will not import (at least into Rackspace). I
also tried resizing the dynamic vhd
(with vhd-util
resize --debug -n ./test.vhd -s $((32 * 1024 * 1024 )) -j
resize.log
), which looked correct in vhd-util
output (Current
disk size
grew) but Rackspace would not import this image.
qemu patch¶
Applies against qemu at HEAD
16884391c750d0c5e863f55ad7aaaa146fc5181e
diff --git a/block/vpc.c b/block/vpc.c
index a655502..d1716f0 100644
--- a/block/vpc.c
+++ b/block/vpc.c
@@ -60,6 +60,7 @@ enum vhd_type {
#define VHD_MAX_GEOMETRY (VHD_CHS_MAX_C * VHD_CHS_MAX_H * VHD_CHS_MAX_S)
#define VPC_OPT_FORCE_SIZE "force_size"
+#define VPC_OPT_XENSERVER_COMPAT "xenserver_compat"
/* always big-endian */
typedef struct vhd_footer {
@@ -1042,12 +1043,16 @@ static int coroutine_fn vpc_co_create(BlockdevCreateOptions *opts,
memset(buf, 0, 1024);
memcpy(footer->creator, "conectix", 8);
- if (vpc_opts->force_size) {
+ if (vpc_opts->xenserver_compat) {
+ memcpy(footer->creator_app, "tap\0", 4);
+ memcpy(footer->creator_os, "\0\0\0\0", 4);
+ } else if (vpc_opts->force_size) {
memcpy(footer->creator_app, "qem2", 4);
+ memcpy(footer->creator_os, "Wi2k", 4);
} else {
memcpy(footer->creator_app, "qemu", 4);
+ memcpy(footer->creator_os, "Wi2k", 4);
}
- memcpy(footer->creator_os, "Wi2k", 4);
footer->features = cpu_to_be32(0x02);
footer->version = cpu_to_be32(0x00010000);
@@ -1058,9 +1063,14 @@ static int coroutine_fn vpc_co_create(BlockdevCreateOptions *opts,
}
footer->timestamp = cpu_to_be32(time(NULL) - VHD_TIMESTAMP_BASE);
- /* Version of Virtual PC 2007 */
- footer->major = cpu_to_be16(0x0005);
- footer->minor = cpu_to_be16(0x0003);
+ if (vpc_opts->xenserver_compat) {
+ footer->major = cpu_to_be16(0x0001);
+ footer->minor = cpu_to_be16(0x0000);
+ } else {
+ /* Version of Virtual PC 2007 */
+ footer->major = cpu_to_be16(0x0005);
+ footer->minor = cpu_to_be16(0x0003);
+ }
footer->orig_size = cpu_to_be64(total_size);
footer->current_size = cpu_to_be64(total_size);
footer->cyls = cpu_to_be16(cyls);
@@ -1101,6 +1111,7 @@ static int coroutine_fn vpc_co_create_opts(const char *filename,
static const QDictRenames opt_renames[] = {
{ VPC_OPT_FORCE_SIZE, "force-size" },
+ { VPC_OPT_XENSERVER_COMPAT, "xenserver-compat" },
{ NULL, NULL },
};
@@ -1220,6 +1231,11 @@ static QemuOptsList vpc_create_opts = {
"specified, rather than using the nearest CHS-based "
"calculation"
},
+ {
+ .name = VPC_OPT_XENSERVER_COMPAT,
+ .type = QEMU_OPT_BOOL,
+ .help = "Set creator to tap"
+ },
{ /* end of list */ }
}
};
diff --git a/qapi/block-core.json b/qapi/block-core.json
index aa97ee2..68daa3c 100644
--- a/qapi/block-core.json
+++ b/qapi/block-core.json
@@ -4669,6 +4669,7 @@
# @force-size Force use of the exact byte size instead of rounding to the
# next size that can be represented in CHS geometry
# (default: false)
+# @xenserver-compat Xenserver comapt
#
# Since: 2.12
##
@@ -4676,7 +4677,9 @@
'data': { 'file': 'BlockdevRef',
'size': 'size',
'*subformat': 'BlockdevVpcSubformat',
- '*force-size': 'bool' } }
+ '*force-size': 'bool',
+ '*xenserver-compat': 'bool'
+ } }
##
# @BlockdevCreateOptions: