1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
|
#!/bin/bash -norc
# Magic startup script for a UML instance. As long as unique
# instances are started, more than one of them can be concurrently
# in use on a single system. All their network interfaces are
# bridged together onto the virtual bridge "virbr0" which is
# supplied by the "libvirt" package.
#
# Note that a DHCP server is started for that interface. It's
# configured in this file:
# /etc/libvirt/qemu/networks/default.xml
# Unfortunately what I see there serves all possible DHCP addresses,
# so stealing them like we do here isn't really kosher. To fix
# it, that configuration should change to serve a smaller subset
# of the available address range.
#
# Each instance uses its own tun/tap device, created using the
# "tunctl" command. The assigned tap device will correspond with
# the guest id (a small integer representing the instance), i.e.,
# guest id 1 uses tap1, etc. The tap device is attached to the
# virtual bridge, which will have its own subnet associated with it.
# The guest side of that interface will have the same subnet as the
# bridge interface, with the bottom bits representing (normally) 100
# more than the guest id. So for subnet 192.168.122.0/24, guest
# id 1 will use ip 192.168.122.101, guest id 2 will use ip
# 192.168.122.102, and so on. Because these interfaces are bridged,
# they can all communicate with each other.
# You will want to override this by setting and exporting the
# "CEPH_TOP" environment variable to be the directory that contains
# the "ceph-client" source tree.
CEPH_TOP="${CEPH_TOP:-/home/elder/ceph}"
# You may want to change this too, if you want guest UML instances
# to have a diffeerent IP address range. The guest IP will be based
# on this plus GUEST_ID (defined below).
GUEST_IP_OFFSET="${GUEST_IP_OFFSET:-100}"
#############################
if [ $# -gt 1 ]; then
echo "" >&2
echo "Usage: $(basename $0) [guest_id]" >&2
echo "" >&2
echo " guest_id is a small integer (default 1)" >&2
echo " (each UML instance needs a distinct guest_id)" >&2
echo "" >&2
exit 1
elif [ $# -eq 1 ]; then
GUEST_ID="$1"
else
GUEST_ID=1
fi
# This will be what the guest host calls itself.
GUEST_HOSTNAME="uml-${GUEST_ID}"
# This is the path to the boot disk image used by UML.
DISK_IMAGE_A="${CEPH_TOP}/ceph-client/uml.${GUEST_ID}"
if [ ! -f "${DISK_IMAGE_A}" ]; then
echo "root disk image not found (or not a file)" >&2
exit 2
fi
# Hostid 1 uses tun/tap device tap1, hostid 2 uses tap2, etc.
TAP_ID="${GUEST_ID}"
# This is the tap device used for this UML instance
TAP="tap${TAP_ID}"
# This is just used to mount an image temporarily
TMP_MNT="/tmp/m$$"
# Where to put a config file generated for this tap device
TAP_IFUPDOWN_CONFIG="/tmp/interface-${TAP}"
# Compute the HOST_IP and BROADCAST address values to use,
# and assign shell variables with those names to their values.
# Also compute BITS, which is the network prefix length used.
# The NETMASK is then computed using that BITS value.
eval $(
ip addr show virbr0 | awk '
/inet/ {
split($2, a, "/")
printf("HOST_IP=%s\n", a[1]);
printf("BROADCAST=%s\n", $4);
printf("BITS=%s\n", a[2]);
exit(0);
}')
# Use bc to avoid 32-bit wrap when computing netmask
eval $(
echo -n "NETMASK="
bc <<! | fmt | sed 's/ /./g'
m = 2 ^ 32 - 2 ^ (32 - ${BITS})
for (p = 24; p >= 0; p = p - 8)
m / (2 ^ p) % 256
!
)
# Now use the netmask and the host IP to compute the subnet address
# and from that the guest IP address to use.
eval $(
awk '
function from_quad(addr, a, val, i) {
if (split(addr, a, ".") != 4)
exit(1); # address not in dotted quad format
val = 0;
for (i = 1; i <= 4; i++)
val = val * 256 + a[i];
return val;
}
function to_quad(val, addr, i) {
addr = "";
for (i = 1; i <= 4; i++) {
addr = sprintf("%u%s%s", val % 256, i > 1 ? "." : "", addr);
val = int(val / 256);
}
if ((val + 0) != 0)
exit(1); # value provided exceeded 32 bits
return addr;
}
BEGIN {
host_ip = from_quad("'${HOST_IP}'");
netmask = from_quad("'${NETMASK}'");
guest_net_ip = '${GUEST_IP_OFFSET}' + '${GUEST_ID}';
if (and(netmask, guest_net_ip))
exit(1); # address too big for subnet
subnet = and(host_ip, netmask);
guest_ip = or(subnet, guest_net_ip);
if (guest_ip == host_ip)
exit(1); # computed guest ip matches host ip
printf("SUBNET=%s\n", to_quad(subnet));
printf("GUEST_IP=%s\n", to_quad(guest_ip));
}
' < /dev/null
)
############## OK, we now know all our network parameters...
# There is a series of things that need to be done as superuser,
# so group them all into one big (and sort of nested!) sudo request.
sudo -s <<EnD_Of_sUdO
# Mount the boot disk for the UML and set up some configuration
# files there.
mkdir -p "${TMP_MNT}"
mount -o loop "${DISK_IMAGE_A}" "${TMP_MNT}"
# Arrange for loopback and eth0 to load automatically,
# and for eth0 to have our desired network parameters.
cat > "${TMP_MNT}/etc/network/interfaces" <<!
# Used by ifup(8) and ifdown(8). See the interfaces(5) manpage or
# /usr/share/doc/ifupdown/examples for more information.
auto lo
iface lo inet loopback
auto eth0
# iface eth0 inet dhcp
iface eth0 inet static
address ${GUEST_IP}
netmask ${NETMASK}
broadcast ${BROADCAST}
gateway ${HOST_IP}
!
# Have the guest start with an appropriate host name.
# Also record an entry for it in its own hosts file.
echo "${GUEST_HOSTNAME}" > "${TMP_MNT}/etc/hostname"
echo "${GUEST_IP} ${GUEST_HOSTNAME}" >> "${TMP_MNT}/etc/hosts"
# The host will serve as the name server also
cat > "${TMP_MNT}/etc/resolv.conf" <<!
nameserver ${HOST_IP}
!
# OK, done tweaking the boot image.
sync
umount "${DISK_IMAGE_A}"
rmdir "${TMP_MNT}"
# Set up a config file for "ifup" and "ifdown" (on the host) to use.
# All the backslashes below are needed because we're sitting inside
# a double here-document...
cat > "${TAP_IFUPDOWN_CONFIG}" <<!
iface ${TAP} inet manual
up brctl addif virbr0 "\\\${IFACE}"
up ip link set dev "\\\${IFACE}" up
pre-down brctl delif virbr0 "\\\${IFACE}"
pre-down ip link del dev "\\\${IFACE}"
tunctl_user $(whoami)
!
# OK, bring up the tap device using our config file
ifup -i "${TAP_IFUPDOWN_CONFIG}" "${TAP}"
EnD_Of_sUdO
# Finally ready to launch the UML instance.
./linux \
umid="${GUEST_HOSTNAME}" \
ubda="${DISK_IMAGE_A}" \
eth0="tuntap,${TAP}" \
mem=1024M
# When we're done, clean up. Bring down the tap interface and
# delete the config file.
#
# Note that if the above "./linux" crashes, you'll need to run the
# following commands manually in order to clean up state.
sudo ifdown -i "${TAP_IFUPDOWN_CONFIG}" "${TAP}"
sudo rm -f "${TAP_IFUPDOWN_CONFIG}"
exit 0
|
