Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
Merge branch 'bonding-support-aggregator-selection-based-on-port-priority'
Hangbin Liu says:
====================
bonding: support aggregator selection based on port priority
This patchset introduces a new per-port bonding option: `ad_actor_port_prio`.
It allows users to configure the actor's port priority, which can then be used
by the bonding driver for aggregator selection based on port priority.
This provides finer control over LACP aggregator choice, especially in setups
with multiple eligible aggregators over 2 switches.
====================
Link: https://patch.msgid.link/20250902064501.360822-1-liuhangbin@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
+247
-33
+21
-4
Documentation/networking/bonding.rst
+21
-4
Documentation/networking/bonding.rst
···
193
193
This parameter has effect only in 802.3ad mode and is available through
194
194
SysFs interface.
195
195
196
+
actor_port_prio
197
+
198
+
In an AD system, this specifies the port priority. The allowed range
199
+
is 1 - 65535. If the value is not specified, it takes 255 as the
200
+
default value.
201
+
202
+
This parameter has effect only in 802.3ad mode and is available through
203
+
netlink interface.
204
+
196
205
ad_actor_system
197
206
198
207
In an AD system, this specifies the mac-address for the actor in
···
250
241
ports (slaves). Reselection occurs as described under the
251
242
"bandwidth" setting, above.
252
243
253
-
The bandwidth and count selection policies permit failover of
254
-
802.3ad aggregations when partial failure of the active aggregator
255
-
occurs. This keeps the aggregator with the highest availability
256
-
(either in bandwidth or in number of ports) active at all times.
244
+
actor_port_prio or 3
245
+
246
+
The active aggregator is chosen by the highest total sum of
247
+
actor port priorities across its active ports. Note this
248
+
priority is actor_port_prio, not per port prio, which is
249
+
used for primary reselect.
250
+
251
+
The bandwidth, count and actor_port_prio selection policies permit
252
+
failover of 802.3ad aggregations when partial failure of the active
253
+
aggregator occurs. This keeps the aggregator with the highest
254
+
availability (either in bandwidth, number of ports, or total value
255
+
of port priorities) active at all times.
257
256
258
257
This option was added in bonding version 3.4.0.
259
258
+31
drivers/net/bonding/bond_3ad.c
+31
drivers/net/bonding/bond_3ad.c
···
436
436
437
437
port->actor_system = BOND_AD_INFO(bond).system.sys_mac_addr;
438
438
port->actor_system_priority = BOND_AD_INFO(bond).system.sys_priority;
439
+
port->actor_port_priority = SLAVE_AD_INFO(port->slave)->port_priority;
439
440
}
440
441
441
442
/* Conversions */
···
745
744
}
746
745
747
746
return active;
747
+
}
748
+
749
+
static unsigned int __agg_ports_priority(const struct aggregator *agg)
750
+
{
751
+
struct port *port = agg->lag_ports;
752
+
unsigned int prio = 0;
753
+
754
+
for (; port; port = port->next_port_in_aggregator)
755
+
if (port->is_enabled)
756
+
prio += port->actor_port_priority;
757
+
758
+
return prio;
748
759
}
749
760
750
761
/**
···
1720
1707
* 4. Therefore, current and best both have partner replies or
1721
1708
* both do not, so perform selection policy:
1722
1709
*
1710
+
* BOND_AD_PRIO: Select by total priority of ports. If priority
1711
+
* is equal, select by count.
1712
+
*
1723
1713
* BOND_AD_COUNT: Select by count of ports. If count is equal,
1724
1714
* select by bandwidth.
1725
1715
*
···
1744
1728
return best;
1745
1729
1746
1730
switch (__get_agg_selection_mode(curr->lag_ports)) {
1731
+
case BOND_AD_PRIO:
1732
+
if (__agg_ports_priority(curr) > __agg_ports_priority(best))
1733
+
return curr;
1734
+
1735
+
if (__agg_ports_priority(curr) < __agg_ports_priority(best))
1736
+
return best;
1737
+
1738
+
fallthrough;
1747
1739
case BOND_AD_COUNT:
1748
1740
if (__agg_active_ports(curr) > __agg_active_ports(best))
1749
1741
return curr;
···
1814
1790
* set of slaves in the bond changes.
1815
1791
*
1816
1792
* BOND_AD_COUNT: select the aggregator with largest number of ports
1793
+
* (slaves), and reselect whenever a link state change takes place or the
1794
+
* set of slaves in the bond changes.
1795
+
*
1796
+
* BOND_AD_PRIO: select the aggregator with highest total priority of ports
1817
1797
* (slaves), and reselect whenever a link state change takes place or the
1818
1798
* set of slaves in the bond changes.
1819
1799
*
···
2236
2208
port = &(SLAVE_AD_INFO(slave)->port);
2237
2209
2238
2210
ad_initialize_port(port, &bond->params);
2211
+
2212
+
/* Port priority is initialized. Update it to slave's ad info */
2213
+
SLAVE_AD_INFO(slave)->port_priority = port->actor_port_priority;
2239
2214
2240
2215
port->slave = slave;
2241
2216
port->actor_port_number = SLAVE_AD_INFO(slave)->id;
+16
drivers/net/bonding/bond_netlink.c
+16
drivers/net/bonding/bond_netlink.c
···
28
28
nla_total_size(sizeof(u8)) + /* IFLA_BOND_SLAVE_AD_ACTOR_OPER_PORT_STATE */
29
29
nla_total_size(sizeof(u16)) + /* IFLA_BOND_SLAVE_AD_PARTNER_OPER_PORT_STATE */
30
30
nla_total_size(sizeof(s32)) + /* IFLA_BOND_SLAVE_PRIO */
31
+
nla_total_size(sizeof(u16)) + /* IFLA_BOND_SLAVE_ACTOR_PORT_PRIO */
31
32
0;
32
33
}
33
34
···
78
77
ad_port->partner_oper.port_state))
79
78
goto nla_put_failure;
80
79
}
80
+
81
+
if (nla_put_u16(skb, IFLA_BOND_SLAVE_ACTOR_PORT_PRIO,
82
+
SLAVE_AD_INFO(slave)->port_priority))
83
+
goto nla_put_failure;
81
84
}
82
85
83
86
return 0;
···
135
130
static const struct nla_policy bond_slave_policy[IFLA_BOND_SLAVE_MAX + 1] = {
136
131
[IFLA_BOND_SLAVE_QUEUE_ID] = { .type = NLA_U16 },
137
132
[IFLA_BOND_SLAVE_PRIO] = { .type = NLA_S32 },
133
+
[IFLA_BOND_SLAVE_ACTOR_PORT_PRIO] = { .type = NLA_U16 },
138
134
};
139
135
140
136
static int bond_validate(struct nlattr *tb[], struct nlattr *data[],
···
182
176
bond_opt_slave_initval(&newval, &slave_dev, prio);
183
177
err = __bond_opt_set(bond, BOND_OPT_PRIO, &newval,
184
178
data[IFLA_BOND_SLAVE_PRIO], extack);
179
+
if (err)
180
+
return err;
181
+
}
182
+
183
+
if (data[IFLA_BOND_SLAVE_ACTOR_PORT_PRIO]) {
184
+
u16 ad_prio = nla_get_u16(data[IFLA_BOND_SLAVE_ACTOR_PORT_PRIO]);
185
+
186
+
bond_opt_slave_initval(&newval, &slave_dev, ad_prio);
187
+
err = __bond_opt_set(bond, BOND_OPT_ACTOR_PORT_PRIO, &newval,
188
+
data[IFLA_BOND_SLAVE_ACTOR_PORT_PRIO], extack);
185
189
if (err)
186
190
return err;
187
191
}
+41
-4
drivers/net/bonding/bond_options.c
+41
-4
drivers/net/bonding/bond_options.c
···
79
79
const struct bond_opt_value *newval);
80
80
static int bond_option_ad_actor_sys_prio_set(struct bonding *bond,
81
81
const struct bond_opt_value *newval);
82
+
static int bond_option_actor_port_prio_set(struct bonding *bond,
83
+
const struct bond_opt_value *newval);
82
84
static int bond_option_ad_actor_system_set(struct bonding *bond,
83
85
const struct bond_opt_value *newval);
84
86
static int bond_option_ad_user_port_key_set(struct bonding *bond,
···
162
160
};
163
161
164
162
static const struct bond_opt_value bond_ad_select_tbl[] = {
165
-
{ "stable", BOND_AD_STABLE, BOND_VALFLAG_DEFAULT},
166
-
{ "bandwidth", BOND_AD_BANDWIDTH, 0},
167
-
{ "count", BOND_AD_COUNT, 0},
168
-
{ NULL, -1, 0},
163
+
{ "stable", BOND_AD_STABLE, BOND_VALFLAG_DEFAULT},
164
+
{ "bandwidth", BOND_AD_BANDWIDTH, 0},
165
+
{ "count", BOND_AD_COUNT, 0},
166
+
{ "actor_port_prio", BOND_AD_PRIO, 0},
167
+
{ NULL, -1, 0},
169
168
};
170
169
171
170
static const struct bond_opt_value bond_num_peer_notif_tbl[] = {
···
222
219
static const struct bond_opt_value bond_ad_actor_sys_prio_tbl[] = {
223
220
{ "minval", 1, BOND_VALFLAG_MIN},
224
221
{ "maxval", 65535, BOND_VALFLAG_MAX | BOND_VALFLAG_DEFAULT},
222
+
{ NULL, -1, 0},
223
+
};
224
+
225
+
static const struct bond_opt_value bond_actor_port_prio_tbl[] = {
226
+
{ "minval", 0, BOND_VALFLAG_MIN},
227
+
{ "maxval", 65535, BOND_VALFLAG_MAX},
228
+
{ "default", 255, BOND_VALFLAG_DEFAULT},
225
229
{ NULL, -1, 0},
226
230
};
227
231
···
492
482
.unsuppmodes = BOND_MODE_ALL_EX(BIT(BOND_MODE_8023AD)),
493
483
.values = bond_ad_actor_sys_prio_tbl,
494
484
.set = bond_option_ad_actor_sys_prio_set,
485
+
},
486
+
[BOND_OPT_ACTOR_PORT_PRIO] = {
487
+
.id = BOND_OPT_ACTOR_PORT_PRIO,
488
+
.name = "actor_port_prio",
489
+
.unsuppmodes = BOND_MODE_ALL_EX(BIT(BOND_MODE_8023AD)),
490
+
.values = bond_actor_port_prio_tbl,
491
+
.set = bond_option_actor_port_prio_set,
495
492
},
496
493
[BOND_OPT_AD_ACTOR_SYSTEM] = {
497
494
.id = BOND_OPT_AD_ACTOR_SYSTEM,
···
1824
1807
newval->value);
1825
1808
1826
1809
bond->params.ad_actor_sys_prio = newval->value;
1810
+
bond_3ad_update_ad_actor_settings(bond);
1811
+
1812
+
return 0;
1813
+
}
1814
+
1815
+
static int bond_option_actor_port_prio_set(struct bonding *bond,
1816
+
const struct bond_opt_value *newval)
1817
+
{
1818
+
struct slave *slave;
1819
+
1820
+
slave = bond_slave_get_rtnl(newval->slave_dev);
1821
+
if (!slave) {
1822
+
netdev_dbg(bond->dev, "%s called on NULL slave\n", __func__);
1823
+
return -ENODEV;
1824
+
}
1825
+
1826
+
netdev_dbg(newval->slave_dev, "Setting actor_port_prio to %llu\n",
1827
+
newval->value);
1828
+
1829
+
SLAVE_AD_INFO(slave)->port_priority = newval->value;
1827
1830
bond_3ad_update_ad_actor_settings(bond);
1828
1831
1829
1832
return 0;
+2
include/net/bond_3ad.h
+2
include/net/bond_3ad.h
···
26
26
BOND_AD_STABLE = 0,
27
27
BOND_AD_BANDWIDTH = 1,
28
28
BOND_AD_COUNT = 2,
29
+
BOND_AD_PRIO = 3,
29
30
};
30
31
31
32
/* rx machine states(43.4.11 in the 802.3ad standard) */
···
275
274
struct port port; /* 802.3ad port structure */
276
275
struct bond_3ad_stats stats;
277
276
u16 id;
277
+
u16 port_priority;
278
278
};
279
279
280
280
static inline const char *bond_3ad_churn_desc(churn_state_t state)
+1
include/net/bond_options.h
+1
include/net/bond_options.h
+1
include/uapi/linux/if_link.h
+1
include/uapi/linux/if_link.h
+2
-1
tools/testing/selftests/drivers/net/bonding/Makefile
+2
-1
tools/testing/selftests/drivers/net/bonding/Makefile
+108
tools/testing/selftests/drivers/net/bonding/bond_lacp_prio.sh
+108
tools/testing/selftests/drivers/net/bonding/bond_lacp_prio.sh
···
1
+
#!/bin/bash
2
+
# SPDX-License-Identifier: GPL-2.0
3
+
#
4
+
# Testing if bond lacp per port priority works
5
+
#
6
+
# Switch (s_ns) Backup Switch (b_ns)
7
+
# +-------------------------+ +-------------------------+
8
+
# | bond0 | | bond0 |
9
+
# | + | | + |
10
+
# | eth0 | eth1 | | eth0 | eth1 |
11
+
# | +---+---+ | | +---+---+ |
12
+
# | | | | | | | |
13
+
# +-------------------------+ +-------------------------+
14
+
# | | | |
15
+
# +-----------------------------------------------------+
16
+
# | | | | | |
17
+
# | +-------+---------+---------+-------+ |
18
+
# | eth0 eth1 | eth2 eth3 |
19
+
# | + |
20
+
# | bond0 |
21
+
# +-----------------------------------------------------+
22
+
# Client (c_ns)
23
+
24
+
lib_dir=$(dirname "$0")
25
+
# shellcheck disable=SC1091
26
+
source "$lib_dir"/../../../net/lib.sh
27
+
28
+
setup_links()
29
+
{
30
+
# shellcheck disable=SC2154
31
+
ip -n "${c_ns}" link add eth0 type veth peer name eth0 netns "${s_ns}"
32
+
ip -n "${c_ns}" link add eth1 type veth peer name eth1 netns "${s_ns}"
33
+
# shellcheck disable=SC2154
34
+
ip -n "${c_ns}" link add eth2 type veth peer name eth0 netns "${b_ns}"
35
+
ip -n "${c_ns}" link add eth3 type veth peer name eth1 netns "${b_ns}"
36
+
37
+
ip -n "${c_ns}" link add bond0 type bond mode 802.3ad miimon 100 \
38
+
lacp_rate fast ad_select actor_port_prio
39
+
ip -n "${s_ns}" link add bond0 type bond mode 802.3ad miimon 100 \
40
+
lacp_rate fast
41
+
ip -n "${b_ns}" link add bond0 type bond mode 802.3ad miimon 100 \
42
+
lacp_rate fast
43
+
44
+
ip -n "${c_ns}" link set eth0 master bond0
45
+
ip -n "${c_ns}" link set eth1 master bond0
46
+
ip -n "${c_ns}" link set eth2 master bond0
47
+
ip -n "${c_ns}" link set eth3 master bond0
48
+
ip -n "${s_ns}" link set eth0 master bond0
49
+
ip -n "${s_ns}" link set eth1 master bond0
50
+
ip -n "${b_ns}" link set eth0 master bond0
51
+
ip -n "${b_ns}" link set eth1 master bond0
52
+
53
+
ip -n "${c_ns}" link set bond0 up
54
+
ip -n "${s_ns}" link set bond0 up
55
+
ip -n "${b_ns}" link set bond0 up
56
+
}
57
+
58
+
test_port_prio_setting()
59
+
{
60
+
RET=0
61
+
ip -n "${c_ns}" link set eth0 type bond_slave actor_port_prio 1000
62
+
prio=$(cmd_jq "ip -n ${c_ns} -d -j link show eth0" \
63
+
".[].linkinfo.info_slave_data.actor_port_prio")
64
+
[ "$prio" -ne 1000 ] && RET=1
65
+
ip -n "${c_ns}" link set eth2 type bond_slave actor_port_prio 10
66
+
prio=$(cmd_jq "ip -n ${c_ns} -d -j link show eth2" \
67
+
".[].linkinfo.info_slave_data.actor_port_prio")
68
+
[ "$prio" -ne 10 ] && RET=1
69
+
}
70
+
71
+
test_agg_reselect()
72
+
{
73
+
local bond_agg_id slave_agg_id
74
+
local expect_slave="$1"
75
+
RET=0
76
+
77
+
# Trigger link state change to reselect the aggregator
78
+
ip -n "${c_ns}" link set eth1 down
79
+
sleep 0.5
80
+
ip -n "${c_ns}" link set eth1 up
81
+
sleep 0.5
82
+
83
+
bond_agg_id=$(cmd_jq "ip -n ${c_ns} -d -j link show bond0" \
84
+
".[].linkinfo.info_data.ad_info.aggregator")
85
+
slave_agg_id=$(cmd_jq "ip -n ${c_ns} -d -j link show $expect_slave" \
86
+
".[].linkinfo.info_slave_data.ad_aggregator_id")
87
+
# shellcheck disable=SC2034
88
+
[ "${bond_agg_id}" -ne "${slave_agg_id}" ] && \
89
+
RET=1
90
+
}
91
+
92
+
trap cleanup_all_ns EXIT
93
+
setup_ns c_ns s_ns b_ns
94
+
setup_links
95
+
96
+
test_port_prio_setting
97
+
log_test "bond 802.3ad" "actor_port_prio setting"
98
+
99
+
test_agg_reselect eth0
100
+
log_test "bond 802.3ad" "actor_port_prio select"
101
+
102
+
# Change the actor port prio and re-test
103
+
ip -n "${c_ns}" link set eth0 type bond_slave actor_port_prio 10
104
+
ip -n "${c_ns}" link set eth2 type bond_slave actor_port_prio 1000
105
+
test_agg_reselect eth2
106
+
log_test "bond 802.3ad" "actor_port_prio switch"
107
+
108
+
exit "${EXIT_STATUS}"
-24
tools/testing/selftests/net/forwarding/lib.sh
-24
tools/testing/selftests/net/forwarding/lib.sh
···
571
571
fi
572
572
}
573
573
574
-
cmd_jq()
575
-
{
576
-
local cmd=$1
577
-
local jq_exp=$2
578
-
local jq_opts=$3
579
-
local ret
580
-
local output
581
-
582
-
output="$($cmd)"
583
-
# it the command fails, return error right away
584
-
ret=$?
585
-
if [[ $ret -ne 0 ]]; then
586
-
return $ret
587
-
fi
588
-
output=$(echo $output | jq -r $jq_opts "$jq_exp")
589
-
ret=$?
590
-
if [[ $ret -ne 0 ]]; then
591
-
return $ret
592
-
fi
593
-
echo $output
594
-
# return success only in case of non-empty output
595
-
[ ! -z "$output" ]
596
-
}
597
-
598
574
pre_cleanup()
599
575
{
600
576
if [ "${PAUSE_ON_CLEANUP}" = "yes" ]; then
+24
tools/testing/selftests/net/lib.sh
+24
tools/testing/selftests/net/lib.sh
···
645
645
sleep 0.1
646
646
done
647
647
}
648
+
649
+
cmd_jq()
650
+
{
651
+
local cmd=$1
652
+
local jq_exp=$2
653
+
local jq_opts=$3
654
+
local ret
655
+
local output
656
+
657
+
output="$($cmd)"
658
+
# it the command fails, return error right away
659
+
ret=$?
660
+
if [[ $ret -ne 0 ]]; then
661
+
return $ret
662
+
fi
663
+
output=$(echo $output | jq -r $jq_opts "$jq_exp")
664
+
ret=$?
665
+
if [[ $ret -ne 0 ]]; then
666
+
return $ret
667
+
fi
668
+
echo $output
669
+
# return success only in case of non-empty output
670
+
[ ! -z "$output" ]
671
+
}