主页 / ubuntu / 问题 / 1333453
Accepted
Doug Smythies
Asked: 2021-04-24 08:40:05 +0800 CST

KVM/Qemu LAN 中的桥接网络处理丢弃的数据包

  • 772

摘要：在桥接 KVM/QEMU 配置中，发往来宾 VM 的网络数据包无法到达那里。

配置：主机是最新的 Ubuntu 20.04.2 LTS 服务器；来宾是 3 个虚拟机中的任何一个，一个非常旧的 16.04 Ubuntu 服务器，一个旧的 Ubuntu 20.04 桌面和一个全新的 Ubuntu 21.04 桌面。前 2 个 VM 正在从非桥接、NAT'd 转换，第 3 个 VM 是指定桥接网络创建的。最终，VM 将通过 DHCP 从主 LAN 获取其 IP 地址，但目前为了获得更好的调试信息，它们使用静态 IP 地址。

主机桥定义 , /etc/netplan/01-netcfg.yaml（这是众多尝试之一）：

# This file describes the network interfaces available on your system
# For more information, see netplan(5).
network:
  version: 2
  renderer: networkd
  ethernets:
    enp3s0:
      dhcp4: no
  bridges:
    br0:
#      interfaces: [ enp3s0 ]
      dhcp4: yes
#      dhcp6: no
#      link-local: [ ]
      interfaces:
        - enp3s0
#      parameters:
#        stp: true
#        forward-delay: 4

虚拟的东西，/etc/libvirt/qemu/networks/br0.xml：

<!--
WARNING: THIS IS AN AUTO-GENERATED FILE. CHANGES TO IT ARE LIKELY TO BE
OVERWRITTEN AND LOST. Changes to this xml configuration should be made using:
  virsh net-edit br0
or other application using the libvirt API.
-->

<network>
  <name>br0</name>
  <uuid>40a8752c-d074-4802-bae8-b0aef95d9c99</uuid>
  <forward mode='bridge'/>
  <bridge name='br0'/>
</network>

注意：已经尝试了许多版本的桥接 .xml 文件，包括不同的名称，不同的引用使用不同的技术。Ubuntu Serverguide 引用了这个，它说名称和网桥名称必须相同，但其他参考文献中的名称和网桥名称不同。使用 nano 创建裸文件后，这些命令：

virsh net-define br0.xml
virsh net-autostart br0
virsh net-start br0

用于添加和配置它。默认的 NAT 方式是从自动启动目录中取消链接的，以免取消链接。最终，它是未定义的。结果：

$ virsh net-list --all
 Name      State      Autostart   Persistent
----------------------------------------------
 br0       active     yes         yes

此时重启后完全没有iptables规则。但是，VM 没有网络访问权限。请注意，一些参考文献提到了 br_netfilter 模块的特殊 iptables 规则和特殊属性，所有这些都已尝试过。这个问题已经够长了，没有详细介绍这里尝试的变化的所有细节。

调试详细信息：无论配置变体如何，基本问题始终相同，VM 目标数据包似乎没有到达主机，至少在 tcpdump 中看到。但是，广播类型的数据包确实会到达并确实到达客户端 VM。

此示例将在 LAN 上使用 192.168.111.59 (MAC: 52:54:00:60:ea:0e)、16.04 服务器 VM 和 192.168.111.132（树莓派）。20.04 主机服务器位于 192.168.111.136。网络掩码为 24 位，255.255.255.0。网关和 DHCP 服务器是 Debian 服务器（顺便说一下，桥接来宾 VM 在其上工作正常）。

在 ping 期间从 raspberry-pi 看到的第一个 tpcudmp：

doug@rpi2:~ $ sudo tcpdump -n -tttt -i eth0 ether host 52:54:00:60:ea:0e
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on eth0, link-type EN10MB (Ethernet), capture size 262144 bytes
2021-04-23 08:33:19.363553 ARP, Request who-has 192.168.111.1 tell 192.168.111.59, length 46
2021-04-23 08:33:19.487239 IP 192.168.111.132 > 192.168.111.59: ICMP echo request, id 27848, seq 14, length 64
2021-04-23 08:33:20.363542 ARP, Request who-has 192.168.111.1 tell 192.168.111.59, length 46
2021-04-23 08:33:20.527250 IP 192.168.111.132 > 192.168.111.59: ICMP echo request, id 27848, seq 15, length 64
2021-04-23 08:33:21.567215 IP 192.168.111.132 > 192.168.111.59: ICMP echo request, id 27848, seq 16, length 64
2021-04-23 08:33:22.607228 IP 192.168.111.132 > 192.168.111.59: ICMP echo request, id 27848, seq 17, length 64
2021-04-23 08:33:23.372351 ARP, Request who-has 192.168.111.1 tell 192.168.111.59, length 46
2021-04-23 08:33:23.647228 IP 192.168.111.132 > 192.168.111.59: ICMP echo request, id 27848, seq 18, length 64
2021-04-23 08:33:24.371431 ARP, Request who-has 192.168.111.1 tell 192.168.111.59, length 46

从所有 ARP 活动中可以看出，VM 正在正常发送数据包。但是，它从不回复任何内容。现在让我们观察来自主机的相同活动，注意 tcpdump 输出对于任何接口 br0 或 enp3s0 或 vnet0 都是相同的。

$ sudo tcpdump -n -tttt -i br0 ether host 52:54:00:60:ea:0e
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on br0, link-type EN10MB (Ethernet), capture size 262144 bytes
2021-04-23 08:40:38.837608 ARP, Request who-has 192.168.111.1 tell 192.168.111.59, length 28
2021-04-23 08:40:39.837159 ARP, Request who-has 192.168.111.1 tell 192.168.111.59, length 28
2021-04-23 08:40:40.837122 ARP, Request who-has 192.168.111.1 tell 192.168.111.59, length 28
2021-04-23 08:40:43.842985 ARP, Request who-has 192.168.111.1 tell 192.168.111.59, length 28
2021-04-23 08:40:44.840895 ARP, Request who-has 192.168.111.1 tell 192.168.111.59, length 28
2021-04-23 08:40:45.840991 ARP, Request who-has 192.168.111.1 tell 192.168.111.59, length 28
2021-04-23 08:40:48.848508 ARP, Request who-has 192.168.111.1 tell 192.168.111.59, length 28
2021-04-23 08:40:49.848895 ARP, Request who-has 192.168.111.1 tell 192.168.111.59, length 28
2021-04-23 08:40:50.848871 ARP, Request who-has 192.168.111.1 tell 192.168.111.59, length 28
2021-04-23 08:40:51.514011 ARP, Reply 192.168.111.59 is-at 52:54:00:60:ea:0e, length 28
2021-04-23 08:40:52.928400 ARP, Reply 192.168.111.59 is-at 52:54:00:60:ea:0e, length 28
2021-04-23 08:40:53.853881 ARP, Request who-has 192.168.111.1 tell 192.168.111.59, length 28
2021-04-23 08:40:54.852472 ARP, Request who-has 192.168.111.1 tell 192.168.111.59, length 28

每隔一段时间观察一下 VM 是否响应，但稍后我们会看到它是针对广播数据包的。似乎还有一个问题是 192.168.111.1 没有响应。无论出于何种原因，在 tcpdump 级别都看不到数据包。还要注意没有来自树莓派的 ICMP 数据包。现在，显示网关响应（这是另一台计算机上的“br0”。编辑：替换为更好的捕获示例，因此时间戳不同）：

$ sudo tcpdump -n -tttt -e -i br0 ether host 52:54:00:60:ea:0e
tcpdump: verbose output suppressed, use -v[v]... for full protocol decode
listening on br0, link-type EN10MB (Ethernet), snapshot length 262144 bytes
2021-04-23 22:25:17.434415 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 60: Request who-has 192.168.111.1 tell 192.168.111.59, length 46
2021-04-23 22:25:17.434432 xx:xx:xx:xx:xx:xx > 52:54:00:60:ea:0e, ethertype ARP (0x0806), length 42: Reply 192.168.111.1 is-at xx:xx:xx:xx:xx:xx, length 28
2021-04-23 22:25:20.440843 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 60: Request who-has 192.168.111.1 tell 192.168.111.59, length 46
2021-04-23 22:25:20.440859 xx:xx:xx:xx:xx:xx > 52:54:00:60:ea:0e, ethertype ARP (0x0806), length 42: Reply 192.168.111.1 is-at xx:xx:xx:xx:xx:xx, length 28
2021-04-23 22:25:21.438316 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 60: Request who-has 192.168.111.1 tell 192.168.111.59, length 46
2021-04-23 22:25:21.438332 xx:xx:xx:xx:xx:xx > 52:54:00:60:ea:0e, ethertype ARP (0x0806), length 42: Reply 192.168.111.1 is-at xx:xx:xx:xx:xx:xx, length 28
2021-04-23 22:25:22.438266 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 60: Request who-has 192.168.111.1 tell 192.168.111.59, length 46
2021-04-23 22:25:22.438283 xx:xx:xx:xx:xx:xx > 52:54:00:60:ea:0e, ethertype ARP (0x0806), length 42: Reply 192.168.111.1 is-at xx:xx:xx:xx:xx:xx, length 28
2021-04-23 22:25:25.446312 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 60: Request who-has 192.168.111.1 tell 192.168.111.59, length 46
2021-04-23 22:25:25.446329 xx:xx:xx:xx:xx:xx > 52:54:00:60:ea:0e, ethertype ARP (0x0806), length 42: Reply 192.168.111.1 is-at xx:xx:xx:xx:xx:xx, length 28
2021-04-23 22:25:26.446195 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 60: Request who-has 192.168.111.1 tell 192.168.111.59, length 46
2021-04-23 22:25:26.446211 xx:xx:xx:xx:xx:xx > 52:54:00:60:ea:0e, ethertype ARP (0x0806), length 42: Reply 192.168.111.1 is-at xx:xx:xx:xx:xx:xx, length 28

观察虚拟机传出数据包，ARP 完成。我不知道如何从我通过 VNC 与之通信的 VM 复制和粘贴，但它显示了一些已完成但陈旧的 ARP 条目以响应ip neigh命令，并且 tcpdump 显示来自 LAN 的一些 ARP 和广播数据包。

其他信息（与此问题无关的 MAC 已隐藏）：

$ brctl show br0
bridge name     bridge id               STP enabled     interfaces
br0             8000.3c7c3f0d9983       no              enp3s0
                                                        vnet0
$ brctl showmacs br0
port no mac addr                is local?       ageing timer
  1     xx:xx:xx:xx:xx:xx       no                 0.00
  1     3c:7c:3f:0d:99:83       yes                0.00
  1     3c:7c:3f:0d:99:83       yes                0.00
  2     52:54:00:60:ea:0e       no                 1.68
  1     xx:xx:xx:xx:xx:xx       no                 2.14
  1     xx:xx:xx:xx:xx:xx       no                36.84
  1     xx:xx:xx:xx:xx:xx       no                89.57
  1     xx:xx:xx:xx:xx:xx       no               226.51
  1     xx:xx:xx:xx:xx:xx       no                13.28
  1     xx:xx:xx:xx:xx:xx       no               165.68
  1     xx:xx:xx:xx:xx:xx       no               165.68
  1     xx:xx:xx:xx:xx:xx       no               265.02
  1     xx:xx:xx:xx:xx:xx       no                27.62
  2     fe:54:00:60:ea:0e       yes                0.00
  2     fe:54:00:60:ea:0e       yes                0.00

$ ip addr
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
       valid_lft forever preferred_lft forever
2: enp3s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq master br0 state UP group default qlen 1000
    link/ether 3c:7c:3f:0d:99:83 brd ff:ff:ff:ff:ff:ff
3: br0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
    link/ether 3c:7c:3f:0d:99:83 brd ff:ff:ff:ff:ff:ff
    inet 192.168.111.136/24 brd 192.168.111.255 scope global dynamic br0
       valid_lft 51547sec preferred_lft 51547sec
7: vnet0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel master br0 state UNKNOWN group default qlen 1000
    link/ether fe:54:00:60:ea:0e brd ff:ff:ff:ff:ff:ff

编辑：有趣的是，来自配置为交换机的 D-Link AC2600 Wi-Fi 千兆路由器的所有 ARP 数据包总是出现在主机上并到达虚拟机并回复：

$ sudo tcpdump -n -tttt -e -i br0 ether host aa:aa:aa:aa:aa:aa
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on br0, link-type EN10MB (Ethernet), capture size 262144 bytes
2021-04-23 22:45:51.463524 aa:aa:aa:aa:aa:aa > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 60: Request who-has 192.168.111.59 (ff:ff:ff:ff:ff:ff) tell 192.168.111.58, length 46
2021-04-23 22:45:51.463631 52:54:00:60:ea:0e > aa:aa:aa:aa:aa:aa, ethertype ARP (0x0806), length 42: Reply 192.168.111.59 is-at 52:54:00:60:ea:0e, length 28
2021-04-23 22:46:51.466955 aa:aa:aa:aa:aa:aa > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 60: Request who-has 192.168.111.59 (ff:ff:ff:ff:ff:ff) tell 192.168.111.58, length 46
2021-04-23 22:46:51.467030 52:54:00:60:ea:0e > aa:aa:aa:aa:aa:aa, ethertype ARP (0x0806), length 42: Reply 192.168.111.59 is-at 52:54:00:60:ea:0e, length 28
2021-04-23 22:47:51.466889 aa:aa:aa:aa:aa:aa > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 60: Request who-has 192.168.111.59 (ff:ff:ff:ff:ff:ff) tell 192.168.111.58, length 46
2021-04-23 22:47:51.466965 52:54:00:60:ea:0e > aa:aa:aa:aa:aa:aa, ethertype ARP (0x0806), length 42: Reply 192.168.111.59 is-at 52:54:00:60:ea:0e, length 28
2021-04-23 22:48:51.479096 aa:aa:aa:aa:aa:aa > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 60: Request who-has 192.168.111.59 (ff:ff:ff:ff:ff:ff) tell 192.168.111.58, length 46
2021-04-23 22:48:51.479178 52:54:00:60:ea:0e > aa:aa:aa:aa:aa:aa, ethertype ARP (0x0806), length 42: Reply 192.168.111.59 is-at 52:54:00:60:ea:0e, length 28

编辑 3 - 新配置测试：为了减少变量的数量，完成了以下操作：

  • 一台即将退役的 Ubuntu 16.04 服务器启动，提供了一个新的隔离 LAN。
  • 主机 Ubuntu 20.04 服务器直接连接到 16.04 服务器的 LAN 端 NIC。根本不涉及任何交换机，只有一根长以太网电缆。
  • 事情经过测试，似乎工作正常。通过 ssh 从我的主 LAN 访问所有内容，通过我的主静态 WAN IP 并通过我的测试 WAN 静态 IP 回到旧的 16.04 服务器。然后从那里到 20.04 主机服务器的链式 ssh 会话。
  • Ubuntu 16.04 VM 客户端已在主机上启动。
  • 尝试从旧的 16.04 网关服务器到客户端的“ping”。
  • 结果与原始配置相同。

网关旧 16.04 服务器上的 tcpdump：

doug@DOUG-64:~$ sudo tcpdump -n -tttt -e -i enp2s0 ether host 52:54:00:60:ea:0e
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on enp2s0, link-type EN10MB (Ethernet), capture size 262144 bytes
2021-04-26 15:10:00.701941 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 60: Request who-has 192.168.111.1 tell 192.168.111.59, length 46
2021-04-26 15:10:00.701965 00:19:b9:0d:af:fa > 52:54:00:60:ea:0e, ethertype ARP (0x0806), length 42: Reply 192.168.111.1 is-at 00:19:b9:0d:af:fa, length 28
2021-04-26 15:10:01.699156 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 60: Request who-has 192.168.111.1 tell 192.168.111.59, length 46
2021-04-26 15:10:01.699169 00:19:b9:0d:af:fa > 52:54:00:60:ea:0e, ethertype ARP (0x0806), length 42: Reply 192.168.111.1 is-at 00:19:b9:0d:af:fa, length 28
2021-04-26 15:10:02.699141 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 60: Request who-has 192.168.111.1 tell 192.168.111.59, length 46
2021-04-26 15:10:02.699154 00:19:b9:0d:af:fa > 52:54:00:60:ea:0e, ethertype ARP (0x0806), length 42: Reply 192.168.111.1 is-at 00:19:b9:0d:af:fa, length 28
2021-04-26 15:10:05.707404 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 60: Request who-has 192.168.111.1 tell 192.168.111.59, length 46
2021-04-26 15:10:05.707417 00:19:b9:0d:af:fa > 52:54:00:60:ea:0e, ethertype ARP (0x0806), length 42: Reply 192.168.111.1 is-at 00:19:b9:0d:af:fa, length 28
2021-04-26 15:10:06.707097 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 60: Request who-has 192.168.111.1 tell 192.168.111.59, length 46
2021-04-26 15:10:06.707110 00:19:b9:0d:af:fa > 52:54:00:60:ea:0e, ethertype ARP (0x0806), length 42: Reply 192.168.111.1 is-at 00:19:b9:0d:af:fa, length 28
2021-04-26 15:10:07.707094 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 60: Request who-has 192.168.111.1 tell 192.168.111.59, length 46
2021-04-26 15:10:07.707107 00:19:b9:0d:af:fa > 52:54:00:60:ea:0e, ethertype ARP (0x0806), length 42: Reply 192.168.111.1 is-at 00:19:b9:0d:af:fa, length 28

20.04 主机服务器上的 tcpdump：

doug@s19:~$ sudo tcpdump -n -tttt -e -i br0 ether host 52:54:00:60:ea:0e
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on br0, link-type EN10MB (Ethernet), capture size 262144 bytes
2021-04-26 15:11:35.801771 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 42: Request who-has 192.168.111.1 tell 192.168.111.59, length 28
2021-04-26 15:11:36.800497 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 42: Request who-has 192.168.111.1 tell 192.168.111.59, length 28
2021-04-26 15:11:37.800491 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 42: Request who-has 192.168.111.1 tell 192.168.111.59, length 28
2021-04-26 15:11:40.807062 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 42: Request who-has 192.168.111.1 tell 192.168.111.59, length 28
2021-04-26 15:11:41.804469 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 42: Request who-has 192.168.111.1 tell 192.168.111.59, length 28
2021-04-26 15:11:42.804444 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 42: Request who-has 192.168.111.1 tell 192.168.111.59, length 28
2021-04-26 15:11:45.812553 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 42: Request who-has 192.168.111.1 tell 192.168.111.59, length 28
2021-04-26 15:11:46.812405 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 42: Request who-has 192.168.111.1 tell 192.168.111.59, length 28
2021-04-26 15:11:47.812398 52:54:00:60:ea:0e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 42: Request who-has 192.168.111.1 tell 192.168.111.59, length 28

附带说明：从我的链接 ssh 会话到 20.04 主机服务器，我可以再次链接并 ssh 到 VM 客户端就好了。

结论： Ubuntu 20.04 服务器上的链路层出了点问题，因此 tcpdump 甚至无法“看到”传入的数据包，也无法到达 VM 来宾。图。

编辑 4：与 Christian Ehrhardt 提供的信息相比，我的系统上的潜在差异是 br0 MAC 列表可能不正确，替换了第一个字节。注意：不相关的 MAC 已删除，3 个 VM 正在运行：

doug@s19:~$ brctl showmacs br0
port no mac addr                is local?       ageing timer
  1     3c:7c:3f:0d:99:83       yes                0.00  <<< enp3s0, br0
  1     3c:7c:3f:0d:99:83       yes                0.00  <<< enp3s0, br0
  4     52:54:00:22:2f:dc       no                 5.15  <<< VM 3
  2     52:54:00:60:ea:0e       no                 3.29  <<< VM 1
  3     52:54:00:60:ea:3e       no                12.67  <<< VM 2
  4     fe:54:00:22:2f:dc       yes                0.00  <<< vnet2
  4     fe:54:00:22:2f:dc       yes                0.00  <<< vnet2
  2     fe:54:00:60:ea:0e       yes                0.00  <<< vnet0
  2     fe:54:00:60:ea:0e       yes                0.00  <<< vnet0
  3     fe:54:00:60:ea:3e       yes                0.00  <<< vnet1
  3     fe:54:00:60:ea:3e       yes                0.00  <<< vnet1

出于某种原因，Christian 没有显示 vnet，或者在他的系统上调用的任何内容，连接到网桥（我不知道它是否相关）：

$ brctl showmacs br0
port no mac addr        is local?   ageing timer
  2 52:54:00:48:40:69   no         2.36   <- Guest
  1 52:54:00:95:e4:2a   no         0.00   <- outside system
  1 52:54:00:9b:9b:0e   yes        0.00   <- Host
  1 52:54:00:9b:9b:0e   yes        0.00   <- Host

编辑 5：与编辑 4 类似的数据，但来自运行 2 个虚拟机的 Debian 服务器，运行正常：

doug@s15:~$ sudo brctl showmacs br0
port no mac addr                is local?       ageing timer
  1     52:54:00:22:2f:dc       no                17.85
  2     52:54:00:27:1b:5e       no                18.48  <<< VM 1
  3     52:54:00:27:1b:ae       no                 2.14  <<< VM 2
  1     f4:8c:eb:c8:08:a0       no                18.48
  2     fe:71:fa:75:16:93       yes                0.00  <<< tap0 (VM1)
  2     fe:71:fa:75:16:93       yes                0.00  <<< tap0
  3     fe:e1:c5:2a:c7:e3       yes                0.00  <<< tap1 (VM2)
  3     fe:e1:c5:2a:c7:e3       yes                0.00  <<< tap1

编辑 6：来自 的信息networkctl，需要注意的是 Debian 服务器显示“Master: br0”，而 Ubuntu 服务器没有：

Ubuntu：

doug@s19:~$ networkctl
IDX LINK   TYPE     OPERATIONAL SETUP
  1 lo     loopback carrier     unmanaged
  2 enp3s0 ether    enslaved    configured
  3 br0    bridge   routable    configured
  4 vnet0  ether    carrier     unmanaged

4 links listed.
doug@s19:~$ man networkctl
doug@s19:~$ networkctl^Cnetworkctl
doug@s19:~$ networkctl status vnet0
● 4: vnet0
             Link File: /usr/lib/systemd/network/99-default.link
          Network File: n/a
                  Type: ether
                 State: carrier (unmanaged)
                Driver: tun
            HW Address: fe:54:00:60:ea:0e
                   MTU: 1500 (min: 68, max: 65521)
  Queue Length (Tx/Rx): 1/1
      Auto negotiation: no
                 Speed: 10Mbps
                Duplex: full
                  Port: tp

Apr 30 07:40:51 s19 systemd-networkd[530]: vnet0: Link UP
Apr 30 07:40:51 s19 systemd-networkd[530]: vnet0: Gained carrier

德比安：

doug@s15:~$ networkctl
IDX LINK   TYPE     OPERATIONAL SETUP
  1 lo     loopback carrier     unmanaged
  2 enp3s0 ether    enslaved    configured
  3 enp1s0 ether    routable    configured
  4 br0    bridge   routable    configured
 10 tap0   ether    carrier     unmanaged

5 links listed.
doug@s15:~$ networkctl status tap0
● 10: tap0
             Link File: /usr/lib/systemd/network/99-default.link
          Network File: n/a
                  Type: ether
                 State: carrier (unmanaged)
                Driver: tun
            HW Address: fe:8a:6a:ce:18:9c
                   MTU: 1500 (min: 68, max: 65521)
                 QDisc: pfifo_fast
                Master: br0   <<<<< Different than Ubuntu
  Queue Length (Tx/Rx): 1/1
      Auto negotiation: no
                 Speed: 10Mbps
                Duplex: full
                  Port: tp

问题：出了什么问题，如何让桥接的 VM 正常工作？

networking

2 个回答

  1. 我们在 PowerEdge R740 和 Broadcomm 10GB NIC 上存在相同的 kvm vlan 网桥问题。我们发现的是

    　　博通进阶。双 10Gb 以太网 21.65.33.33 和内核 5.4 -> 坏
　　博通进阶。双 10Gb 以太网 21.65.33.33 和内核 5.5 -> 好
　　博通进阶。双 10Gb 以太网 21.80.16.95 和内核 5.5 -> 坏
　　博通进阶。双 10Gb 以太网 21.80.16.95 和内核 5.12 -> 坏

    但是，固件回滚不起作用，所以要小心。

    • 2
  2. Best Answer

    问题是内核。主机 20.04 服务器未使用 HWE 堆栈，因此当前内核为 5.4.0.72。

    然而，所有从主线 5.5-rc1 及更高版本测试的内核都可以正常工作，包括 5.12-rc6（我还没有安装 5.12）。还测试了一些较旧的 5.4 系列内核，包括主线 5.4.0 和 5.4.117，它们都失败了。

    似乎内核二分法不能向后工作，即当好的内核在坏内核之后。但是，如果只是颠倒好与坏的定义，那么也许。参考。

    编辑 1：这个 20.04 服务器需要 HWE 堆栈，因为硬件对于 5.4 系列内核来说是新的。我无法将内核一分为二，因为git bisect skip需要太多 s，因为太多内核无法启动或什至无法编译。这整件事是相当大的浪费时间。

    • 1

相关问题