Hi; this is a bug report I'm forwarding for somebody else. The problem is that QEMU's tun/tap support in net/tap-linux.c only supports the legacy /dev/net/tun interface (the filename is hardcoded). If you created the tap device via macvtap, then this is the wrong filename, but there's no code in QEMU to get the correct /dev/tapN filename from the user-provided ifname=whatever option.
Talking on IRC, the answer suggested was that we ought to do a SIOCGIFINDEX ioctl passing in the user-specified ifname string, which then gives you the index N to construct the /dev/tapN filename. There is probably complexity in working out whether we should do that or use the legacy interface (or try both always?) This is probably overall not a very large amount of code, though. The usual suggested workaround is to use the -netdev fd option, like fd=3 3<>/dev/tap$(< /sys/class/net/tap0/ifindex) (which gets the shell to open the right /dev/tap device). Unfortunately this isn't compatible with multi-queue support because netdev complains "ifname=, script=, downscript=, vnet_hdr=, helper=, queues=, fds=, and vhostfds= are invalid with fd=" so you can't pass options like "queues=4"... Detailed repro case below if required (uses aarch64 but none of this is arch-specific as far as I know). ------------<Begin Reproducer>-------------- To reproduce, here are the steps I followed. Create a tap device on the host =============================== # ip link add link eth0 name tap0 type macvtap mode bridge To create the tap device the host kernel needs CONFIG_MACVTAP enabled. You can check that the device is created by running $ ip a 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1 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 inet6 ::1/128 scope host valid_lft forever preferred_lft forever 2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP group default qlen 1000 link/ether 02:a1:a2:a3:a4:a5 brd ff:ff:ff:ff:ff:ff inet 10.1.194.30/23 brd 10.1.195.255 scope global eth0 valid_lft forever preferred_lft forever inet6 fe80::a1:a2ff:fea3:a4a5/64 scope link valid_lft forever preferred_lft forever [...] 8: tap0@eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP group default qlen 500 link/ether 5e:fc:d2:28:f2:f9 brd ff:ff:ff:ff:ff:ff inet6 fe80::5cfc:d2ff:fe28:f2f9/64 scope link valid_lft forever preferred_lft forever Creating a tap device creates a character device. The character device created is "/dev/tap" followed by the interface index (8 in our example above). The interface index can also be discovered by reading a node in sysfs. $ cat /sys/class/net/tap0/ifindex 8 Modify tap device permissions ============================= By default, the device node is created with permissions root:root. $ ls -l /dev/tap8 crw------- 1 root root 242, 1 Nov 30 11:57 /dev/tap8 Update the permissions so that qemu-system-aarch64 can access the node. # chgrp kvm /dev/tap8 # chmod g+rw /dev/tap8 $ ls -l /dev/tap8 crw-rw---- 1 root kvm 242, 1 Nov 30 11:57 /dev/tap8 Run qemu ======== I am using a locally built qemu (tag: v2.8.0-rc2). When attempting to start a guest I get the following complaint - $ qemu-system-aarch64 -M virt -m 4096 -cpu host -smp 4 -enable-kvm -nographic \ -pflash flash0.img -pflash flash1.img \ -drive if=none,file=hda.qcow2,id=hd0" \ -device virtio-blk-device,drive=hd0 \ -netdev tap,id=eth0,script=,ifname=tap0,queues=4 \ -device virtio-net-pci,netdev=eth0,mq=on,vectors=12,mac=22:22:3a:57:07:ed qemu-system-aarch64: -netdev tap,id=eth0,script=,ifname=tap0,queues=4: could not open /dev/net/tun: No such file or directory ------------<End of Reproducer>-------------- thanks -- PMM