Hi, vde shows some strange behavior when vde_switch is started without the -m and -g command line options. Without a modified /etc/init.d/vde script: --exec /usr/bin/vde_switch -- -tap ${VDE_TAP} -sock /var/run/vde.ctl -daemon the sockets vde.ctl.xxxxx will be placed directly in the /var/run - directory and not in the /var/run/vde.ctl/ directory! The result in my case is that my qemu/kvm guests are not able to communicate to the net... When I add "-m 770" and "-g kvm" (because I use qemu/kvm) and "-M /var/run/vde.mgmt" (because I like to control my vlans using unixterm) to the vde_switch line in /etc/init.d/vde everything works. I think these options should have their concerning env.var in /etc/conf.d/vde ########################## here my modified /etc/conf.d/vde: # load the tun module VDE_MODPROBE_TUN="yes" # virtual tap networking device to be used for vde VDE_TAP="tap0" # mode and group for the socket VDE_SOCK_CHMOD="770" # # instead of VDE_SOCK_CHOWN now VDE_SOCK_GROUP #VDE_SOCK_CHOWN=":qemu" VDE_SOCK_GROUP="qemu" # VDE_SOCK_CTLDIR="/var/run/vde.ctl" VDE_SOCK_MGMT="/var/run/vde.mgmt" # # #################################### here the modified part of /etc/init.d/vde: start() { ebegin "Starting vde" [ "${VDE_MODPROBE_TUN}" == "yes" ] && modprobe tun start-stop-daemon --start --quiet \ --exec /usr/bin/vde_switch -- -tap ${VDE_TAP} \ -m ${VDE_SOCK_CHMOD} \ -g ${VDE_SOCK_GROUP} \ -sock ${VDE_SOCK_CTLDIR} \ -M ${VDE_SOCK_MGMT} -daemon eend $? "Failed to start vde" chmod -R ${VDE_SOCK_CHMOD} ${VDE_SOCK_CTLDIR} chown -R :${VDE_SOCK_GROUP} ${VDE_SOCK_CTLDIR} chmod -R ${VDE_SOCK_CHMOD} ${VDE_SOCK_MGMT} chown -R :${VDE_SOCK_GROUP} ${VDE_SOCK_MGMT} } Reproducible: Always
I thought the modified config/startscript solves the problem but after a reboot I had the same problem(no working network) again when I start qemu/kvm using the following line. :-( vdeq qemu-system-x86_64 \ -net nic,macaddr=52:54:00:00:EE:A2 \ -net vde,sock=/var/run/vde.ctl \ -m 256 -localtime \ -soundhw all \ -boot c \ -hda fedora32.img \ 1>fc32.log 2>fc32.err & Further investigation shows that vde_switch must have some problems with a still existing /var/run/vde.ctl directory + files in this directory from former runs (normally this directory is automatically deleted). So I have to remove a eventually existing /var/run/vde.ctl directory before I can run vde_switch. This behavior is reproducible: Kill vde_switch after last guest was terminated and then restart vde with /etc/init.d/vde restart. Start a new qemu guest and then no working network at all. Note: the mode variable in qemu/vdeq.c is sometimes not initialized. And the vdeq -m parameter for mode settings" may conflict with qemu's -m parameter.
Just back from holiday, expect some more later this week/early next week.
paludis complains that there are no whitespaces in the DEPEND String after '(' and before ')'
I'm awfully sorry for posting in the wrong bug :(
I had similar issues which were fixed by the recent ebuild bump (vde-2.2.2). Give it a try!
Can anyone else reproduce this in recent ebuilds? I can't. Please close otherwise..
This is working with 2.2.2, as far as I can tell.
