semantic: 0.176 other: 0.150 network: 0.089 graphic: 0.086 KVM: 0.084 PID: 0.070 device: 0.068 socket: 0.067 performance: 0.046 debug: 0.039 files: 0.036 permissions: 0.035 boot: 0.033 vnc: 0.019 debug: 0.608 network: 0.131 KVM: 0.114 PID: 0.037 other: 0.020 performance: 0.016 socket: 0.016 semantic: 0.015 files: 0.015 device: 0.009 vnc: 0.006 boot: 0.006 graphic: 0.004 permissions: 0.003 Silently fail with wrong vde socket dir Hi, Using qemu 0.12.5, kvm silently fail with exit code 1 when using -net vde and a wrong path for sock. Actually, the sock option is mean to be the socket dir of the vde_switch, not the socket itself. With -net vde,sock=/var/run/vde/vde0/ctl , strace ends with the following messages : connect(7, {sa_family=AF_FILE, path="/var/run/vde/vde0/ctl/ctl"}, 110) = -1 ENOTDIR (Not a directory) close(7) = 0 close(8) = 0 exit_group(1) = ? root ~# Please add a meaningful message. Regards, Étienne Hello, Could you please provide more data, what kinda of system and version are you running on? Jes There's no need to add any more specific information. The bug's in the code in qemu. net/vde.c: static int net_vde_init(VLANState *vlan, const char *model, const char *name, const char *sock, int port, const char *group, int mode) { ... vde = vde_open(init_sock, (char *)"QEMU", &args); if (!vde){ return -1; } ... } There's no message generated there. Callers merely pass the failure up the road, where it's finally handled as exit(1). If _anything_ is wrong in vde_open() (which can fail due to variety of reasons, including wrong path to the listening socket and what not), nothing will indicate that, just a trivial, silent exit. Given that you know what the problem is, it would probably have been faster to post a patch than just updating the bug and marking it confirmed.... A fix for this problem has finally been contributed here: https://git.qemu.org/?p=qemu.git;a=commitdiff;h=7587855cd23755a7a6bd