Re: [Qemu-devel] libvirt fails to start guest with latest code checkins

[Laszlo Ersek]

On 02/01/13 23:41, Bruce Rogers wrote:

> Here is the invocation by libvirt, as recorded in 
> /var/log/libvirt/qemu/<vmname>:
> 2013-02-01 21:35:45.597+0000: starting up
> LC_ALL=C
> PATH=/sbin:/usr/sbin:/usr/local/sbin:/root/bin:/usr/local/bin:/usr/bin:
> /bin:/usr/bin/X11:/usr/X11R6/bin:/usr/games:/usr/lib/mit/bin:/usr/lib/mit/sbin
> HOME=/root USER=root LOGNAME=root TMPDIR=/tmp QEMU_AUDIO_DRV=none
> /usr/bin/qemu-kvm -name sles11 -S -M pc -cpu
> core2duo,+lahf_lm,+pdcm,+xtpr,+cx16,+tm2,+est,+vmx,+ds_cpl,+dtes64,+pbe,+tm,+ht,+ss,+acpi,+ds
> -enable-kvm -m 512 -smp 2,sockets=2,cores=1,threads=1 -uuid
> 36975709-16ae-85e0-c90f-09eb4ec7dec3 -no-user-config -nodefaults
> -chardev
> socket,id=charmonitor,path=/var/lib/libvirt/qemu/sles11.monitor,server,nowait
> -mon chardev=charmonitor,id=monitor,mode=control -rtc base=utc
> -no-shutdown -device piix3-usb-uhci,id=usb,bus=pci.0,addr=0x1.0x2
> -drive
> file=/var/lib/kvm/images/sles11/disk0.raw,if=none,id=drive-virtio-disk0,format=raw
> -device
> virtio-blk-pci,scsi=off,bus=pci.0,addr=0x4,drive=drive-virtio-disk0,id=virtio-disk0,bootindex=1
> -drive if=none,id=drive-ide0-0-0,readonly=on,format=raw -device
> ide-cd,bus=ide.0,unit=0,drive=drive-ide0-0-0,id=ide0-0-0 -netdev
> tap,fd=18,id=hostnet0 -device
> virtio-net-pci,netdev=hostnet0,id=net0,mac=52:54:00:59:31:da,bus=pci.0,addr=0x3
> -vnc 127.0.0.1:0 -vga cirrus -device
> virtio-balloon-pci,id=balloon0,bus=pci.0,addr=0x5
> WARNING: failed to find acpi-dsdt.aml
> qemu-kvm: -device
> virtio-net-pci,netdev=hostnet0,id=net0,mac=52:54:00:59:31:da,bus=pci.0,addr=0x3:
> Property 'virtio-net-pci.netdev' can't find value 'hostnet0'
> 2013-02-01 21:35:46.491+0000: shutting down

(The "WARNING: failed to find acpi-dsdt.aml" message is fixed by commit 
7e973bb2.)

I think the following could be happening.

    -netdev tap,fd=18,id=hostnet0

is specified on the command line (generated by libvirt). Accordingly,
the net_init_tap() function is entered with name="hostnet0".

Looking at the tree at 264986e2^ (= e5dc0b4):
- "tap->has_fd" is true,
- fd / vnet_hdr / model are set up in that branch,
- the net_init_tap() tail-calls net_init_tap_one(), with the configured
  local variables, including "name" (input param):

    return net_init_tap_one(tap, peer, model, name, ifname, script,
                    downscript, tap->has_vhostfd ? tap->vhostfd : NULL,
                    vnet_hdr, fd);

Commit 264986e2, among other things, instantiates the net_init_tap_one()
call for each branch of net_init_tap(), open-coding the parameters at
each call site that originally used to be "buffered up" for the common
final call. Unfortunately, there's a typo in the "tap->has_fd" branch:

        if (net_init_tap_one(tap, peer, "tap", NULL, NULL,
                             script, downscript,
                             vhostfdname, vnet_hdr, fd)) {
            return -1;
        }

The "name" parameter, which would ultimately go to
qemu_net_client_setup(), is lost, hence assign_name() is used instead.

The rest of the arguments seems OK. Suggested patch:

----[cut here]----
diff --git a/net/tap.c b/net/tap.c
index 1bf7609..48c254e 100644
--- a/net/tap.c
+++ b/net/tap.c
@@ -711,7 +711,7 @@ int net_init_tap(const NetClientOptions *opts, const char 
*name,
 
         vnet_hdr = tap_probe_vnet_hdr(fd);
 
-        if (net_init_tap_one(tap, peer, "tap", NULL, NULL,
+        if (net_init_tap_one(tap, peer, "tap", name, NULL,
                              script, downscript,
                              vhostfdname, vnet_hdr, fd)) {
             return -1;
----[cut here]----

Laszlo