December 2009 - Devel - libvirt List Archives

[libvirt] [PATCH 0/2] Xen: Add support for interface model='netfront'
by Jiri Denemark 04 Dec '09

04 Dec '09

Xen HVM guests with PV drivers end up with two network interfaces for each configured interface. One of them being emulated by qemu and the other one paravirtual. As this might not be desirable, the attached patch provides a way for users to specify that only paravirtual network interface should be presented to the guest. The configuration was inspired by qemu/kvm driver, for which users can specify model='virtio' to use paravirtual network interface. The patch adds support for model='netfront' which results in type=netfront instead of type=ioemu (or nothing for newer xen versions) in guests native configuration. Xen's qemu ignores interfaces with type != ioemu and only paravirtual network device will be seen in the guest. Four possible configuration scenarios follow: - no model specified in domain's XML - libvirt will behave like before this change; it will set type=ioemu for HVM guests on xen host which is not newer than XEND_CONFIG_MAX_VERS_NET_TYPE_IOEMU - covered by existing tests - PV guest, any model - no functional change, model is passed as is (and ignored by the hypervisor) - covered by existing tests (e.g., *-net-e1000.*) - HVM guest, model=netfront - type is set to "netfront", model is not specified - covered by new *-net-netfront.* tests - HVM guest, model != netfront - type is set to "ioemu", model is passed as is - covered by new *-net-ioemu.* tests The fourth scenario feels like a regression for xen newer than XEND_CONFIG_MAX_VERS_NET_TYPE_IOEMU as users who had a model specified in their guest's configuration won't see a paravirtual interface in their guests any more. On the other hand, the reason for specifying a model is most likely the fact that they want to use such model which implies emulated interface. Users of older xen won't be affected at all as their xen provides paravirtual interface regardless of the type used. In previous version of this patch, I used "xenpv" instead of "netfront" as it looked more like kvm/qemu's virtio. However, you can find several tips how to force PV network interface for HVM guest by using type="netfront" so I changed it to conform with those tips. It also matches guest's driver name used for that device. Jiri Denemark (2): Support for interface model='netfront' Tests for interface type/model configuration src/xen/xend_internal.c | 32 +++++++++++---- src/xen/xm_internal.c | 38 +++++++++++++++-- tests/sexpr2xmldata/sexpr2xml-fv-net-ioemu.sexpr | 1 + tests/sexpr2xmldata/sexpr2xml-fv-net-ioemu.xml | 42 +++++++++++++++++++ .../sexpr2xmldata/sexpr2xml-fv-net-netfront.sexpr | 1 + tests/sexpr2xmldata/sexpr2xml-fv-net-netfront.xml | 42 +++++++++++++++++++ tests/sexpr2xmltest.c | 3 + tests/xmconfigdata/test-fullvirt-net-ioemu.cfg | 25 +++++++++++ tests/xmconfigdata/test-fullvirt-net-ioemu.xml | 43 ++++++++++++++++++++ tests/xmconfigdata/test-fullvirt-net-netfront.cfg | 25 +++++++++++ tests/xmconfigdata/test-fullvirt-net-netfront.xml | 43 ++++++++++++++++++++ tests/xmconfigtest.c | 3 + tests/xml2sexprdata/xml2sexpr-fv-net-ioemu.sexpr | 1 + tests/xml2sexprdata/xml2sexpr-fv-net-ioemu.xml | 31 ++++++++++++++ .../xml2sexprdata/xml2sexpr-fv-net-netfront.sexpr | 1 + tests/xml2sexprdata/xml2sexpr-fv-net-netfront.xml | 31 ++++++++++++++ tests/xml2sexprtest.c | 3 + 17 files changed, 352 insertions(+), 13 deletions(-) create mode 100644 tests/sexpr2xmldata/sexpr2xml-fv-net-ioemu.sexpr create mode 100644 tests/sexpr2xmldata/sexpr2xml-fv-net-ioemu.xml create mode 100644 tests/sexpr2xmldata/sexpr2xml-fv-net-netfront.sexpr create mode 100644 tests/sexpr2xmldata/sexpr2xml-fv-net-netfront.xml create mode 100644 tests/xmconfigdata/test-fullvirt-net-ioemu.cfg create mode 100644 tests/xmconfigdata/test-fullvirt-net-ioemu.xml create mode 100644 tests/xmconfigdata/test-fullvirt-net-netfront.cfg create mode 100644 tests/xmconfigdata/test-fullvirt-net-netfront.xml create mode 100644 tests/xml2sexprdata/xml2sexpr-fv-net-ioemu.sexpr create mode 100644 tests/xml2sexprdata/xml2sexpr-fv-net-ioemu.xml create mode 100644 tests/xml2sexprdata/xml2sexpr-fv-net-netfront.sexpr create mode 100644 tests/xml2sexprdata/xml2sexpr-fv-net-netfront.xml

3 4

[libvirt] [PATCH] Support for virtualbox version 3.1
by pritesh 04 Dec '09

04 Dec '09

Hi All, I have just added support for VirtualBox Version 3.1 (works for the beta as well) in the libvirt vbox driver. The patch for the same is attached below. Regards, Pritesh

3 8

[libvirt] libvirtd crashes
by Shi Jin 04 Dec '09

04 Dec '09

Hi there, My libvirtd built from the latest git code keeps on crashing on all machines. I turned on debugging and this is the information I have in the log file before crashing: 14:31:50.828: debug : virEventUpdateHandleImpl:146 : Update handle w=110 e=12 14:31:50.828: debug : virEventInterruptLocked:662 : Interrupting 14:31:50.828: debug : qemuMonitorCommandWithHandler:271 : Receive command reply ret=-1 errno=104 0 bytes '(null)' 14:31:50.828: error : qemuMonitorCommandWithHandler:290 : cannot send monitor command 'info cpus': Connection reset by peer 14:31:50.828: error : qemuMonitorTextGetCPUInfo:436 : internal error cannot run monitor command to fetch CPU thread info I am not sure if there is any other information needed to help identify the problem. My building options are: ./autogen.sh --prefix=/usr --sysconfdir=/etc --localstatedir=/var --without-xen --with-qemu --with-qemu-user=oneadmin --with-qemu-group=oneadmin --without-uml --without-vbox --without-openvz --without-lxc Please help me here. I can accept the service failing queries from time to time since I have error handling written so that they can be re-tried. But a crashing libvirtd takes the whole thing down. Thanks a lot. Shi -- Shi Jin, PhD

2 3

[libvirt] ANNOUNCE: New releases virt-manager 0.8.1 and virtinst 0.500.1
by Cole Robinson 03 Dec '09

03 Dec '09

I'm happy to announce two new releases: virt-manager 0.8.1: virt-manager is a desktop application for managing KVM and Xen virtual machines via libvirt. virtinst 0.500.1: virtinst is a collection of command line tools for provisioning libvirt virtual machines, including virt-install and virt-clone. The releases can be downloaded from: http://virt-manager.org/download.html The direct download links are: http://virt-manager.org/download/sources/virt-manager/virt-manager-0.8.1.ta… http://virt-manager.org/download/sources/virtinst/virtinst-0.500.1.tar.gz The virt-manager release includes: - VM Migration wizard, exposing various migration options - Enumerate CDROM and bridge devices on remote connections - Can once again list multiple graphs in the manager window (Jon Nordby) - Support disabling dhcp (Michal Novotny), and specifying 'routed' type for new virtual networks - Support storage pool source enumeration for LVM, NFS, and SCSI - Allow changing VM ACPI, APIC, clock offset, individual vcpu pinning, and video model (vga, cirrus, etc.) - Many improvements and bugfixes The virtinst release includes: - virt-install now attempts --os-variant detection by default. This can be disabled with '--os-variant none' (distro detection currently only works for URL installs) - New --disk option 'format', for creating image disks formats such as qcow2 or vmdk - Many improvements and bugfixes Thanks to everyone who has contributed to this release through testing, bug reporting, submitting patches, and otherwise sending in feedback! Thanks, Cole

1 0

[libvirt] both libvirt 0.7.4 and latest git crashing
by Nikola Ciprich 03 Dec '09

03 Dec '09

Hi, few days ago, I've reported problem with libvirt getting stuck while saving kvm. Today, I tried latest git and it seems the problem is solved which is great. Unfortunately, there seems to be another problem (which I noticed even for 0.7.4 although haven't reported yet) is that libvirtd sometimes crashes. I'm not 100% sure which action causes it to crash as it's being executed by cluster resource manager, but the backtrace might help those familiar with it's internals. (I think it's domain start) (gdb) bt #0 0x00007f4f46d03cd0 in pthread_mutex_unlock () from /lib64/libpthread.so.0 #1 0x000000000042da91 in qemuDomainObjExitMonitor (obj=0x980a80) at qemu/qemu_driver.c:287 #2 0x000000000043ca84 in qemudDomainGetInfo (dom=0x978d50, info=0x42690de0) at qemu/qemu_driver.c:3294 #3 0x00007f4f47ddf3fb in virDomainGetInfo (domain=0x978d50, info=0x42690de0) at libvirt.c:2786 #4 0x000000000041d79e in remoteDispatchDomainGetInfo (server=<value optimized out>, client=<value optimized out>, conn=0x961840, hdr=<value optimized out>, rerr=0x42690e70, args=<value optimized out>, ret=0x42690f50) at remote.c:1076 #5 0x000000000041f35c in remoteDispatchClientCall (server=0x95b780, client=0x97e310, msg=0x980b00) at dispatch.c:506 #6 0x000000000041f712 in remoteDispatchClientRequest (server=0x95b780, client=0x97e310, msg=0x980b00) at dispatch.c:388 #7 0x00000000004147e2 in qemudWorker (data=<value optimized out>) at libvirtd.c:1518 #8 0x00007f4f46d004a7 in start_thread () from /lib64/libpthread.so.0 #9 0x00007f4f46a76c2d in clone () from /lib64/libc.so.6 Could anyone please have a look at it? thanks a lot in advance nik -- ------------------------------------- Nikola CIPRICH LinuxBox.cz, s.r.o. 28. rijna 168, 709 01 Ostrava tel.: +420 596 603 142 fax: +420 596 621 273 mobil: +420 777 093 799 www.linuxbox.cz mobil servis: +420 737 238 656 email servis: servis(a)linuxbox.cz -------------------------------------

2 1

[libvirt] cannot read libvirt.org website...
by Matthew Woehlke 03 Dec '09

03 Dec '09

...because of this (in http://libvirt.org/generic.css) body { margin: 0em; padding: 0px; color: rgb(0,0,0); font-family: Verdana, Arial, Helvetica, sans-serif; font-size: 80%; // font-size: 83%; } Black text on unspecified background is not legible on my setup. A background color *must* be specified when a foreground color is specified. Please fix this. (Apologies if this is not the right spot, but alas there is no webmaster AT libvirt.org, and I did not find anything obvious on who is the webmaster. If I should send this elsewhere, please let me know where.) -- Matthew Please do not quote my e-mail address unobfuscated in message bodies. -- Obscurity: +5

2 2

[libvirt] Move esxVMX_IndexToDiskName to util.c
by Matthias Bolte 03 Dec '09

03 Dec '09

I needed the inverse function to virDiskNameToIndex() for the ESX driver and added it to esx_vmx.c. The pending VirtualBox 3.1 patch needs disk index to disk name mapping too. So I want to move esxVMX_IndexToDiskName() to util.c. esxVMX_IndexToDiskName() handles indices up to 701. This limit comes from a gap in the disk name to disk index mapping of virDiskNameToIndex(): sdzy -> 700 sdzz -> 701 sdaaa -> 728 sdaab -> 729 This line in virDiskNameToIndex() causes this gap: idx = (idx + i) * 26; It can be fixed by altering this line to: idx = (idx + (i < 1 ? 0 : 1)) * 26; But this change breaks compatibility for indices > 701. So, I made two patches for either option and ask you for comments. Matthias

3 3

[libvirt] About libvirt API Support event notification mechanism problem
by ajia 03 Dec '09

03 Dec '09

Dear Daniel, Recently, I am writing test plan about libvirt asynchronous event mechanism, refering to the following bugs from the bugzilla: spice/vnc: https://bugzilla.redhat.com/show_bug.cgi?id=515268 disk error: https://bugzilla.redhat.com/show_bug.cgi?id=515270 I know libvirt can communicate with qemu monitor by QMP at present, can I use libvirt API to test it now?libvirt-0.7.4-1is ok? In addition,In qemu monitor with rhel6(2.6.31-31.el6.x86_64),"notify" command is nonexistent (Fedora12 is existent),if current qemu can't support event notification mechanism in rhel6,from libvirt point of view,I don't clear libvirt how to get asyns event notification?so I don't know how to test it.and I see qemu monitor protocol being patched,I am not sure current QMP status. I would be most grateful for any response . thank you very much. Best Regards, Alex Jia

2 1

[libvirt] [PATCH] Fix some locking issues
by Matthias Bolte 03 Dec '09

03 Dec '09

* src/conf/domain_conf.c: don't call virDomainObjUnlock twice * src/qemu/qemu_driver.c: relock driver lock if an error occurs in qemuDomainObjBeginJobWithDriver, enter/exit monitor with driver in qemudDomainSave --- src/conf/domain_conf.c | 1 - src/qemu/qemu_driver.c | 9 +++++---- 2 files changed, 5 insertions(+), 5 deletions(-) diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c index e9f79b7..21d509d 100644 --- a/src/conf/domain_conf.c +++ b/src/conf/domain_conf.c @@ -5230,7 +5230,6 @@ virDomainObjIsDuplicate(virDomainObjListPtr doms, } dupVM = 1; - virDomainObjUnlock(vm); } else { /* UUID does not match, but if a name matches, refuse it */ vm = virDomainFindByName(doms, def->name); diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index bf4557e..2a63de9 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -353,6 +353,7 @@ static int qemuDomainObjBeginJobWithDriver(struct qemud_driver *driver, else virReportSystemError(NULL, errno, "%s", _("cannot acquire job mutex")); + qemuDriverLock(driver); return -1; } } @@ -3612,9 +3613,9 @@ static int qemudDomainSave(virDomainPtr dom, if (header.compressed == QEMUD_SAVE_FORMAT_RAW) { const char *args[] = { "cat", NULL }; qemuDomainObjPrivatePtr priv = vm->privateData; - qemuDomainObjEnterMonitor(vm); + qemuDomainObjEnterMonitorWithDriver(driver, vm); rc = qemuMonitorMigrateToCommand(priv->mon, 0, args, path); - qemuDomainObjExitMonitor(vm); + qemuDomainObjExitMonitorWithDriver(driver, vm); } else { const char *prog = qemudSaveCompressionTypeToString(header.compressed); qemuDomainObjPrivatePtr priv = vm->privateData; @@ -3623,9 +3624,9 @@ static int qemudDomainSave(virDomainPtr dom, "-c", NULL }; - qemuDomainObjEnterMonitor(vm); + qemuDomainObjEnterMonitorWithDriver(driver, vm); rc = qemuMonitorMigrateToCommand(priv->mon, 0, args, path); - qemuDomainObjExitMonitor(vm); + qemuDomainObjExitMonitorWithDriver(driver, vm); } if (rc < 0) -- 1.6.0.4

2 1

Re: [libvirt] [PATCH v2] Disk- and Controller Hotplug
by Wolfgang Mauerer 03 Dec '09

03 Dec '09

Hi, Some time ago, Daniel P. Berrange wrote: > On Tue, Nov 24, 2009 at 02:15:58PM +0100, Wolfgang Mauerer wrote: >> Daniel P. Berrange wrote: >>> On Mon, Nov 23, 2009 at 02:15:06PM +0100, Wolfgang Mauerer wrote: >>>> Daniel P. Berrange wrote: >>>>> Wolfgang Mauerer wrote: >> okay, to avoid flooding the mailing list with nearly identical >> patches, I've put the rebased versions into a repository at >> git://git.kiszka.org/libvirt.git queue > > I've been taking a closer look at this, and I think the final state > of patch series as a whole looks good. The minor issue is that the > intermediate patches don't compile, since they rely on compile fixes > at the end of the series. Rather than ask you to re-format the > patches yet again, I'm going to just merge the patches into a smaller > set myself. yes, I've sacrificed individual compilability when I did the forward porting because I the amount of time I can dedicate to this right now is a bit restricted for various reasons. Thanks for doing the missing cleanups! > I've got just one question I'd like to understand before doing this. > On the <disk> element's new <controller> element, you are allowing > two mutually exclusive ways of specifying the controller > > <disk> ... <controller name="<identifier>" bus="<number>" > unit="<number>"/> </disk> > > and > > <disk> ... <controller pci_addr="<addr>" bus="<number>" > unit="<number>"/> </disk> > > I think it is going to cause unneccessary pain for applications to > have to cope with two different ways of specifying the same > relationship. So my intention is to remove the 'pci_addr' variant, > since that information can easily be got directly from the separate > top level <controller> device itself Getting rid of pci_addr is perfectly okay. I've kept it in the current series for setups that don't provide <controller>s, but it's supposed to be eliminated by design. Just go ahead if you want to do this it already now ;-) > On a side note, once we've applied this initial series, I think we'll > also need to be automatically adding a <controller> element in all > guest domains which have IDE or SCSI controllers present by default. > This is going to also force us to hook into QEMU's 'info pci' monitor > command to figure out their PCI address. This is long overdue though > and needed for NIC & Disk devices added at startup too, so this is > good motivation :-) agreed. I can try to address this when I bring back hot-removal. Best regards, Wolfgang -- Siemens AG, Corporate Technology CT T DE IT 1 Corporate Competence Centre Embedded Linux

2 1