September 2009 - Devel - libvirt List Archives

[libvirt] Some problem with the save function
by Daniel Berteaud 25 Sep '09

25 Sep '09

Hi everyone. This is my first post to this list, so excuse me if it's not the right place to poste info like this. I've found some problems in qemu_driver.c in the latest versions of libvirt: - the first problem was also present in 0.7.0. If we run qemu guests unprivileged, we cannot use the save function (virsh save guestname /path/to/file). The reason seems to be that libvirt first create the destination file to write the header and the XML definition. The file is owned by root with 600 permission. Then, we ask qemu to append the dump of the memory (via the migrate exec: call) to this file. This part doesn't work because the qemu user cannot write to this file. So we endup with a file which only contains the header and the XML, and the guest is stoped :/ - the second problem is present since libvirt 0.7.1. Now that the saved file can be compressed, it seems we cannot save in a raw format any more. This is due to this part in the code (qemu_driver.c): if (STREQ (prog, "raw")) prog = "cat"; internalret = virAsprintf(&command, "migrate \"exec:" "%s -c >> '%s' 2>/dev/null\"", prog, safe_path); which result in "migrate \"exec cat -c >> safe_path 2>/dev/null\"" But cat doesn't support the -c argument, so once again, the save fails, as we end up with a save file which only contains the header and the XML definition. Unfortunately, my C knowledge is near 0, so I cannot correct this myself. - the third problem, is that restore doesn't work any more. I haven't dig it yet, but when I try to restore a guest using a saved file (uncompressed produced under libvirt 0.7.0), I've this error: virsh restore /tmp/guest.state error: Failed to restore domain from /tmp/guest.state error: internal error unable to start guest: Nothing interesting in /var/log/libvirt/qemu/guest.log: LC_ALL=C PATH=/sbin:/usr/sbin:/bin:/usr/bin HOME=/root USER=root LOGNAME=root /usr/bin/qemu-kvm -S -M pc -m 2048 -smp 4 -name guest -uuid 37b5937b-eacf-fda6-f4d0-33537d61b414 -monitor unix:/var/lib/libvirt/qemu/guest.monitor,server,nowait -localtime -no-reboot -boot c -drive file=/dev/data/guest,if=ide,index=0,boot=on -drive file=/var/lib/libvirt/iso/boot.iso,if=ide,media=cdrom,index=2 -net nic,macaddr=52:54:00:48:c4:ca,vlan=0,name=nic.0 -net tap,fd=25,vlan=0,name=tap.0 -serial pty -parallel none -usb -usbdevice tablet -vnc 127.0.0.1:0 -k fr -vga cirrus -incoming exec:cat The qemu process eats 100% of CPU during a few seconds, then stops (if I strace it during this time, I see a lot of error like bad file descriptor). Anyone have save/restore functions working with 0.7.1 ? Cheers, Daniel -- Daniel Berteaud FIREWALL-SERVICES SARL. Société de Services en Logiciels Libres Technopôle Montesquieu 33650 MARTILLAC Tel : 05 56 64 15 32 Fax : 05 56 64 15 32 Mail: daniel(a)firewall-services.com Web : http://www.firewall-services.com

5 14

[libvirt] qemu_driver migrateuri handling broken?
by Gregor Schaffrath 24 Sep '09

24 Sep '09

Hi all. Short summary on DV's request ;) I ran into a problem migrating kvm machines with libvirt-0.6.5: 1) At first, using the same syntax for the migrateuri as with xen (just the IP) did not work... looking into the source code (! ;) ), I found a different syntax for qemu. 2) using tcp://<ip>:<port> just produced an 'unknown failure' on the receiving side: root@loadgen137:~> virsh -c qemu:///system migrate --live kvm-testnode-vnode3 qemu+tcp://10.192.11.136/system?no_verify=1 tcp://10.192.11.136:12345 error: Unknown failure (Note: it was working like a charm when I eliminated the migrateuri altogether, but this was not an option in all my work environments, since one of them doesn't feature DNS, breaking migration on the hostname alone (and keeping /etc/hosts in sync on n machines is something I'd like to avoid, if possible ;) )) 3) removing the case distinction and the handling of the migrateuri in the qemudDomainMigratePrepare2 function in qemu_driver.c entirely (if-statement, and full else-part) solved both my issues. --- The issue seems to be current with 0.7.1 as well (at least I did receive the same error message - I did not investigate further) --- My 2ct: ;) I think that a wrapper like libvirt should consequently abstract lower level details (such as 'qemu', or 'xen') as much as possible (that's what libvirt is for, I guess :) ). Since the removal of the URI handling solves both the issue _and_ eliminates a case distinction between xen and qemu (painlessly, as far as I can see), I'd suggest perhaps considering the removal. :) As to the sending side specification of transfer ports: I think there should always be a possiblity to leave port selection up to libvirt, as the sending side does not necessarily an oversight over the free ports on the receiving side. (I feel that the reason for the different syntax including mandatory :port part was motivated by the wish to allow users to specify ports on the sending side) Cheers, Gregor Schaffrath.

3 5

[libvirt] [PATCH] Fix a typo in virDiskHasValidPciAddr()
by Jiri Denemark 24 Sep '09

24 Sep '09

Signed-off-by: Jiri Denemark <jdenemar(a)redhat.com> --- src/conf/domain_conf.h | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/src/conf/domain_conf.h b/src/conf/domain_conf.h index 09368d9..d494e54 100644 --- a/src/conf/domain_conf.h +++ b/src/conf/domain_conf.h @@ -125,7 +125,7 @@ struct _virDomainDiskDef { static inline int virDiskHasValidPciAddr(virDomainDiskDefPtr def) { - return def->pci_addr.domain || def->pci_addr.domain || def->pci_addr.slot; + return def->pci_addr.domain || def->pci_addr.bus || def->pci_addr.slot; } -- 1.6.5.rc1

3 2

[libvirt] "Unable to read QEMU help output: Interrupted system call" doing virDomainRestore() on current master
by Charles Duffy 23 Sep '09

23 Sep '09

Howdy. I've had this issue since yesterday, but avoided reporting it until determining today that it occurs on an unmodified upstream tree as well as my own local branch: $ virsh restore ramsave error: Failed to restore domain from ramsave error: Unable to read QEMU help output: Interrupted system call This happens immediately (no delay) and only on restore; "virsh start" behaves as usual. When running libvirtd under gdb, the failure mode is different: $ virsh restore ramsave error: Failed to restore domain from ramsave error: internal error Timed out while reading monitor startup output When running libvirtd under strace, this shows the binary being run (["/usr/bin/qemu-kvm", "-help"]) failing with a SIGPIPE while attempting to write to stdout. strace also shows a SIGSEGV being delivered to a libvirtd thread (a few clone()s removed from the parent) after that thread attempts to close all its open file descriptors above 0-2, but without being able to run gdb I'm at a bit of a loss as to what that thread is doing. Any suggestions?

2 1

[libvirt] [PATCH 0/5] Python fixes
by Cole Robinson 23 Sep '09

23 Sep '09

The following series fixes some functional (and cosmetic) issues with the python bindings. Thanks, Cole Cole Robinson (5): python: Don't generate conflicting conn.createXML functions. python: Add a newline between virConnect and virNode for readability. python: Don't generate bindings for vir*Ref python: Use a pure python implementation of 'vir*GetConnect' python: Fix generated virInterface method names python/generator.py | 40 +++++++++++++++++++++++++++++++++++++--- python/virConnect.py | 1 + 2 files changed, 38 insertions(+), 3 deletions(-)

2 8

[libvirt] building libvirtd fails on libxml-2
by Dan Bar Dov 23 Sep '09

23 Sep '09

I'm running on Centos5.3 Building using: ./autogen.sh --prefix=/usr configure fails with checking libxml2 xml2-config >= 2.6.0 ... configure: error: Could not find libxml2 anywhere (see config.log for details). but rpm -qa | grep libxml2 libxml2-python-2.6.26-2.1.2.7 libxml2-2.6.26-2.1.2.7 libxml2-2.6.26-2.1.2.7 So what's the problem, or better yet, how to fix it? Thanks, Dan

3 4

[libvirt] RFC: Rename / move / delete files in GIT
by Daniel P. Berrange 23 Sep '09

23 Sep '09

With the 0.7.1 relesae out of the way I'd like to suggest that we take this time to move around some files in GIT to correct some long standing wierd/bad naming decisions :-) The qemud/ directory is better named 'daemon', and some of the things in there should really have been in the src/ directory. So... * qemud/ -> daemon/ * qemud/qemud.{h,c} daemon/main.{h,c} * qemud/default-network.xml -> src/network/default.xml * qemud/libvirtd_qemu.aug src/qemu/qemu.aug * qemud/test_libvirtd_qemu.aug src/qemu/test_qemu.aug * qemud/remote_protocol.x -> src/remote/remote_protocol.x In the src/ directory we should move the rest of the drivers into their own sub-directories. Basically want one sub directory for each of the 'mod_LTLIBRARIES' declared in the src/Makefile.am. This will ensure we keep separate build dependancies, not accidentally including files that we shouldn't. * src/qemu_*.{c,h} -> src/qemu/ * src/xen_unified.{c,h} -> src/xen/xen_driver.{c,h} * src/xend_*, src/xm_* src/xen_* -> src/xen/ * src/test.{c,h} -> src/test/test_driver.{c,h} * src/storage*.{c,h} -> src/storage/ * src/security*.{c,h} -> src/security/ * src/remote_internal.{c,h} -> src/remote/remote_driver.{c,h} * src/interface_driver.c -> src/interface/netcf_driver.c * src/network_driver.c -> src/network/network_driver.c * src/lxc_* -> src/lxc/ * src/openvz_* -> src/openvz/ * src/node_device* -> src/nodedev/ That would just leave all the shared source files in src/. We could leave them there, or create a src/util/ directory for that stuff. Move virsh into the tools directory * src/virsh.c -> tools/virsh.c * docs/virsh.pod -> tools/virsh.pod * virsh.1: delete from GIT Sanitise naming in python/ directory, to make it clear which are the manually overriden methods. * python/libvir.c -> python/override-api.c * python/libvir.py -> python/override-api.py * python/libvirt-python-api.xml python/override-api.xml * python/libvirt_wrap.h -> python/types.h * python/virConnect.py -> python/override-virConnect.py Cleanup the docs/ directory * docs/*.html.in -> docs/website/ * docs/*.html: delete from GIT * docs/devhelp: delete from GIT * docs/html: delete from GIT * docs/libvirt-{api,refs}.xml: delete from GIT * docs/examples/ -> examples/ * docs/test*xml -> examples/xml/test/ * docs/storage -> examples/xml/storage/ Remove include/libvirt/libvirt.h from GIT For there places where I list 'delete from GIT', we would ensure that when you run 'make dist' the files are still included in the tar.gz NB, we would also update the cron job that deploys the website on libvirt.org soo that it runs 'make' in the docs/website directory to generate the .html files. Regards, Daniel -- |: Red Hat, Engineering, London -o- http://people.redhat.com/berrange/ :| |: http://libvirt.org -o- http://virt-manager.org -o- http://ovirt.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|

4 8

[libvirt] [PATCH] Allow control over QEMU audio backend
by Daniel P. Berrange 23 Sep '09

23 Sep '09

When using VNC for graphics + keyboard + mouse, we shouldn't then use the host OS for audio. Audio should go back over VNC. When using SDL for graphics, we should use the host OS for audio since that's where the display is. We need to allow certain QEMU env variables to be passed through to guest too to allow choice of QEMU audio backend. * qemud/libvirtd.sysconf: Mention QEMU/SDL audio env vars * src/qemu_conf.c: Passthrough QEMU/SDL audio env for SDL display, disable host audio for VNC display --- daemon/libvirtd.sysconf | 8 ++++++++ src/qemu/qemu_conf.c | 14 ++++++++++++++ 2 files changed, 22 insertions(+), 0 deletions(-) diff --git a/daemon/libvirtd.sysconf b/daemon/libvirtd.sysconf index fe4596a..28080a0 100644 --- a/daemon/libvirtd.sysconf +++ b/daemon/libvirtd.sysconf @@ -7,3 +7,11 @@ # Override Kerberos service keytab for SASL/GSSAPI #KRB5_KTNAME=/etc/libvirt/krb5.tab + +# Override the QEMU/SDL default audio driver probing when +# starting virtual machines using SDL graphics +# +# NB these have no effect for VMs using VNC +#QEMU_AUDIO_DRV=sdl +# +#SDL_AUDIODRIVER=pulse diff --git a/src/qemu/qemu_conf.c b/src/qemu/qemu_conf.c index 273aeca..74ca52b 100644 --- a/src/qemu/qemu_conf.c +++ b/src/qemu/qemu_conf.c @@ -2109,6 +2109,13 @@ int qemudBuildCommandLine(virConnectPtr conn, ADD_ARG_LIT("-k"); ADD_ARG_LIT(def->graphics[0]->data.vnc.keymap); } + + /* QEMU implements a VNC extension for providing audio, so we + * set the audio backend to none, to prevent it opening the + * host OS audio devices since that causes security issues + * and is non-sensical when using VNC. + */ + ADD_ENV_LIT("QEMU_AUDIO_DRV=none"); } else if ((def->ngraphics == 1) && def->graphics[0]->type == VIR_DOMAIN_GRAPHICS_TYPE_SDL) { char *xauth = NULL; @@ -2131,6 +2138,13 @@ int qemudBuildCommandLine(virConnectPtr conn, ADD_ENV(display); if (def->graphics[0]->data.sdl.fullscreen) ADD_ARG_LIT("-full-screen"); + + /* If using SDL for video, then we should just let it + * use QEMU's host audio drivers, possibly SDL too + * User can set these two before starting libvirtd + */ + ADD_ENV_COPY("QEMU_AUDIO_DRV"); + ADD_ENV_COPY("SDL_AUDIODRIVER"); } if (def->nvideos) { -- 1.6.2.5

2 2

[libvirt] ESX: Check if a datastore is accessible before accessing its URL property
by Matthias Bolte 23 Sep '09

23 Sep '09

An inaccessible datastore has no valid URL property. * src/esx/esx_vi.c: esxVI_LookupDatastoreByName(): check if datastore is accessible before accessing its URL property * src/esx/esx_vmx.c: update to changed datastore properties PS: This patch has to be applied after the 12er patch series. Matthias

3 2

[libvirt] ESX [01/12]: Add esxNodeGetFreeMemory()
by Matthias Bolte 23 Sep '09

23 Sep '09

* src/esx/esx_driver.c: add esxNodeGetFreeMemory(), cache IP address * src/esx/esx_vi.[ch]: refactor resource pool query into esxVI_GetResourcePool() * src/esx/esx_vi_types.[ch]: bind VI type ResourcePoolResourceUsage

3 2