July 2009 - Devel - libvirt List Archives

[libvirt] [PATCH] Extra filesystem options during backend filesystem mount.
by Harshavardhana 22 Jul '09

22 Jul '09

New option index added to support -o options for various netfs. Currently added an option for glusterfs. --- src/storage_backend_fs.c | 32 ++++++++++++++++++++++++++++++++ 1 files changed, 32 insertions(+), 0 deletions(-) diff --git a/src/storage_backend_fs.c b/src/storage_backend_fs.c index c3d66b5..606fb47 100644 --- a/src/storage_backend_fs.c +++ b/src/storage_backend_fs.c @@ -604,6 +604,7 @@ static int virStorageBackendFileSystemMount(virConnectPtr conn, virStoragePoolObjPtr pool) { char *src; + char *options; const char **mntargv; /* 'mount -t auto' doesn't seem to auto determine nfs (or cifs), @@ -611,6 +612,10 @@ virStorageBackendFileSystemMount(virConnectPtr conn, * accommodate this */ int netauto = (pool->def->type == VIR_STORAGE_POOL_NETFS && pool->def->source.format == VIR_STORAGE_POOL_NETFS_AUTO); + int glusterfs = (pool->def->type == VIR_STORAGE_POOL_NETFS && + pool->def->source.format == VIR_STORAGE_POOL_NETFS_GLUSTERFS); + + int option_index; int source_index; const char *netfs_auto_argv[] = { @@ -632,9 +637,26 @@ virStorageBackendFileSystemMount(virConnectPtr conn, NULL, }; + const char *glusterfs_argv[] = { + MOUNT, + "-t", + pool->def->type == VIR_STORAGE_POOL_FS ? + virStoragePoolFormatFileSystemTypeToString(pool->def->source.format) : + virStoragePoolFormatFileSystemNetTypeToString(pool->def->source.format), + NULL, + "-o", + NULL, + pool->def->target.path, + NULL, + }; + if (netauto) { mntargv = netfs_auto_argv; source_index = 1; + } else if (glusterfs) { + mntargv = glusterfs_argv; + source_index = 3; + option_index = 5; } else { mntargv = fs_argv; source_index = 3; @@ -670,6 +692,12 @@ virStorageBackendFileSystemMount(virConnectPtr conn, } if (pool->def->type == VIR_STORAGE_POOL_NETFS) { + if (pool->def->source.format = VIR_STORAGE_POOL_NETFS_GLUSTERFS) { + if (virAsprintf(&options, "direct-io-mode=1") == -1) { + virReportOOMError(conn); + return -1; + } + } if (virAsprintf(&src, "%s:%s", pool->def->source.host.name, pool->def->source.dir) == -1) { @@ -685,6 +713,10 @@ virStorageBackendFileSystemMount(virConnectPtr conn, } mntargv[source_index] = src; + if (glusterfs) { + mntargv[option_index] = options; + } + if (virRun(conn, mntargv, NULL) < 0) { VIR_FREE(src); return -1; -- 1.6.0.6

3 6

[libvirt] Workaround to start xen inside a buggy kvm
by Thomas Fricke 21 Jul '09

21 Jul '09

Buggy kvm crashes with kernel panic if xen started inside a kvm machine. kvm needs the -no-kvm-pit, this argument is not available with libvirt. I have tested the following workaround on ubuntu 9.04 Rename kvm cd /usr/bin mv kvm kvm.orig Use the following script instead the original kvm #!/bin/sh exec kvm.orig -no-kvm-pit $* Do not forget chmod +x kvm Have fun, Thomas

1 0

[libvirt] [PATCH 0/6] virInterface bug fixes, test driver, and netcf backend driver
by Laine Stump 21 Jul '09

21 Jul '09

This series of patches first corrects a couple of bugs/deficiencies in existing virInterface code, then adds test and netcf-based backend interface drivers.

3 25

[libvirt] [patch] initialize a local variable in qemudOpenMonitorUnix()
by Jun Koi 21 Jul '09

21 Jul '09

This patch initializes a local variable in qemudOpenMonitorUnix(), thus also eliminates a compilation warning. diff --git a/src/qemu_driver.c b/src/qemu_driver.c index 00dc6e5..d2db1a2 100644 --- a/src/qemu_driver.c +++ b/src/qemu_driver.c @@ -874,7 +874,7 @@ qemudOpenMonitorUnix(virConnectPtr conn, struct sockaddr_un addr; int monfd; int timeout = 3; /* In seconds */ - int ret, i; + int ret, i = 0; if ((monfd = socket(AF_UNIX, SOCK_STREAM, 0)) < 0) { virReportSystemError(conn, errno,

2 1

[libvirt] [PATCH] rpm spec cleanup and split off client only requirements
by Daniel Veillard 21 Jul '09

21 Jul '09

This early patch changes the spec file to: - remove the old Obsoletes: livir and such, that was 3 years ago - make sure the inter packages require %{version}-%{release} and not just %{version} - compresses the ChangeLog which is getting close to a megabyte - split out a separate libvirt-client binary rpm including only what is needed to access the libvirt service from a client viewpioint. The main package requires the client anyway So I moved in the client: - the shared library, I guess it's uncontroversial - the virsh/virt-xml-validate binaries, I guess it it doesn't block i386/x86_64 to be parallel installable that's fine - the man pages, obvious based on previous - the proxy when built, more controversial, it assumes that there is a service running locally so could be left on the main package - sasl and policykit data, more controversial, I guess it's better to associate them with the client - the schemas to allow checking on the client Some of this is still open, including the name of the subpackage which I initially expected to be libvirt-libs but in retrospect maybe -client will be more in line with the content, Comments welcome :-) Daniel -- Daniel Veillard | libxml Gnome XML XSLT toolkit http://xmlsoft.org/ daniel(a)veillard.com | Rpmfind RPM search engine http://rpmfind.net/ http://veillard.com/ | virtualization library http://libvirt.org/

2 2

[libvirt] [PATCH 0/7] Support cgroups in QEMU driver
by Daniel P. Berrange 20 Jul '09

20 Jul '09

This series of patches introduces cgroups support to the QEMU driver. At this time it uses the 'devices' controller to whitelist block devices for QEMU guests to prevent unauthorized access. It uses the 'cpu_shares' controller to allow schedular tunables on a per guest basis. In the future we should use the 'memory' controller to enforce the limit set in the balloon driver ie if the guest does not honour the balloon request, then force the guest into swap, avoiding host overcommit. There is also some refactoring of the cgroups code to remove the assumption that libvirtd is starting in the root cgroup, remove the requirement that all controllers be active, and allow for use in non-privileged drivers. Daniel P. Berrange (7): Use enums for cgroup controller types / labels Use virFileReadAll/virFileWriteStr for key cgroup read/write helpers Make cgroups a little more efficient Place every QEMU guest in a private cgroup Implement schedular tunables API using cgroups Use cgroups for block device whitelisting in QEMU guests Refactor cgroups to allow a group per driver to be managed directly src/cgroup.c | 861 ++++++++++++++++++++++++++-------------------- src/cgroup.h | 28 +- src/libvirt_private.syms | 1 + src/lxc_conf.h | 2 + src/lxc_controller.c | 19 +- src/lxc_driver.c | 25 +- src/qemu_conf.h | 2 + src/qemu_driver.c | 355 ++++++++++++++++++- src/util.c | 27 ++- src/util.h | 2 + 10 files changed, 904 insertions(+), 420 deletions(-)

2 17

[libvirt] changing files in /var/lib/libvirt/images
by Gene Czarcinski 20 Jul '09

20 Jul '09

Assume I am working with virtual disk image files in /var/lib/libvirt/images such as removing a file, defining a new file (with qemi-img create or qemu-img convert), or simply copying a file ("cp") from elsewhere such as VMware. When I do this, Guest definition or adding hardware to an existing guest under virt- manager does not "see" the file change. That is, until I terminate virt- manager and restart /etc/init.d/libvirtd. Is there some better/easier/less-disruptive means of notifying "whatever" that the list of files has changed? I noticed that /etc/init.d/libvirtd force-reload has no effect. I also notice that restarting libvirtd and/or terminating and restarting virt- manager while there are guests running really screws things up. If such a "means" does not exist, should one be created? [should this be an RFE in bugzilla?] Gene

2 2

[libvirt] Create a new VM from an existent image?
by Jun Koi 19 Jul '09

19 Jul '09

Hi, To use libvirt, I am strying to use virt-install to create a new VM from an existent (KVM) VM image. I did the following steps: 1) run libvirtd with "libvirtd -d" (I use the latest libvirt git tree) 2) Run virt-install (0.400.3) with my existent image in img/img.winxp # virt-install --import -n winxp -f img/img.xp3 --vnc --hvm --accelerate --ram 800 However, this step returns error like: "ERROR Could not find usable default libvirt connection." Actually I am not sure that is the right way to use the new --import option with virt-install. So that might be the problem? All the code are updated, and compiled from source on Ubuntu 8.04. Any help? Thanks a lot, J

2 4

[libvirt] [PATCH] qemu: Try multiple times to open unix monitor socket
by Cole Robinson 18 Jul '09

18 Jul '09

Unlike the pty monitor (which we know exists since we scrape its path from stdout), we have no way of knowing that the unix monitor socket should exist/ be initialized. As a result, some of my KVM guests randomly fail to start on F10 host. Try to open the unix socket in a 3 second timeout loop. Ignore EACCES (path does not exist if a first time run) and ECONNREFUSED (leftover socket from a previous run hasn't been removed yet). Fixes things for me. --- src/qemu_driver.c | 22 +++++++++++++++++++++- 1 files changed, 21 insertions(+), 1 deletions(-) diff --git a/src/qemu_driver.c b/src/qemu_driver.c index e2b7acb..200718b 100644 --- a/src/qemu_driver.c +++ b/src/qemu_driver.c @@ -874,6 +874,8 @@ qemudOpenMonitorUnix(virConnectPtr conn, { struct sockaddr_un addr; int monfd; + int timeout = 3; /* In seconds */ + int ret, i; if ((monfd = socket(AF_UNIX, SOCK_STREAM, 0)) < 0) { virReportSystemError(conn, errno, @@ -885,10 +887,28 @@ qemudOpenMonitorUnix(virConnectPtr conn, addr.sun_family = AF_UNIX; strncpy(addr.sun_path, monitor, sizeof(addr.sun_path)); - if (connect(monfd, (struct sockaddr *) &addr, sizeof(addr)) < 0) { + do { + ret = connect(monfd, (struct sockaddr *) &addr, sizeof(addr)); + + if (ret == 0) + break; + + if (errno == EACCES || errno == ECONNREFUSED) { + /* EACCES : Socket may not have shown up yet + * ECONNREFUSED : Leftover socket hasn't been removed yet */ + continue; + } + virReportSystemError(conn, errno, "%s", _("failed to connect to monitor socket")); goto error; + + } while ((++i <= timeout*5) && (usleep(.2 * 1000000) <= 0)); + + if (ret != 0) { + virReportSystemError(conn, errno, "%s", + _("monitor socket did not show up.")); + goto error; } if (qemudOpenMonitorCommon(conn, driver, vm, monfd, reconnect) < 0) -- 1.6.0.6

4 5

[libvirt] adding scsi disk failed (code=9)
by Massimo Canonico 18 Jul '09

18 Jul '09

Hi, I'm newbie and sorry if this question was already post in this ML (is there any search engine for the ML archive?). I'm trying to attach a volume in my instance. The instance is running and the volume is available, but i got this error: ---- libvirt: operation failed: adding scsi disk failed (code=9) virDomainAttachDevice() failed (err=-1) XML=<disk type='block'><driver name='phy'/><source dev='/dev/etherd/e0.8'/><target dev='sdb'/></disk> doAttachVolume() failed error=1 ---- I'm not sure where I can start to debug this error. I'm using Fedora 11, x86_64, with these packages [root@minicloud eucalyptus]# rpm -qa | grep libv libvirt-devel-0.6.2-12.fc11.x86_64 libvorbis-1.2.0-7.fc11.x86_64 libvirt-0.6.2-12.fc11.x86_64 libvirt-python-0.6.2-12.fc11.x86_64 libvolume_id-141-3.fc11.x86_64 May you help me, please? Thanks in advance, M

4 5