- Devel - libvirt List Archives

Re: [libvirt] [Xen-devel] libvirt, libxl and QDISKs
by Jim Fehlig 29 Apr '13

29 Apr '13

David Scott wrote: >> Something like the attached, which seems to work well for me when >> specifying driverName = qemu, e.g. >> >> <disk type='file' device='disk'> >> <driver name='qemu'/> >> <source file='/var/lib/xen/images/sles11sp2-pv/disk0.raw'/> >> <target dev='xvda' bus='xen'/> >> </disk> > > This also works for me! Good to hear. I'll send the patch to the libvirt list. > > On a related note, what do you think about the attached patch? It > allows the user to select a non-default qemu via the <emulator> > element. My domain XML looked like this: > > <devices> > <emulator>/usr/lib/xen/bin/qemu-system-i386</emulator> IMO, the possible emulators should be exposed in the capabilities. E.g. on a machine with both kvm and qemu, both emulators are shown as possibilities for an hvm, x86_64 guest # virsh capabilities ... <guest> <os_type>hvm</os_type> <arch name='x86_64'> <wordsize>64</wordsize> <domain type='qemu'> <emulator>/usr/bin/qemu-system-x86_64</emulator> <machine>pc-1.1</machine> <machine canonical='pc-1.1'>pc</machine> <machine>pc-1.0</machine> <machine>pc-0.15</machine> <machine>pc-0.14</machine> <machine>pc-0.13</machine> <machine>pc-0.12</machine> <machine>pc-0.11</machine> <machine>pc-0.10</machine> <machine>isapc</machine> </domain> <domain type='kvm'> <emulator>/usr/bin/qemu-kvm</emulator> <machine>pc-1.1</machine> <machine canonical='pc-1.1'>pc</machine> <machine>pc-1.0</machine> <machine>pc-0.15</machine> <machine>pc-0.14</machine> <machine>pc-0.13</machine> <machine>pc-0.12</machine> <machine>pc-0.11</machine> <machine>pc-0.10</machine> <machine>isapc</machine> </domain> .... </guest> ... > <disk device="disk" type="network"> > <driver name='qemu'/> > <source protocol="rbd" name="rbd:rbd/ubuntu1204.img"/> > <target dev="hda"/> > </disk> > <graphics type="vnc" port="-1" autoport="yes" listen="0.0.0.0"/> > </devices> > > Now that upstream qemu is the default in xen-4.3, it seems useful to > be able to select the traditional qemu for older VMs. Agreed. And reporting that both emulators exist via the capabilities would be helpful for users. > > Also I backported this to my xen-4.2 system and used this patch + your > patch + the previous 'stat()' fix to successfully start a VM on ceph > storage via libvirt + libxl (my quest is almost complete!) Nice! > diff --git a/src/libxl/libxl_conf.c b/src/libxl/libxl_conf.c > index f549a5d..abbd3c0 100644 > --- a/src/libxl/libxl_conf.c > +++ b/src/libxl/libxl_conf.c > @@ -811,6 +811,30 @@ libxlMakeCapabilities(libxl_ctx *ctx) > } > > int > +libxlMakeEmulator(virDomainDefPtr def, libxl_domain_config *d_config) > +{ > + /* No explicit override means use the default */ > + if (!def->emulator) { > + return 0; > + } > + if (strstr(def->emulator, "/qemu-system-")) { > + d_config->b_info.device_model_version = > + LIBXL_DEVICE_MODEL_VERSION_QEMU_XEN; > + return 0; > + } > Here we could check the requested emulator against the capabilities, and then do the proper mapping for device_model_version. Do you have time for an upstream libvirt patch to expose the possible emulators in the capabilities, along with this patch allowing the user to specify one? Regards, Jim > + if (strstr(def->emulator, "/qemu-dm")) { > + d_config->b_info.device_model_version = > + LIBXL_DEVICE_MODEL_VERSION_QEMU_XEN_TRADITIONAL; > + return 0; > + } > + virReportError(VIR_ERR_INTERNAL_ERROR, > + _("libxenlight doesn't support emulator '%s'"), > + def->emulator); > + return -1; > +} > + > + > +int > libxlBuildDomainConfig(libxlDriverPrivatePtr driver, > virDomainDefPtr def, libxl_domain_config *d_config) > { > @@ -834,6 +858,10 @@ libxlBuildDomainConfig(libxlDriverPrivatePtr driver, > goto error; > } > > + if (libxlMakeEmulator(def, d_config) < 0) { > + goto error; > + } > + > d_config->on_reboot = def->onReboot; > d_config->on_poweroff = def->onPoweroff; > d_config->on_crash = def->onCrash; > > ------------------------------------------------------------------------

2 1

[libvirt] [PATCH 0/3] PCI mess fixes
by Ján Tomko 27 Apr '13

27 Apr '13

1/3 should allow domains without a PCI bus to start 2/3 prevents invalid memory access when the domain has no PCI bus 3/3 fixes a mising error message on OOM Ján Tomko (3): qemu: don't always reserve PCI addresses for implicit controllers qemu: prevent invalid reads in qemuAssignDevicePCISlots conf: add missing error on OOM src/conf/domain_conf.c | 1 + src/qemu/qemu_command.c | 42 +++++++++++++++++++++++++----------------- 2 files changed, 26 insertions(+), 17 deletions(-) -- 1.8.1.5

2 8

Re: [libvirt] KVM Forum 2013 Save the Date
by tiejun.chen 27 Apr '13

27 Apr '13

On 04/27/2013 02:25 PM, Will Huck wrote: > Hi, > > Where can get slides in 2012 KVM Forum? http://www.linux-kvm.org/page/KVM_Forum_2012 Tiejun

1 0

[libvirt] [PATCH 00/10 v3] Support scsi device passthrough via scsi-generic
by Osier Yang 27 Apr '13

27 Apr '13

This adds the support for qemu's scsi-generic device, which can be used to passthrough scsi host device. An example of the XML: <hostdev mode='subsystem' type='scsi' managed='no'> <source> <adapter name='scsi_host7'/> <address bus='0' target='0' unit='0'/> </source> <address type='drive' controller='0' bus='0' target='0' unit='1'/> </hostdev> Mainly changes from v2.5 (details in each patch): * 1/10 and 2/10 of v2.5 are pushed * 8/10 of v2.5 is splitted * A better way to generate the address for scsi host device instead of the rigid one. * Support all scsi controllers when building the qemu command line. * Figure out the limits of the qemu properties, and do checking * Don't forget DAC and apparmor security backends * Orgnize the patches in a better way. * Fix the make check failure Han Cheng (7): conf: Introduce XMLs for scsi hostdev qemu: New cap flags for scsi-generic utils: util functions for scsi hostdev qemu: Build qemu command line for scsi host device qemu: Introduce activeScsiHostdevs list for scsi host devices qemu: Allow the scsi-generic device in cgroup qemu: Add hotplug support for scsi host device Osier Yang (3): security: Manage the security label for scsi host device qemu: Refactor helpers for USB device attachment conf: Generate address for scsi host device automatically docs/formatdomain.html.in | 38 +- docs/schemas/domaincommon.rng | 29 ++ po/POTFILES.in | 1 + src/Makefile.am | 1 + src/conf/domain_audit.c | 10 + src/conf/domain_conf.c | 312 +++++++++++++++- src/conf/domain_conf.h | 8 + src/libvirt_private.syms | 22 ++ src/qemu/qemu_capabilities.c | 14 +- src/qemu/qemu_capabilities.h | 2 + src/qemu/qemu_cgroup.c | 72 +++- src/qemu/qemu_cgroup.h | 3 + src/qemu/qemu_command.c | 154 +++++++- src/qemu/qemu_command.h | 6 + src/qemu/qemu_conf.h | 2 + src/qemu/qemu_driver.c | 3 + src/qemu/qemu_hostdev.c | 213 +++++++++++ src/qemu/qemu_hostdev.h | 10 + src/qemu/qemu_hotplug.c | 224 +++++++++--- src/qemu/qemu_process.c | 3 + src/security/security_apparmor.c | 49 ++- src/security/security_dac.c | 77 +++- src/security/security_selinux.c | 72 +++- src/util/virscsi.c | 407 +++++++++++++++++++++ src/util/virscsi.h | 84 +++++ tests/qemuhelpdata/qemu-1.0-device | 10 + tests/qemuhelpdata/qemu-1.1.0-device | 10 + tests/qemuhelpdata/qemu-1.2.0-device | 5 + tests/qemuhelpdata/qemu-kvm-1.2.0-device | 5 + tests/qemuhelptest.c | 19 +- .../qemuxml2argv-hostdev-scsi-boot.args | 9 + .../qemuxml2argv-hostdev-scsi-boot.xml | 34 ++ .../qemuxml2argv-hostdev-scsi-readonly.args | 9 + .../qemuxml2argv-hostdev-scsi-readonly.xml | 35 ++ tests/qemuxml2argvtest.c | 9 + tests/qemuxml2xmltest.c | 2 + 36 files changed, 1842 insertions(+), 121 deletions(-) create mode 100644 src/util/virscsi.c create mode 100644 src/util/virscsi.h create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-hostdev-scsi-boot.args create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-hostdev-scsi-boot.xml create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-hostdev-scsi-readonly.args create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-hostdev-scsi-readonly.xml -- 1.8.1.4

2 15

[libvirt] [PATCH 0/2] VFIO bugfix and naming tweak
by Laine Stump 27 Apr '13

27 Apr '13

Laine Stump (2): network: support <driver name='vfio'/> in network definitions conf: remove extraneous _TYPE from driver backend enums src/conf/domain_conf.c | 6 +++--- src/conf/domain_conf.h | 6 +++--- src/conf/network_conf.c | 39 +++++++++++++++++++++++++++++++++- src/conf/network_conf.h | 17 ++++++++++++++- src/network/bridge_driver.c | 23 ++++++++++++++++++++ src/qemu/qemu_command.c | 8 +++---- src/qemu/qemu_hostdev.c | 4 ++-- src/qemu/qemu_hotplug.c | 4 ++-- src/security/security_apparmor.c | 2 +- src/security/security_dac.c | 4 ++-- src/security/security_selinux.c | 4 ++-- tests/networkxml2xmlin/hostdev-pf.xml | 1 + tests/networkxml2xmlout/hostdev-pf.xml | 1 + 13 files changed, 98 insertions(+), 21 deletions(-) -- 1.7.11.7

2 4

[libvirt] [PATCH v2 0/2] qemu: invoke qemu-bridge-helper from libvirtd
by Paolo Bonzini 26 Apr '13

26 Apr '13

The <interface type='bridge'> is working mostly because of a peculiar design decision in Linux. Ideally, QEMU would run with an empty capability bounding set and would not be able to do any privileged operation (not even by running a helper program). This is not the case because dropping capabilities from the bounding set requires a capability of its own, CAP_SETPCAP; thus QEMU does *not* run with an empty bounding set if invoked via qemu:///session. This is apparently for security reasons, to avoid that dropping _some_ caps but not all of them lets you exploit untested error paths in suid binaries. This series lets libvirtd invoke the privileged helper program on its own, which is a cleaner design that would work even if the above Linux quirk was not there. Also, this adds a <target dev='tap0'/> element to the XML of an active domain using <interface type='bridge'>. Thanks to the patches that have already been committed, the recvfd and virCommand APIs make the task almost trivial. v1->v2: OOM fix in patch 1, change label name in patch 2, rebase Paolo Bonzini (2): virnetdevtap: add virNetDevTapGetName qemu: launch bridge helper from libvirtd src/libvirt_private.syms | 1 + src/qemu/qemu_command.c | 133 ++++++++++++++++++++++++++++++++++------------- src/qemu/qemu_command.h | 1 - src/qemu/qemu_hotplug.c | 25 +++------ src/util/virnetdevtap.c | 33 ++++++++++++ src/util/virnetdevtap.h | 3 ++ 6 files changed, 143 insertions(+), 53 deletions(-) -- 1.8.2

3 6

[libvirt] [PATCH-v4 0/2] Static Route related updates
by Gene Czarcinski 26 Apr '13

26 Apr '13

This update includes two patch files: 1. The first file adds virSocketAddrGetIpPrefix() to determine the prefix for a network. This function is used by the static route code and has also been used to update (replace the code in) virNetworkIpDefPrefix() in src/conf/network_conf.c 2. The second and major update adds functionality to implement static route for both IPv4 and IPv6. Having static route provides the routing information needed to forward packets not directly reachable from a host. As far as I can determine, this update has adopted all suggestions made against the previous version (except that I am still using CommandRun ... that will be addressed in a future update). I also plan to take a look at <ip> processing such as handling ULong return codes. I also raise the question as to whether address= should be manditory rather than (currently) optional ... I am not sure what <ip> means with address= being optional. Gene Czarcinski (2): create virSocketAddrGetIpPrefix utility function Support for static routes on a virtual bridge docs/formatnetwork.html.in | 80 +++++ docs/schemas/network.rng | 22 ++ src/conf/network_conf.c | 364 +++++++++++++++++++-- src/conf/network_conf.h | 20 ++ src/libvirt_private.syms | 2 + src/network/bridge_driver.c | 41 +++ src/util/virnetdev.c | 44 +++ src/util/virnetdev.h | 5 + src/util/virsocketaddr.c | 43 +++ src/util/virsocketaddr.h | 3 + .../networkxml2xmlin/dhcp6host-routed-network.xml | 2 + .../networkxml2xmlout/dhcp6host-routed-network.xml | 2 + 12 files changed, 600 insertions(+), 28 deletions(-) -- 1.8.1.4

3 11

[libvirt] [PATCH] qemu: don't assign a PCI address to 'none' USB controller
by Ján Tomko 26 Apr '13

26 Apr '13

--- src/qemu/qemu_command.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/src/qemu/qemu_command.c b/src/qemu/qemu_command.c index 9737609..126cced 100644 --- a/src/qemu/qemu_command.c +++ b/src/qemu/qemu_command.c @@ -1925,6 +1925,11 @@ qemuAssignDevicePCISlots(virDomainDefPtr def, if (def->controllers[i]->type == VIR_DOMAIN_CONTROLLER_TYPE_PCI) continue; + /* USB controller model 'none' doesn't need a PCI address */ + if (def->controllers[i]->type == VIR_DOMAIN_CONTROLLER_TYPE_USB && + def->controllers[i]->model == VIR_DOMAIN_CONTROLLER_MODEL_USB_NONE) + continue; + /* FDC lives behind the ISA bridge; CCID is a usb device */ if (def->controllers[i]->type == VIR_DOMAIN_CONTROLLER_TYPE_FDC || def->controllers[i]->type == VIR_DOMAIN_CONTROLLER_TYPE_CCID) -- 1.8.1.5

3 5

[libvirt] libvirtd segfault
by AL13N 26 Apr '13

26 Apr '13

regarding http://www.redhat.com/archives/libvir-list/2012-December/msg01365.html and https://www.redhat.com/archives/libvir-list/2012-December/msg00935.html This seems like the problem i'm having... is there a permanent solution for this? (i'm using libvirtd with XEN using libxl) almost every time i shutdown a domain libvirtd segfaults...

4 13

[libvirt] [PATCH] fix segfault during virsh save in pv guest
by Bamvor Jian Zhang 26 Apr '13

26 Apr '13

this patch fix the wrong sequence for fd and timeout register. the sequence was right in dfa1e1dd for fd register, but it changed in e0622ca2. in this patch, set priv, xl_priv in info and increase info->priv ref count before virEventAddHandle. if do this after virEventAddHandle, the fd callback or fd deregister maybe got the empty priv, xl_priv or wrong ref count. after apply this patch, test more than 100 rounds passed compare to fail within 3 rounds without this patch. each round includes define -> start -> destroy -> create -> suspend -> resume -> reboot -> shutdown -> save -> resotre -> dump -> destroy -> create -> setmem -> setvcpus -> destroy. Signed-off-by: Bamvor Jian Zhang <bjzhang(a)suse.com> --- src/libxl/libxl_driver.c | 39 +++++++++++++++++++++------------------ 1 files changed, 21 insertions(+), 18 deletions(-) diff --git a/src/libxl/libxl_driver.c b/src/libxl/libxl_driver.c index b4f1889..212d0fc 100644 --- a/src/libxl/libxl_driver.c +++ b/src/libxl/libxl_driver.c @@ -181,26 +181,28 @@ libxlFDRegisterEventHook(void *priv, int fd, void **hndp, return -1; } + info->priv = priv; + /* + * Take a reference on the domain object. Reference is dropped in + * libxlEventHookInfoFree, ensuring the domain object outlives the fd + * event objects. + */ + virObjectRef(info->priv); + info->xl_priv = xl_priv; + if (events & POLLIN) vir_events |= VIR_EVENT_HANDLE_READABLE; if (events & POLLOUT) vir_events |= VIR_EVENT_HANDLE_WRITABLE; + info->id = virEventAddHandle(fd, vir_events, libxlFDEventCallback, info, libxlEventHookInfoFree); if (info->id < 0) { + virObjectUnref(info->priv); VIR_FREE(info); return -1; } - info->priv = priv; - /* - * Take a reference on the domain object. Reference is dropped in - * libxlEventHookInfoFree, ensuring the domain object outlives the fd - * event objects. - */ - virObjectRef(info->priv); - - info->xl_priv = xl_priv; *hndp = info; return 0; @@ -283,6 +285,15 @@ libxlTimeoutRegisterEventHook(void *priv, return -1; } + info->priv = priv; + /* + * Also take a reference on the domain object. Reference is dropped in + * libxlEventHookInfoFree, ensuring the domain object outlives the timeout + * event objects. + */ + virObjectRef(info->priv); + info->xl_priv = xl_priv; + gettimeofday(&now, NULL); timersub(&abs_t, &now, &res); /* Ensure timeout is not overflowed */ @@ -296,22 +307,14 @@ libxlTimeoutRegisterEventHook(void *priv, info->id = virEventAddTimeout(timeout, libxlTimerCallback, info, libxlEventHookInfoFree); if (info->id < 0) { + virObjectUnref(info->priv); VIR_FREE(info); return -1; } - info->priv = priv; - /* - * Also take a reference on the domain object. Reference is dropped in - * libxlEventHookInfoFree, ensuring the domain object outlives the timeout - * event objects. - */ - virObjectRef(info->priv); - virObjectLock(info->priv); LIBXL_EV_REG_APPEND(info->priv->timerRegistrations, info); virObjectUnlock(info->priv); - info->xl_priv = xl_priv; *hndp = info; return 0; -- 1.6.0.2

2 1