- Devel - libvirt List Archives

[libvirt] [PATCH 0/3] Add callback mech for qemu and nodedev
by John Ferlan 09 Feb '17

09 Feb '17

Add a callback mechanism for the nodedev driver (mimics the nwfilter) to allow the (c)ability to directly enumerate node devices and add/remove devices based on udev events. This is a set up of sorts for the ability to add a vHBA definition to a domain, create the vHBA, and allow the LUN's associated with the vHBA to be automagially added/removed to/from the domain. The vHBA code exists partially in some local branches I have, but I figure I need to get this part of the logic accepted. See bz: https://bugzilla.redhat.com/show_bug.cgi?id=1404964 for some "thoughts" on where this is headed. An alternative solution would require the qemu driver to handle or recognize nodedev connection events, but would require a few more calls in order to get the data needed for vHBA handling. The logic ends up being "hidden" behind a new qemu configuration variable "enumerate_nodedev" (I'm fine using a different name). The default is for the code to be disabled. The end result may be usable by the (on list) vGPU mdev code as well as a way to find/recogize an mdev create and the various "children" of the mdev that would be created (and of course deletion). But that's just me considering other uses. John Ferlan (3): nodedev: Add driver callback mechanism to add/remove devices qemu: Use nodedev callback mechanism to to get a nodedev data qemu: Add configuration variable to control nodedev enumeration src/check-aclrules.pl | 3 +- src/check-driverimpls.pl | 1 + src/conf/node_device_conf.c | 125 ++++++++++++++++++++++++++++++++++++- src/conf/node_device_conf.h | 18 ++++++ src/libvirt_private.syms | 3 + src/node_device/node_device_udev.c | 29 +++++++++ src/qemu/libvirtd_qemu.aug | 1 + src/qemu/qemu.conf | 19 ++++++ src/qemu/qemu_conf.c | 53 ++++++++++++++++ src/qemu/qemu_conf.h | 17 +++++ src/qemu/qemu_driver.c | 34 ++++++++++ src/qemu/test_libvirtd_qemu.aug.in | 1 + 12 files changed, 302 insertions(+), 2 deletions(-) -- 2.7.4

2 5

[libvirt] [PATCH 1/1] qemu: libvirt live migration over RDMA of ipv6 addr failed
by David Dai 09 Feb '17

09 Feb '17

Using libvirt to do live migration over RDMA via ip v6 address failed. For example: # virsh migrate --live --migrateuri rdma://[deba::2222]:49152 \ rhel73_host1_guest1 qemu+ssh://[deba::2222]/system --verbose root@deba::2222's password: error: internal error: unable to execute QEMU command 'migrate': RDMA ERROR: could not rdma_getaddrinfo address deba As we can see, the ip v6 address used by rdma_getaddrinfo() has only "deba" part. It should be "deba::2222". 1) According to rfc 3986, a literal ip v6 address should be enclosed in '[' and ']'. When using virsh command to do live migration via ip v6 addresss, user will input the ip v6 address with brackets (i.e. rdma://[deba::2222]:49152). libvirt will parse command line option by calling virURIParse(). Inside it calls virStringStripIPv6Brackets() to strip off the brackets. The uri passed in to virURIParse() is: "uri = rdma://[deba::2222]:49152" Inside virURIParse() routine, it will strip off the bracket '[' and ']' if it's ip v6 address. Then save the ip v6 address in this format "deba::2222" in the virURI->server field, and to be passed to qemu. 2) At the beginning of migration, in qemu's qemu_rdma_data_init(host_port) routine, it calls inet_parse(host_port) routine to parse the ip v6 address and port string obtained from libvirt. The input string host_port passed to qemu_rdma_data_init() can be: "hostname:port", or "ipv4address:port", or "[ipv6address]:port" (i.e "[deba::2222]:49152"), or "ipv6address:port" (i.e "deba::2222:49152"). Existing qemu api inet_parse() can handle the above first 3 cases properly, but didn't handle the last case ("ipv6address:port") correctly. In this live migration over rdma via ip v6 address case, the server ip v6 address obtained from libvirt doesn't contain the brackets '[' and ']' (i.e. "deba::2222:49152"). It caused inet_parse() to parse only "deba" part, and stopped at the 1st colon ':'. As the result, the subsequent rdma_getaddrinfo() with ip address "deba" will fail. NOTE: If using libvirt to do live migration over TCP via ip v6 address: # virsh migrate --live --migrateuri tcp://[deba::2222]:49152 \ rhel73_host1_guest1 qemu+ssh://[deba::2222]/system --verbose It works fine. In migrateuri of tcp case, libvirt will call virNetSocketNewConnectTCP() directly to connect to remote "deba::2222" after it strips off the brackets '[' and ']' for an ip v6 address. On qemu side, fd_start_outgoing_migration() will be called to do migration. It doesn't call inet_parse(). So we don't see issue in tcp case. Solution: Originally the patch was proposed to Qemu-devel commuity to changed code in inet_parse() to handle ipv6 address w/o brackets ("ipv6:port" format). Daniel's review comment is it's mandatory to use [] when providing a numeric IPv6 address. So there's nothing broken in QEMU. libvirt needs fixing to pass correct data to QEMU. Now the new proposed patch is in libvirt's qemu driver, in qemuMonitorMigrateToHost() routine, it will assemble uri based on host and port values passed in. If the host passed in is an ipv6 address, we will enclose brackets [] for the ipv6 address here before calling qemuMonitorJSONMigrate(). The -incoming CLI part of enclosed brackets ipv6 address is already taken care of in qemuMigrationPrepareIncoming() routine. Signed-off-by: David Dai <zdai(a)linux.vnet.ibm.com> --- src/qemu/qemu_monitor.c | 8 ++++++-- 1 files changed, 6 insertions(+), 2 deletions(-) diff --git a/src/qemu/qemu_monitor.c b/src/qemu/qemu_monitor.c index b7be5e7..258665c 100644 --- a/src/qemu/qemu_monitor.c +++ b/src/qemu/qemu_monitor.c @@ -2577,8 +2577,12 @@ qemuMonitorMigrateToHost(qemuMonitorPtr mon, QEMU_CHECK_MONITOR(mon); - if (virAsprintf(&uri, "%s:%s:%d", protocol, hostname, port) < 0) - return -1; + if (strchr(hostname, ':')) { + if (virAsprintf(&uri, "%s:[%s]:%d", protocol, hostname, port) < 0) + return -1; + } else if (virAsprintf(&uri, "%s:%s:%d", protocol, hostname, port) < 0) { + return -1; + } if (mon->json) ret = qemuMonitorJSONMigrate(mon, flags, uri); -- 1.7.1

2 1

[libvirt] [PATCH V2 0/2] libxl: a few domain maximum memory fixes
by Jim Fehlig 09 Feb '17

09 Feb '17

Patch 1 fixes reporting of domain maximum memory for running domains. When creating a virDomainDef object to represent dom0, max memory was not set correctly, which is fixed by patch2. V2: Check return value of libxl_get_physinfo and libxlDriverGetDom0MaxmemConf in patch2. Jim Fehlig (2): libxl: fix reporting of maximum memory libxl: fix dom0 maximum memory setting src/libxl/libxl_conf.c | 77 ++++++++++++++++++++++++++++++++++++++++++++++++ src/libxl/libxl_conf.h | 4 +++ src/libxl/libxl_driver.c | 10 ++++--- 3 files changed, 87 insertions(+), 4 deletions(-) -- 2.9.2

2 8

[libvirt] [PATCH V2 0/2] libxl: fix handling of <disk> driver and format settings
by Jim Fehlig 09 Feb '17

09 Feb '17

V2 of https://www.redhat.com/archives/libvir-list/2017-February/msg00185.html Patch1 is new in V2 and essentially moves default setting of disk format from domain build time to device post-parse, which addresses Michal's comment from V1. Patch2 is mostly unchanged. The hunk setting format of virDomainDiskDef in libxlMakeDisk is removed since that is now handled by patch1. Jim Fehlig (2): libxl: set default disk format in device post-parse libxl: fix disk detach when <driver> not specified src/libxl/libxl_conf.c | 42 ++++++++++++++++++++++++++++++++++-------- src/libxl/libxl_conf.h | 4 ++++ src/libxl/libxl_domain.c | 32 +++++++++++++++++++++++++++++++- src/libxl/libxl_driver.c | 1 + 4 files changed, 70 insertions(+), 9 deletions(-) -- 2.9.2

2 5

[libvirt] [PATCH v3 0/3] network: make openvswitch call timeout configurable
by Boris Fiuczynski 09 Feb '17

09 Feb '17

Since a successful completion of openvswitch calls is expected a longer timeout should be able to be chosen in order to account for loaded systems. Therefore this patch series provides the ability to specify the timeout value for openvswitch calls in the libvirtd configuration file. Boris Fiuczynski (3): libvirtd: add openvitch timeout value network: allow to specify timeout for openvswitch calls libvirtd: set openvswitch timeout value based on config data daemon/libvirtd-config.c | 6 ++++ daemon/libvirtd-config.h | 2 ++ daemon/libvirtd.aug | 1 + daemon/libvirtd.c | 13 +++++++++ daemon/libvirtd.conf | 9 ++++++ daemon/test_libvirtd.aug.in | 1 + src/libvirt_private.syms | 1 + src/util/virnetdevopenvswitch.c | 64 +++++++++++++++++++++++++++++++++++------ src/util/virnetdevopenvswitch.h | 5 ++++ 9 files changed, 94 insertions(+), 8 deletions(-) -- 2.5.5

2 12

[libvirt] question about connection between vms in two physical hosts
by 无敌浪子 09 Feb '17

09 Feb '17

Hi there, I have some vms in two different physical hosts, the hosts are connected directly instead of a switcher, I'd like to make all these vms connected. the condition right now is that vm network type is nat, vm behind nat cannot get the new connection inbound, I suppose if I can make the inbound connection acceptable or if there is any other solution ? Thank you best wishes bill

2 1

[libvirt] [PATCHv5 0/4] Add support for memory backing with file source
by Jaroslav Safka 09 Feb '17

09 Feb '17

Hi, we would like to introduce 3 new elements source,access and allocation in memoryBacking element. For now it was made for numa topology. <memoryBacking> <source type="file|anonymous"/> <access mode="shared|private"/> <allocation mode="immediate|ondemand"/> </memoryBacking> If allocation is immediate then -mem-prealloc should be added to the qemu commanline. If source is file then -object memory-backend-file,id=mem,size=1024M,mem-path=*lib dir path* -numa node,memdev=mem Will be added to the qemu commandline If access is shared then the "share=on" parameter will be added to the memory-backend-file e.g. -object memory-backend-file,id=mem,size=1024M,mem-path=/var/lib/libvirt/qemu,share=on The access mode can be overriden by specifying token memAccess in numa cell. The mem-path can be defined by new config parameter memory_backing_dir. PS: the sometime failing test qemuhotplug was failing also before this patch :) Jaroslav Safka (4): qemu,conf: Rename virNumaMemAccess to virDomainMemoryAccess qemu_conf: Add param memory_backing_dir conf: Add new xml elements for file memorybacking support qemu: Add args generation for file memory backing docs/formatdomain.html.in | 9 ++ docs/schemas/domaincommon.rng | 30 +++++ src/conf/domain_conf.c | 133 ++++++++++++++++----- src/conf/domain_conf.h | 22 ++++ src/conf/numa_conf.c | 15 +-- src/conf/numa_conf.h | 14 +-- src/qemu/libvirtd_qemu.aug | 3 + src/qemu/qemu.conf | 4 + src/qemu/qemu_command.c | 55 ++++++--- src/qemu/qemu_conf.c | 9 ++ src/qemu/qemu_conf.h | 2 + src/qemu/qemu_process.c | 2 +- src/qemu/test_libvirtd_qemu.aug.in | 1 + .../qemuxml2argv-fd-memory-no-numa-topology.args | 21 ++++ .../qemuxml2argv-fd-memory-no-numa-topology.xml | 27 +++++ .../qemuxml2argv-fd-memory-numa-topology.args | 24 ++++ .../qemuxml2argv-fd-memory-numa-topology.xml | 30 +++++ .../qemuxml2argv-fd-memory-numa-topology2.args | 27 +++++ .../qemuxml2argv-fd-memory-numa-topology2.xml | 31 +++++ .../qemuxml2argv-fd-memory-numa-topology3.args | 30 +++++ .../qemuxml2argv-fd-memory-numa-topology3.xml | 32 +++++ .../qemuxml2argv-memorybacking-set.xml | 24 ++++ .../qemuxml2argv-memorybacking-unset.xml | 24 ++++ tests/qemuxml2argvtest.c | 13 ++ .../qemuxml2xmlout-memorybacking-set.xml | 32 +++++ .../qemuxml2xmlout-memorybacking-unset.xml | 32 +++++ tests/qemuxml2xmltest.c | 3 + 27 files changed, 584 insertions(+), 65 deletions(-) create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-fd-memory-no-numa-topology.args create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-fd-memory-no-numa-topology.xml create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-fd-memory-numa-topology.args create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-fd-memory-numa-topology.xml create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-fd-memory-numa-topology2.args create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-fd-memory-numa-topology2.xml create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-fd-memory-numa-topology3.args create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-fd-memory-numa-topology3.xml create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-memorybacking-set.xml create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-memorybacking-unset.xml create mode 100644 tests/qemuxml2xmloutdata/qemuxml2xmlout-memorybacking-set.xml create mode 100644 tests/qemuxml2xmloutdata/qemuxml2xmlout-memorybacking-unset.xml -- 2.7.4 -------------------------------------------------------------- Intel Research and Development Ireland Limited Registered in Ireland Registered Office: Collinstown Industrial Park, Leixlip, County Kildare Registered Number: 308263 This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.

3 7

[libvirt] [PATCH v2] cpu: fix typo: rename __kvm_hv_spinlock to __kvm_hv_spinlocks
by Maxim Nestratov 09 Feb '17

09 Feb '17

Strings associated with virDomainHyperv values in domain_conf.c are used to construct HyperV CPU features names to be compared with names defined in cpu_x86_data.h and the names for HyperV "spinlocks" feature don't match. This leads to a misleading warning: "host doesn't support hyperv 'spinlocks' feature" even when it's supported. Let's fix it and rename along with it VIR_CPU_x86_KVM_HV_SPINLOCK to VIR_CPU_x86_KVM_HV_SPINLOCKS. Signed-off-by: Maxim Nestratov <mnestratov(a)virtuozzo.com> --- src/cpu/cpu_x86.c | 4 ++-- src/cpu/cpu_x86_data.h | 8 +++++++- 2 files changed, 9 insertions(+), 3 deletions(-) diff --git a/src/cpu/cpu_x86.c b/src/cpu/cpu_x86.c index 23a519e..40d98ee 100644 --- a/src/cpu/cpu_x86.c +++ b/src/cpu/cpu_x86.c @@ -100,7 +100,7 @@ KVM_FEATURE_DEF(VIR_CPU_x86_KVM_HV_STIMER, 0x40000003, 0x00000008); KVM_FEATURE_DEF(VIR_CPU_x86_KVM_HV_RELAXED, 0x40000003, 0x00000020); -KVM_FEATURE_DEF(VIR_CPU_x86_KVM_HV_SPINLOCK, +KVM_FEATURE_DEF(VIR_CPU_x86_KVM_HV_SPINLOCKS, 0x40000003, 0x00000022); KVM_FEATURE_DEF(VIR_CPU_x86_KVM_HV_VAPIC, 0x40000003, 0x00000030); @@ -124,7 +124,7 @@ static virCPUx86Feature x86_kvm_features[] = KVM_FEATURE(VIR_CPU_x86_KVM_HV_SYNIC), KVM_FEATURE(VIR_CPU_x86_KVM_HV_STIMER), KVM_FEATURE(VIR_CPU_x86_KVM_HV_RELAXED), - KVM_FEATURE(VIR_CPU_x86_KVM_HV_SPINLOCK), + KVM_FEATURE(VIR_CPU_x86_KVM_HV_SPINLOCKS), KVM_FEATURE(VIR_CPU_x86_KVM_HV_VAPIC), KVM_FEATURE(VIR_CPU_x86_KVM_HV_VPINDEX), KVM_FEATURE(VIR_CPU_x86_KVM_HV_RESET), diff --git a/src/cpu/cpu_x86_data.h b/src/cpu/cpu_x86_data.h index 4660ab6..4c51e43 100644 --- a/src/cpu/cpu_x86_data.h +++ b/src/cpu/cpu_x86_data.h @@ -49,11 +49,17 @@ struct _virCPUx86CPUID { # define VIR_CPU_x86_KVM_PV_EOI "__kvm_pv_eoi" # define VIR_CPU_x86_KVM_PV_UNHALT "__kvm_pv_unhalt" # define VIR_CPU_x86_KVM_CLOCKSOURCE_STABLE_BIT "__kvm_clocksource_stable" + +/* + * The following HyperV feature names suffixes must exactly match corresponding + * ones defined in domain_conf.c. E.g "__kvm_runtime" -> "runtime", + * "__kvm_hv_spinlocks" -> "spinlocks" etc. +*/ # define VIR_CPU_x86_KVM_HV_RUNTIME "__kvm_hv_runtime" # define VIR_CPU_x86_KVM_HV_SYNIC "__kvm_hv_synic" # define VIR_CPU_x86_KVM_HV_STIMER "__kvm_hv_stimer" # define VIR_CPU_x86_KVM_HV_RELAXED "__kvm_hv_relaxed" -# define VIR_CPU_x86_KVM_HV_SPINLOCK "__kvm_hv_spinlock" +# define VIR_CPU_x86_KVM_HV_SPINLOCKS "__kvm_hv_spinlocks" # define VIR_CPU_x86_KVM_HV_VAPIC "__kvm_hv_vapic" # define VIR_CPU_x86_KVM_HV_VPINDEX "__kvm_hv_vpindex" # define VIR_CPU_x86_KVM_HV_RESET "__kvm_hv_reset" -- 2.4.11

2 2

[libvirt] [PATCH] cpu_x86: Fix memory leak in virCPUx86Translate
by Jiri Denemark 09 Feb '17

09 Feb '17

virCPUDefStealModel is called with keepVendor == true which means the cpu structure will keep its original vendor/vendor_id values. Thus it makes no sense to copy them to the translated definition as they won't be used there anyway. Except that the translated->vendor pointer might get lost in x86Decode. Signed-off-by: Jiri Denemark <jdenemar(a)redhat.com> --- src/cpu/cpu_x86.c | 4 ---- 1 file changed, 4 deletions(-) diff --git a/src/cpu/cpu_x86.c b/src/cpu/cpu_x86.c index 23a519ec0..9e22f896e 100644 --- a/src/cpu/cpu_x86.c +++ b/src/cpu/cpu_x86.c @@ -2709,10 +2709,6 @@ virCPUx86Translate(virCPUDefPtr cpu, if (!(translated = virCPUDefCopyWithoutModel(cpu))) goto cleanup; - if (VIR_STRDUP(translated->vendor, cpu->vendor) < 0 || - VIR_STRDUP(translated->vendor_id, cpu->vendor_id) < 0) - goto cleanup; - if (x86Decode(translated, &model->data, models, nmodels, NULL, 0) < 0) goto cleanup; -- 2.11.1

2 1

[libvirt] [PATCH 00/11] qemu: Big namespace cleanup
by Michal Privoznik 09 Feb '17

09 Feb '17

Couple of these patches were requested in reviews to my previous namespace patches, couple of them are my own idea. At the same time, 3 bugs are fixed: 1) qemuSecurity wrappers are preferred over virSecurityManager 2) corresponding /dev entry is not created when attaching vhost scsi device 3) corresponding /dev entry is not created when doing blockjobs Michal Privoznik (11): qemuDomainAttachSCSIVHostDevice: Prefer qemuSecurity wrappers syntax-check: Enforce qemuSecurity qemuDomainAttachSCSIVHostDevice: manage /dev entry qemu_security: Kill code duplication qemu_security: Drop qemuSecuritySetRestoreAllLabelData struct qemu_domain: Don't pass virDomainDeviceDefPtr to ns helpers qemuDomainNamespaceSetupDisk: Drop useless @src variable qemuDomainNamespace{Setup,Teardown}Disk: Don't pass pointer to full disk qemuDomainDiskChainElement{Prepare,Revoke}: manage /dev entry qemu_security: Introduce ImageLabel APIs qemuDomainNamespaceSetupDisk: Simplify disk check cfg.mk | 8 ++ src/qemu/qemu_domain.c | 65 ++++++---------- src/qemu/qemu_domain.h | 4 +- src/qemu/qemu_driver.c | 2 +- src/qemu/qemu_hotplug.c | 20 +++-- src/qemu/qemu_security.c | 189 ++++++++++++----------------------------------- src/qemu/qemu_security.h | 8 ++ 7 files changed, 102 insertions(+), 194 deletions(-) -- 2.11.0

3 37