- Devel - libvirt List Archives

[libvirt] [PATCH] qemu: fix the vm deadlock when delete an unexist iothread
by Luyao Huang 04 May '15

04 May '15

https://bugzilla.redhat.com/show_bug.cgi?id=1218145 Jump to endjob instead of cleanup to fix this deadlock. Signed-off-by: Luyao Huang <lhuang(a)redhat.com> --- src/qemu/qemu_driver.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index 3695b26..5e998c9 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -6379,7 +6379,7 @@ qemuDomainChgIOThread(virQEMUDriverPtr driver, _("cannot find IOThread '%u' in persistent " "iothreadids"), iothread_id); - goto cleanup; + goto endjob; } virDomainIOThreadIDDel(persistentDef, iothread_id); -- 1.8.3.1

2 1

[libvirt] Release of libvirt-1.2.15
by Daniel Veillard 04 May '15

04 May '15

It's out ! Libvirt 1.2.15 is released, I tagged it in git and provided signed tarballs and rpms available at the usual place: ftp://libvirt.org/libvirt/ I also pushed a python bindings release too at: ftp://libvirt.org/libvirt/python/ This is again a reasonably large release with around 375 patches merged in, including a fairly large set of bug fixes across all areas. A few user visible features and a large set of improvements are there too: Features: - Implement virDomainAddIOThread and virDomainDelIOThread (John Ferlan) - libxl: Introduce configuration file for libxl driver (Jim Fehlig) - Add VIR_DOMAIN_EVENT_ID_DEVICE_ADDED event (Ján Tomko) - various improvements to parallels driver (Maxim Nestratov, Dmitry Guryanov) Documentation: - qemu: Fix comment for timeDelta (Jiri Denemark) - Fix typo in comment about memory binding (zhang bo) - virhostdev: Fix comments for virHostdevReAttachPCIDevices (John Ferlan) - Add articles to virDomainDeviceDetachFlags docs (Ján Tomko) - lib: snapshot: Explain that only one layer of images is inserted (Peter Krempa) - Visually separate snapshot disk subelements (Ján Tomko) - Add info (where necessary) that paths should be specified as absolute (Erik Skultety) - Typos: Get rid of dependan(t|cies) (Martin Kletzander) - Add Host sFlow into monitoring apps (Martin Kletzander) Portability: - Fix building virnetserverclientmock with MinGW (Martin Kletzander) - Cleanup parameters for mock libraries (Martin Kletzander) - Cleanup srcdir usage (Michal Privoznik) - Add missing linker flags for MinGW build (Pavel Fedin) - build: add pragma directive to fix build on some gcc (Pavel Hrdina) - build: avoid obsolete index() (Eric Blake) - util: fix build on non-Linux (Roman Bogorodskiy) - vircapstest: fix build without LXC, QEMU or XEN (Roman Bogorodskiy) - virbuffer: fix build on rhel-6 (Pavel Hrdina) - tests: fix build on old 32-bit platforms (Eric Blake) - build: provide virNetDevSysfsFile on non-Linux (Eric Blake) - build: include correct header for time() (Eric Blake) - xen: fix build error on rhel-5 (Pavel Hrdina) - qemu: Avoid shadow of 'sync' symbol (Peter Krempa) Bug Fixes: - qemu: Fix bus and lun checks when scsi-disk.channel not present (John Ferlan) - rpm-build: update %files section for libxl (Pavel Hrdina) - qemu: blockjob: Call qemuDomainSupportsBlockJobs only on online VMs (Peter Krempa) - qemu: Check address type for USB disks (Jiri Denemark) - cpu: Honor vendor_id override in host-model (Jiri Denemark) - iscsi: do not fail to stop a stopped pool (Ján Tomko) - qemu: conf: Reject memory device if it would exceed configured max size (Peter Krempa) - qemu: blockCopy: Allow reuse of raw image for shallow block copy (Peter Krempa) - tests: free ChardevInfo correctly in qemumonitorjsontest (Zhang Bo) - storage: fs: Ignore volumes that fail to open with EACCESS/EPERM (Cole Robinson) - storage: If driver startup state syncing fails, delete statefile (Cole Robinson) - storage: Don't leave stale state file if pool startup fails (Cole Robinson) - storage: Fix autostart dir for qemu:///session (Cole Robinson) - qemu: qemuProcessDetectIOThreadPIDs invert checks (John Ferlan) - conf: Resolve some Coverity errors (John Ferlan) - network: check newDef for used bridge names in addition to def (Laine Stump) - network: check for bridge name conflict with existing devices (Laine Stump) - network: move auto-assign of bridge name from XML parser to net driver (Laine Stump) - test: Fix actual vs. expected in virtTestCompareFiles (Laine Stump) - qemu: Resolve Coverity DEADCODE (John Ferlan) - util: fix memleak in virStorageSourceClear (zhang bo) - storage: Resolve Coverity UNINIT (John Ferlan) - Fix check for return value of qemuDomainAgentAvailable() (Martin Kletzander) - conf: fix memleak in virDomainHostdevDefClear (Zhang Bo) - conf: fix memleak in virDomainNetIpParseXML (Zhang Bo) - qemu: fix memleak in virCapabilitiesDomainDataLookup (Zhang Bo) - qemu: fix memleaks in qemuBuildCommandLine (Zhang Bo) - util: fix memleak in virFindSCSIHostByPCI (Zhang Bo) - tests: fix some memleaks in tests (Zhang Bo) - qemu: Fix domain object leak in qemuDomainInterfaceAddresses (Peter Krempa) - libxl: initialize sdl.opengl defbool in libxlMakeVfb (Jim Fehlig) - libxl: initialize vfb defbools in libxlMakeVfb (Olaf Hering) - parallels: fix crash in prlsdkAddNet in case of CT definition (Maxim Nestratov) - bhyve: fix build in tests (Roman Bogorodskiy) - Fix memory leak in virNetSocketNewConnectUNIX (Jiri Denemark) - parallels: fix IS_CT macro (Maxim Nestratov) - rng: fix port number range validation (Daniel P. Berrange) - qemu: Don't fail to reboot domains with unresponsive agent (zhang bo) - bhyve: fix build in tests (Roman Bogorodskiy) - spec: Fix libxl build with driver modules (Cole Robinson) - configure: Fix --loader-nvram typo (Cole Robinson) - parallels: don't forget to unlock domain after successful virDomainObjListFindByUUID call (Maxim Nestratov) - vircommand: fix polling in virCommandProcessIO (Roman Bogorodskiy) - util: storage: Fix possible crash when source path is NULL (Peter Krempa) - qemu: Fix issues with maxMemory in qemuDomainSetMemoryFlags() (Luyao Huang) - qemu: set macvtap physdevs online when macvtap is set online (Laine Stump) - util: set MAC address for VF via netlink message to PF+VF# when possible (Laine Stump) - qemu: Always refresh capabilities if no <guests> found (Cole Robinson) - scsi: Adjust return values from processLU (John Ferlan) - storage: Fix check for stable path check (John Ferlan) - domtop: Properly free cpu status (Michal Privoznik) - parallels: fix return status for parallelsConnectOpen (Nikolay Shirokovskiy) - lxc: fix starting a domain with non-strict numa memory mode (Luyao Huang) - lxc: fix starting a domain with a cpuset but no numatune (Luyao Huang) - rng: Forbid to validate mismatched <disk> 'device' and 'type' attributes (Erik Skultety) - domcaps: Check for architecture more wisely (Michal Privoznik) - virNetSocketNewConnectUNIX: Don't unlink(NULL) (Michal Privoznik) - libxl: drop virDomainObj lock when destroying a domain (Jim Fehlig) - libxl: acquire a job when destroying a domain (Jim Fehlig) - virCondWaitUntil: calculate timespec correctly (Michael Chapman) - qemu: bulk stats: Ignore errors from missing/inaccessible disks (Peter Krempa) - Strip control codes in virBufferEscapeString (Ján Tomko) - Ignore storage volumes with control codes in their names (Ján Tomko) - Strip control characters from sysfs attributes (Ján Tomko) - tests: Resolve Coverity RESOURCE_LEAK (John Ferlan) - virsh: fix regression in 'virsh event' by domain (Eric Blake) - qemu: monitor: Fix qemuMonitorGetAllBlockStatsInfo with HMP (Peter Krempa) - qemu: monitor: Ensure that qemuMonitorSetLink is called with non-null name (Peter Krempa) - virNetSocketNewConnectUNIX: Use flocks when spawning a daemon (Michal Privoznik) - sanlock: Use VIR_ERR_RESOURCE_BUSY if sanlock_acquire fails (Jiri Denemark) - hostdev: fix loop index error when resetvfnetconfig (Huanle Han) - qemu: fix index error when clean up vport profile (Huanle Han) - virt-aa-helper: add unix channels (esp for qemu-guest-agent) (Serge Hallyn) - qemu: save domain status after set memory parameters (Shanzhi Yu) - virsh: improve the error for wrong memtune parameters (Shanzhi Yu) - qemu: drivePivot: Fix assumption when 'block-job-complete' fails (Peter Krempa) - qemu: Fix condition for checking vcpu when pinning vcpus (Peter Krempa) - qemu: Enforce WWN to be unique among VM's disks (Peter Krempa) - virBitmap: Place virBitmapIsAllClear check after virBitmapParse calls (Erik Skultety) - Fix usb device version parsing issues (Ján Tomko) - Support IPv6 in networkGetNetworkAddress (Ján Tomko) - util: Update virNetDevGetIPAddress to get IPv6 addresses (Luyao Huang) - parallels: fix virDomainDefineXML for domain in saved state (Dmitry Guryanov) - conf: fix virDomainDefFormatInternal for parallels (Dmitry Guryanov) - conf: fix virDomainDefParseXML for parallels (Dmitry Guryanov) - storage: Don't duplicate efforts of backend driver (John Ferlan) - storage: Need to update freeExtent at delete primary partition (John Ferlan) - storage: Fix issues in storageVolResize (John Ferlan) - Open /proc/PID/ns/* read-only to avoid getting permission denied (Cédric Bosdonnat) - lxc: create the required directories upon driver start (Lubomir Rintel) - util: file: Don't carelessly sanitize URIs (Peter Krempa) - fix memleak in qemuRestoreCgroupState (Luyao Huang) - qemuProcessHook: Call virNuma*() only when needed (Michal Privoznik) - parallels: delete old networks in prlsdkDoApplyConfig before adding new ones (Maxim Nestratov) - util: fix removal of callbacks in virCloseCallbacksRun (Michael Chapman) - qemu: fix race between disk mirror fail and cancel (Michael Chapman) - qemu: fix error propagation in qemuMigrationBegin (Michael Chapman) - qemu: fix crash in qemuProcessAutoDestroy (Michael Chapman) - virQEMUDriverGetConfig: Fix memleak (Michal Privoznik) - virDomainVirtioSerialAddrSetFree: Fix memleak (Michal Privoznik) - qemuSetupCgroupForVcpu: Fix memleak (Michal Privoznik) - libvirt: virsh: Kill all uses of __FUNCTION__ in error messages (Noella Ashu) - virsh: Fix domifaddr output in quiet mode (Luyao Huang) - libxl: fix dom0 balloon logic (Jim Fehlig) - hostdev: Fix index error in loop after remove an element (Huanle Han) - Fix xlconfigtest with older libxl (Ján Tomko) - iscsi: Fix exit path for virStorageBackendISCSIFindLUs failure (John Ferlan) - qemuDomainBlockCopy: Check @granularity to be a power of two (Luyao Huang) - nodeinfodata: Remove broken symlinks and uneeded files (Michal Privoznik) - qemu: lifecycle: make agent-mode shutdown and reboot timeout (zhang bo) Improvements: - main: add new generated files to .gitignore (Pavel Hrdina) - libxl: provide integration with lock manager (Jim Fehlig) - domain: conf: Drop unused OSTYPE_AIX (Cole Robinson) - qemu: migration: use sync block job helpers (Michael Chapman) - qemuDomainBlockJobAbort: use sync block job helpers (Michael Chapman) - qemuProcessStop: wake up pending sync block jobs (Michael Chapman) - qemuBlockJobSync*: introduce sync block job helpers (Michael Chapman) - qemuBlockJobEventProcess: move to new source file (Michael Chapman) - qemu: Validate available slot count for memory devices (Peter Krempa) - qemu: command: Validate that memory devices slot ID is in range (Peter Krempa) - parallels: implement domainDetachDevice and domainDetachDeviceFlags (Maxim Nestratov) - parallels: add prlsdkDelDisk and prlsdkGetDiskIndex functions (Maxim Nestratov) - qemu: make qemuMonitorChardevInfoFree non-static (Zhang Bo) - storage: Break out storageDriverLoadPoolState (Cole Robinson) - qemu: Remove need for qemuMonitorIOThreadInfoFree (John Ferlan) - qemu: Remove need for qemuDomainParseIOThreadAlias (John Ferlan) - conf: explicitly initialize 'cpumask' variable (Roman Bogorodskiy) - virsh: Add iothreadadd and iothreaddel commands (John Ferlan) - qemu: Add support to Add/Delete IOThreads (John Ferlan) - domain: Introduce virDomainIOThreadSchedDelId (John Ferlan) - remote: Add support for AddIOThread and DelIOThread (John Ferlan) - conf: Adjust the iothreadsched expectations (John Ferlan) - Move iothreadspin information into iothreadids (John Ferlan) - conf: Move virDomainPinIsDuplicate and make static (John Ferlan) - qemu: Use domain iothreadids to IOThread's 'thread_id' (John Ferlan) - conf: Add new domain XML element 'iothreadids' (John Ferlan) - openvz: Drop useless domain lookup (Michal Privoznik) - qemu: Connect to guest agent after channel hotplug (Peter Krempa) - qemu: agent: Differentiate errors when the agent channel was hotplugged (Peter Krempa) - qemu: Reuse qemuDomainAgentAvailable in qemuDomainInterfaceAddresses (Peter Krempa) - qemu: agent: Reuse virJSONValueObjectCreateVArgs in qemuAgentMakeCommand (Peter Krempa) - libxl: remove duplicate check for pci subsystem type (Olaf Hering) - libxl: use pci init and dispose functions (Olaf Hering) - hostdev: fix net config restore error (Huanle Han) - hostdev: Create virHostdevIsPCINetDevice (Huanle Han) - qemu: Build nvram directory at driver startup (Cole Robinson) - qemu: Build channel autosocket directory at driver startup (Cole Robinson) - qemu: chown autoDumpPath on driver startup (Cole Robinson) - qemu: conf: Clarify paths that are relative to libDir (Cole Robinson) - tests: qemu: Couple aarch64 CPU tests (Cole Robinson) - migration: Usable time statistics without requiring NTP (Jiri Denemark) - uml: s/virDomainObjListFindByID/virDomainObjListFindByUUID/ (Michal Privoznik) - virDomainObjList: Introduce yet another hash table (Michal Privoznik) - virDomainObjListFindByName: Return referenced object (Michal Privoznik) - Introduce virDomainObjEndAPI (Michal Privoznik) - virDomainObjListNew: Use virObjectFreeHashData (Michal Privoznik) - virDomainObjListAddLocked: s/false/NULL/ for @oldDef (Michal Privoznik) - parallels: implement .domainGetMaxMemory (Dmitry Guryanov) - Apparmor: allow reading block-rbd.so (Cédric Bosdonnat) - Allow access to vendor and device file for PCI device passthrough (Cédric Bosdonnat) - qemu: cgroup: Fix priorities when setting emulatorpin (Peter Krempa) - tests: Add VIR_TEST_REGENERATE_OUTPUT (Cole Robinson) - tests: Add virtTestCompareToFile (Cole Robinson) - tests: Use *DefParseFile more (Cole Robinson) - tests: Add VIR_TEST_DEBUG and VIR_TEST_VERBOSE (Cole Robinson) - parallels_sdk: Utilize parallelsDomObjFromDomain() (Michal Privoznik) - parallels_driver: Utilize parallelsDomObjFromDomain() (Michal Privoznik) - parallels: Introduce parallelsDomObjFromDomain() (Michal Privoznik) - struct _parallelsConn: Mark @domains as immutable pointer (Michal Privoznik) - spec: Point fedora --with-loader-nvram at nightly firmware repo (Cole Robinson) - configure: Report --with-loader-nvram value in summary (Cole Robinson) - locking: relax PID requirement (Jim Fehlig) - util: storage: Improve error message when requesting image above 'start' (Peter Krempa) - util: storage: Add hint to error message that indexed access was used (Peter Krempa) - util: command: Deduplicate code in virCommandNewArgList (Peter Krempa) - qemu: migration: Refactor hostdev validation in migration check (Peter Krempa) - processNicRxFilterChangedEvent: Take appropriate actions for NET_TYPE_NETWORK too (Michal Privoznik) - virDomainActualNetDefContentsFormat: Format class_id only for status XML (Michal Privoznik) - scsi: Change return values for virStorageBackendSCSIFindLUs (John Ferlan) - scsi: Adjust return value for virStorageBackendSCSINewLun (John Ferlan) - storage: Split out the stable path check (John Ferlan) - xend: Remove a couple of unused function prototypes. (Richard W.M. Jones) - domain: conf: Drop expectedVirtTypes (Cole Robinson) - caps: Use DomainDataLookup to replace GuestDefault* (Cole Robinson) - domain: conf: Use CapabilitiesDomainDataLookup for caps validation (Cole Robinson) - domain: conf: Do ostype/arch/machine parsing earlier (Cole Robinson) - caps: Add virCapabilitiesDomainDataLookup (Cole Robinson) - caps: Convert to use VIR_DOMAIN_VIRT internally (Cole Robinson) - domain: Convert os.type to VIR_DOMAIN_OSTYPE enum (Cole Robinson) - caps: Switch AddGuest to take VIR_DOMAIN_OSTYPE value (Cole Robinson) - caps: Use an enum internally for ostype value (Cole Robinson) - domain: conf: Don't validate VM ostype/arch at daemon startup (Cole Robinson) - domain: conf: Better errors on bad os <type> values (Cole Robinson) - parallels: make device addressing consistent (Nikolay Shirokovskiy) - rng: Move sgio attr definition to a separate block (Erik Skultety) - Apparmor: removed duplicate nscd rules (Cédric Bosdonnat) - schema: Allow multiple machines for VMs and unite architectures (Martin Kletzander) - qemuDomainShutdownFlags: Set fakeReboot more frequently (zhang bo) - Introduce virNetDevBandwidthUpdateFilter (Michal Privoznik) - virnetdevbandwidth.c: Separate tc filter creation to a function (Michal Privoznik) - virNetDevBandwidthSet: Add priority to filter (Michal Privoznik) - virDomainActualNetDefContentsFormat: Format class_id more frequently (Michal Privoznik) - nwfilter: Partly initialize driver even for non-privileged users (Michal Privoznik) - lxc: move wireless PHYs to a network namespace (Lubomir Rintel) - libxl: Move job acquisition in libxlDomainStart to callers (Jim Fehlig) - libxl: support HVM direct kernel boot (Jim Fehlig) - xenconfig: don't use "kernel" for hvmloader (Jim Fehlig) - xenconfig: move <os> parsing/formating to config-specific files (Jim Fehlig) - xenconfig: remove redunant parsing of device_model (Jim Fehlig) - xenconfig: export xenConfigCopyString (Jim Fehlig) - storage: Refactor virStorageBackendSCSINewLun (John Ferlan) - qemu: monitor: Refactor and fix monitor checking (Peter Krempa) - daemon: Prefix sysctl configuration filename with a number (Jiri Denemark) - RNG schema: allow plain @floor to <bandwidth/> (Michal Privoznik) - virNetDevBandwidthPlug: Update function description (Michal Privoznik) - Add functions dealing with control characters in strings (Ján Tomko) - tests: rename testStripIPv6BracketsData to testStripData (Ján Tomko) - Add an example for EVENT_ID_DEVICE_ADDED (Ján Tomko) - Emit VIR_DOMAIN_EVENT_ID_DEVICE_ADDED in the QEMU driver (Ján Tomko) - Cleanup "/sys/class/net" usage (Michal Privoznik) - tests: Add virnetdevtestdata to EXTRA_DIST (Michal Privoznik) - node: udev: Remove some redundant error reports (Peter Krempa) - virPidFileConstructPath: Drop useless VIR_FREE() (Michal Privoznik) - qemu: monitor: @running in qemuMonitorGetStatus is always non-NULL (Peter Krempa) - qemu: monitor: Don't use 'ret' variable where not necessary (Peter Krempa) - qemu: monitor: Sanitize control flow in qemuMonitorSetCapabilities (Peter Krempa) - qemu: monitor: Clean up coding style (Peter Krempa) - json: export non-static functions (Martin Kletzander) - Change virConnectPtr into virObjectLocklable (Martin Kletzander) - closeCallback is already lockable, initialize it as such (Martin Kletzander) - configure: Align messages (Martin Kletzander) - storage: Add duplicate devices check for zfs pool def (John Ferlan) - storage: Add duplicate source pool for Gluster pool def (John Ferlan) - storage: Add duplicate host check for Sheepdog pool def (John Ferlan) - storage: Remove default from switch in virStoragePoolSourceFindDuplicate (John Ferlan) - storage: Use virStoragePoolSourceMatchSingleHost for NETFS (John Ferlan) - storage: Add check for different ports for host duplicate matching (John Ferlan) - storage: Create virStoragePoolSourceMatchSingleHost (John Ferlan) - storage: Refactor iSCSI Source matching (John Ferlan) - Introduce virnetdevtest (Michal Privoznik) - cfg: Disallow usage of 'index' as variable name (John Ferlan) - nwfilter: Change 'index' to 'idx' (John Ferlan) - snapshot: Change 'index' to 'idx' (John Ferlan) - util: Use 'idx' instead of 'index' (John Ferlan) - qemu: Adjust the prototype to match the function (John Ferlan) - sparc: Add default PCI root controller (Martin Kletzander) - qemu: Use 'idx' instead of 'index' for variable name (John Ferlan) - qemu: Refactor qemuDomainBlockJobAbort() (Peter Krempa) - qemu: blockPull: Refactor the rest of qemuDomainBlockJobImpl (Peter Krempa) - qemu: blockjob: Separate qemuDomainBlockJobAbort from qemuDomainBlockJobImpl (Peter Krempa) - qemu: blockjob: Split qemuDomainBlockJobSetSpeed from qemuDomainBlockJobImpl (Peter Krempa) - qemu: monitor: Extract handling of JSON block job error codes (Peter Krempa) - qemu: monitor: json: Refactor error code class checker (Peter Krempa) - conf: Refactor virDomainVcpuPinDefParseXML (Peter Krempa) - conf: Error out if iothread id is missing in iothreadpin (Peter Krempa) - onf: Split up virDomainVcpuPinDefParseXML (Peter Krempa) - conf: Split out parsing of emulatorpin (Peter Krempa) - conf: ABI: Check WWN in disk abi stability check (Peter Krempa) - Convert virDomainPinIsDuplicate into bool return (John Ferlan) - Rename qemuCheckIothreads to qemuCheckIOThreads (John Ferlan) - Rewrite vshParseCPUList (Ján Tomko) - Rewrite vshPrintPinInfo (Ján Tomko) - Parallels: implemented domainAttachDevice (Alexander Burluka) - Parallels: implement domainAttachDeviceFlags (Alexander Burluka) - Parallels: remove disk serial number check (Alexander Burluka) - Remove overengineered loop (Ján Tomko) - Introduce struct _virStorageBackendQemuImgInfo (Ján Tomko) - Rename virStorageBackendCreateQemuImgCmd (Ján Tomko) - Use XPath when parsing snapshot disk definition (Ján Tomko) - Split out storage format 'compat' attribute sanity check (Ján Tomko) - Remove feature formating funcs from pool-specific options (Ján Tomko) - Rewrite usb device version parsing (Ján Tomko) - Do xml->xml test for usb-redir-filter (Ján Tomko) - Do not require virtio-pci devices when using virtio-mmio (Ján Tomko) - qemuMigrationPrecreateStorage: Fix debug message (Michal Privoznik) - qemu_migration.c: sleep first before checking for migration status. (Xing Lin) - conf: Don't output <cpu> tag if it contains no information. (Andrea Bolognani) - qemu: qemuDomainHotplugVcpus - separate out pin adjustment code (John Ferlan) - AUTHORS: Add my email address to AUTHORS.in (Dmitry Guryanov) - util: Replace virNetDevGetIPv4Address with virNetDevGetIPAddress (John Ferlan) - parallels: add implicit input devices (Dmitry Guryanov) - conf: add input device type for parallels containers (Dmitry Guryanov) - conf: return proper default video type for parallels (Dmitry Guryanov) - conf: add VIR_DOMAIN_VIDEO_TYPE_PARALLELS video type (Dmitry Guryanov) - parallels: don't fill net adapter model for containers (Dmitry Guryanov) - parallels: fill adapter model in virDomainNetDef (Dmitry Guryanov) - parallels: add controllers in prlsdkLoadDomain (Dmitry Guryanov) - parallels: report, that cdroms are readonly (Dmitry Guryanov) - parallels: implement virDomainManagedSave (Dmitry Guryanov) - parallels: split prlsdkDomainChangeState function (Dmitry Guryanov) - parallels: fix headers in parallels_sdk.h (Dmitry Guryanov) - qemu: qemuDomainHotplugVcpus - separate out the del cgroup and pin (John Ferlan) - qemu: qemuDomainHotplugVcpus - separate out the add cgroup (John Ferlan) - cgroup: Use virCgroupNewThread (John Ferlan) - cgroup: Introduce virCgroupNewThread (John Ferlan) - qemu: blockPivot: Don't pause the VM any more since we don't use drive-reopen (Peter Krempa) - qemu: Clean up old leftovers in qemuMonitorDrivePivot (Peter Krempa) - qemu: blockjob: Use the new helpers in qemuDomainGetBlockJobInfo (Peter Krempa) - qemu: domain: Add helper to check block job support (Peter Krempa) - qemu: domain: Introduce helper to retrieve domain monitor object (Peter Krempa) - Apparmor qemu abstraction fixes for SLES (Cédric Bosdonnat) - test: Add tests for virFileSanitizePath (Peter Krempa) - virLXCControllerSetupResourceLimits: Call virNuma*() iff needed (Michal Privoznik) - vbox: Implement virDomainSendKey (Dawid Zamirski) - vbox: Register IKeyboard with the unified API. (Dawid Zamirski) - vircgroup: Introduce virCgroupControllerAvailable (Michal Privoznik) - qemu_driver: check caps after starting block job (Michael Chapman) - qemu_migrate: use nested job when adding NBD to cookie (Michael Chapman) - storage: Introduce storagePoolUpdateAllState function (Erik Skultety) - conf: Introduce virStoragePoolLoadAllState && virStoragePoolLoadState (Erik Skultety) - storage: Add support for storage pool state XML (Erik Skultety) - hostdev: Report the domain name for used hostdevs during nodedev-detach (Shivaprasad G Bhat) - virsh: Improve change-media success message (Cole Robinson) - interface: allow multiple IPv4 addresses in interface XML (Laine Stump) - virsh.pod: Remove redundant --config from attach-interface (Jiri Denemark) - esx: esxNodeGetFreeMemory return info from host. (Dawid Zamirski) - esx: add esxVI_GetInt (Dawid Zamirski) - conf: Change virStoragePoolSaveConfig prototype s/configDir/configFile (Erik Skultety) - conf: Introduce virStoragePoolSaveState (Erik Skultety) - conf: Introduce virStoragePoolDefFormatBuf (Erik Skultety) - Auto add virtio-serial controllers (Ján Tomko) - Assign an address when hotplugging a virtio-serial device (Ján Tomko) - Expand the address set when attaching a virtio-serial controller (Ján Tomko) - Allocate virtio-serial addresses when starting a domain (Ján Tomko) - Add functions to track virtio-serial addresses (Ján Tomko) - Add test for virtio serial port assignment (Ján Tomko) - scsi: Remove unused 'type_path' in processLU (John Ferlan) - iscsi: Use error message from virStorageBackendSCSIFindLUs (John Ferlan) - conf: Introduce virStoragePoolSaveXML (Erik Skultety) - storage: Remove unused attribute conn from 'checkPool' callback (Erik Skultety) - Fix indentation in cmdVcpuPin (Ján Tomko) - virsh: remove stale comment (Ján Tomko) - conf: Rename virDomainHasDiskMirror and detect block jobs properly (Shanzhi Yu) - qemu: snapshot: Check for block jobs individually (Peter Krempa) - Remove unnecessary includes from virsh.h (Ján Tomko) - Remove unused includes from virsh (Ján Tomko) - Clean up headers in src/util/virutil.h (Ján Tomko) - Do not include cpu_map.h in libvirtd.c (Ján Tomko) - Use the DEFAULT_DRIVER_DIR macro (Ján Tomko) - Remove unused macros (Ján Tomko) - qemu: Copy bitmap in a sane way (Peter Krempa) - qemu: cgroup: Kill qemuSetupCgroupVcpuPin() (Peter Krempa) - qemu: cgroup: Kill qemuSetupCgroupIOThreadsPin() (Peter Krempa) - qemu: cgroup: Rename qemuSetupCgroupEmulatorPin to qemuSetupCgroupCpusetCpus (Peter Krempa) - qemu: cgroup: Use priv->autoCpuset instead of using qemuPrepareCpumap() (Peter Krempa) - qemu: cgroup: Properly set up vcpu pinning (Peter Krempa) - qemu: cgroup: Refactor setup for IOThread cgroups (Peter Krempa) - qemu: cgroup: Store auto cpuset instead of re-creating it on demand (Peter Krempa) So thanks everybody for your contributions for this release, be it ideas, bug reports, documentations, patches, reviews ... all those releases represent a massive amount of work ! Daniel -- Daniel Veillard | Open Source and Standards, Red Hat veillard(a)redhat.com | libxml Gnome XML XSLT toolkit http://xmlsoft.org/ http://veillard.com/ | virtualization library http://libvirt.org/

1 0

[libvirt] Entering freeze for libvirt-1.2.15
by Daniel Veillard 03 May '15

03 May '15

I'm a little late, but I have now tagged candidate release 1 for 1.2.15 in git and pushed signed tarballs and rpms to the usual place: ftp://libvirt.org/libvirt/ this seems to work for me in my limited testing, but please give it a try, especially on less common platforms. The https://ci.centos.org/ seems to indicate issues on building the Perl tests as well as virt-manager regression testing, can someone dig into what is happening there ? I will likely push an rc2 on Thursday, and then the final release over the week-end, thanks! Daniel -- Daniel Veillard | Open Source and Standards, Red Hat veillard(a)redhat.com | libxml Gnome XML XSLT toolkit http://xmlsoft.org/ http://veillard.com/ | virtualization library http://libvirt.org/

3 4

[libvirt] [PATCH V3] libxl: provide integration with lock manager
by Jim Fehlig 01 May '15

01 May '15

Provide integration with libvirt's lock manager in the libxl driver. Signed-off-by: Jim Fehlig <jfehlig(a)suse.com> --- V3 of https://www.redhat.com/archives/libvir-list/2015-April/msg01006.html In V3, call virDomainLockProcessStart() with 'paused' parameter set to 'true' and add a call to virDomainLockProcessResume(). src/Makefile.am | 12 +++++++++ src/libxl/libvirtd_libxl.aug | 2 ++ src/libxl/libxl.conf | 10 +++++++ src/libxl/libxl_conf.c | 14 ++++++++++ src/libxl/libxl_conf.h | 6 +++++ src/libxl/libxl_domain.c | 51 +++++++++++++++++++++++++++++++++++- src/libxl/libxl_domain.h | 1 + src/libxl/libxl_driver.c | 25 ++++++++++++++++++ src/libxl/libxl_migration.c | 6 +++++ src/libxl/test_libvirtd_libxl.aug.in | 1 + 10 files changed, 127 insertions(+), 1 deletion(-) diff --git a/src/Makefile.am b/src/Makefile.am index 9a5f16c..1438174 100644 --- a/src/Makefile.am +++ b/src/Makefile.am @@ -2246,6 +2246,12 @@ BUILT_SOURCES += locking/qemu-lockd.conf DISTCLEANFILES += locking/qemu-lockd.conf endif WITH_QEMU +if WITH_LIBXL +nodist_conf_DATA += locking/libxl-lockd.conf +BUILT_SOURCES += locking/libxl-lockd.conf +DISTCLEANFILES += locking/libxl-lockd.conf +endif WITH_LIBXL + locking/%-lockd.conf: $(srcdir)/locking/lockd.conf $(AM_V_GEN)$(MKDIR_P) locking ; \ cp $< $@ @@ -2431,6 +2437,12 @@ nodist_conf_DATA += locking/qemu-sanlock.conf BUILT_SOURCES += locking/qemu-sanlock.conf DISTCLEANFILES += locking/qemu-sanlock.conf endif WITH_QEMU + +if WITH_LIBXL +nodist_conf_DATA += locking/libxl-sanlock.conf +BUILT_SOURCES += locking/libxl-sanlock.conf +DISTCLEANFILES += locking/libxl-sanlock.conf +endif WITH_LIBXL else ! WITH_SANLOCK EXTRA_DIST += $(LOCK_DRIVER_SANLOCK_SOURCES) endif ! WITH_SANLOCK diff --git a/src/libxl/libvirtd_libxl.aug b/src/libxl/libvirtd_libxl.aug index f225954..d5aa150 100644 --- a/src/libxl/libvirtd_libxl.aug +++ b/src/libxl/libvirtd_libxl.aug @@ -25,9 +25,11 @@ module Libvirtd_libxl = (* Config entry grouped by function - same order as example config *) let autoballoon_entry = bool_entry "autoballoon" + let lock_entry = str_entry "lock_manager" (* Each entry in the config is one of the following ... *) let entry = autoballoon_entry + | lock_entry let comment = [ label "#comment" . del /#[ \t]*/ "# " . store /([^ \t\n][^\n]*)?/ . del /\n/ "\n" ] let empty = [ label "#empty" . eol ] diff --git a/src/libxl/libxl.conf b/src/libxl/libxl.conf index c104d40..ba3de7a 100644 --- a/src/libxl/libxl.conf +++ b/src/libxl/libxl.conf @@ -10,3 +10,13 @@ # autoballoon setting. # #autoballoon = 1 + + +# In order to prevent accidentally starting two domains that +# share one writable disk, libvirt offers two approaches for +# locking files: sanlock and virtlockd. sanlock is an external +# project which libvirt integrates with via the libvirt-lock-sanlock +# package. virtlockd is a libvirt implementation that is enabled with +# "lockd". Accepted values are "sanlock" and "lockd". +# +#lock_manager = "lockd" diff --git a/src/libxl/libxl_conf.c b/src/libxl/libxl_conf.c index 1b504fa..29498d5 100644 --- a/src/libxl/libxl_conf.c +++ b/src/libxl/libxl_conf.c @@ -102,6 +102,7 @@ libxlDriverConfigDispose(void *obj) VIR_FREE(cfg->libDir); VIR_FREE(cfg->saveDir); VIR_FREE(cfg->autoDumpDir); + VIR_FREE(cfg->lockManagerName); } @@ -1495,6 +1496,7 @@ int libxlDriverConfigLoadFile(libxlDriverConfigPtr cfg, const char *filename) { virConfPtr conf = NULL; + virConfValuePtr p; int ret = -1; /* Check the file is readable before opening it, otherwise @@ -1512,6 +1514,18 @@ int libxlDriverConfigLoadFile(libxlDriverConfigPtr cfg, if (libxlGetAutoballoonConf(cfg, conf) < 0) goto cleanup; + if ((p = virConfGetValue(conf, "lock_manager"))) { + if (p->type != VIR_CONF_STRING) { + virReportError(VIR_ERR_INTERNAL_ERROR, + "%s", + _("Unexpected type for 'lock_manager' setting")); + goto cleanup; + } + + if (VIR_STRDUP(cfg->lockManagerName, p->str) < 0) + goto cleanup; + } + ret = 0; cleanup: diff --git a/src/libxl/libxl_conf.h b/src/libxl/libxl_conf.h index 5ba1a71..0a1c0db 100644 --- a/src/libxl/libxl_conf.h +++ b/src/libxl/libxl_conf.h @@ -38,6 +38,7 @@ # include "virobject.h" # include "virchrdev.h" # include "virhostdev.h" +# include "locking/lock_manager.h" # define LIBXL_DRIVER_NAME "xenlight" # define LIBXL_VNC_PORT_MIN 5900 @@ -98,6 +99,8 @@ struct _libxlDriverConfig { * memory for new domains from domain0. */ bool autoballoon; + char *lockManagerName; + /* Once created, caps are immutable */ virCapsPtr caps; @@ -144,6 +147,9 @@ struct _libxlDriverPrivate { /* Immutable pointer, lockless APIs*/ virSysinfoDefPtr hostsysinfo; + + /* Immutable pointer. lockless access */ + virLockManagerPluginPtr lockManager; }; # define LIBXL_SAVE_MAGIC "libvirt-xml\n \0 \r" diff --git a/src/libxl/libxl_domain.c b/src/libxl/libxl_domain.c index 3039427..5f5f8e5 100644 --- a/src/libxl/libxl_domain.c +++ b/src/libxl/libxl_domain.c @@ -34,6 +34,7 @@ #include "virlog.h" #include "virstring.h" #include "virtime.h" +#include "locking/domain_lock.h" #define VIR_FROM_THIS VIR_FROM_LIBXL @@ -217,12 +218,36 @@ libxlDomainObjPrivateFree(void *data) { libxlDomainObjPrivatePtr priv = data; + VIR_FREE(priv->lockState); virObjectUnref(priv); } +static int +libxlDomainObjPrivateXMLParse(xmlXPathContextPtr ctxt, void *data) +{ + libxlDomainObjPrivatePtr priv = data; + + priv->lockState = virXPathString("string(./lockstate)", ctxt); + + return 0; +} + +static int +libxlDomainObjPrivateXMLFormat(virBufferPtr buf, void *data) +{ + libxlDomainObjPrivatePtr priv = data; + + if (priv->lockState) + virBufferAsprintf(buf, "<lockstate>%s</lockstate>\n", priv->lockState); + + return 0; +} + virDomainXMLPrivateDataCallbacks libxlDomainXMLPrivateDataCallbacks = { .alloc = libxlDomainObjPrivateAlloc, .free = libxlDomainObjPrivateFree, + .parse = libxlDomainObjPrivateXMLParse, + .format = libxlDomainObjPrivateXMLFormat, }; @@ -667,6 +692,11 @@ libxlDomainCleanup(libxlDriverPrivatePtr driver, virHostdevReAttachDomainDevices(hostdev_mgr, LIBXL_DRIVER_NAME, vm->def, VIR_HOSTDEV_SP_PCI, NULL); + VIR_FREE(priv->lockState); + if (virDomainLockProcessPause(driver->lockManager, vm, &priv->lockState) < 0) + VIR_WARN("Unable to release lease on %s", vm->def->name); + VIR_DEBUG("Preserving lock state '%s'", NULLSTR(priv->lockState)); + vm->def->id = -1; if (priv->deathW) { @@ -960,6 +990,20 @@ libxlDomainStart(libxlDriverPrivatePtr driver, virDomainObjPtr vm, vm->def, VIR_HOSTDEV_SP_PCI) < 0) goto cleanup; + if (virDomainLockProcessStart(driver->lockManager, + "xen:///system", + vm, + true, + NULL) < 0) + goto cleanup; + + if (virDomainLockProcessResume(driver->lockManager, + "xen:///system", + vm, + priv->lockState) < 0) + goto cleanup; + VIR_FREE(priv->lockState); + /* Unlock virDomainObj while creating the domain */ virObjectUnlock(vm); @@ -990,7 +1034,7 @@ libxlDomainStart(libxlDriverPrivatePtr driver, virDomainObjPtr vm, virReportError(VIR_ERR_INTERNAL_ERROR, _("libxenlight failed to restore domain '%s'"), d_config.c_info.name); - goto cleanup; + goto release_dom; } /* @@ -1003,6 +1047,7 @@ libxlDomainStart(libxlDriverPrivatePtr driver, virDomainObjPtr vm, if (libxl_evenable_domain_death(cfg->ctx, vm->def->id, 0, &priv->deathW)) goto cleanup_dom; + if ((dom_xml = virDomainDefFormat(vm->def, 0)) == NULL) goto cleanup_dom; @@ -1040,6 +1085,7 @@ libxlDomainStart(libxlDriverPrivatePtr driver, virDomainObjPtr vm, goto cleanup; cleanup_dom: + ret = -1; if (priv->deathW) { libxl_evdisable_domain_death(cfg->ctx, priv->deathW); priv->deathW = NULL; @@ -1048,6 +1094,9 @@ libxlDomainStart(libxlDriverPrivatePtr driver, virDomainObjPtr vm, vm->def->id = -1; virDomainObjSetState(vm, VIR_DOMAIN_SHUTOFF, VIR_DOMAIN_SHUTOFF_FAILED); + release_dom: + virDomainLockProcessPause(driver->lockManager, vm, &priv->lockState); + cleanup: libxl_domain_config_dispose(&d_config); VIR_FREE(dom_xml); diff --git a/src/libxl/libxl_domain.h b/src/libxl/libxl_domain.h index aa647b8..8c73cc4 100644 --- a/src/libxl/libxl_domain.h +++ b/src/libxl/libxl_domain.h @@ -64,6 +64,7 @@ struct _libxlDomainObjPrivate { virChrdevsPtr devs; libxl_evgen_domain_death *deathW; unsigned short migrationPort; + char *lockState; struct libxlDomainJobObj job; }; diff --git a/src/libxl/libxl_driver.c b/src/libxl/libxl_driver.c index 6a54c73..1ecfd71 100644 --- a/src/libxl/libxl_driver.c +++ b/src/libxl/libxl_driver.c @@ -57,6 +57,7 @@ #include "viratomic.h" #include "virhostdev.h" #include "network/bridge_driver.h" +#include "locking/domain_lock.h" #define VIR_FROM_THIS VIR_FROM_LIBXL @@ -411,6 +412,7 @@ libxlStateCleanup(void) virObjectUnref(libxl_driver->domains); virObjectUnref(libxl_driver->reservedVNCPorts); virObjectUnref(libxl_driver->migrationPorts); + virLockManagerPluginUnref(libxl_driver->lockManager); virObjectEventStateFree(libxl_driver->domainEventState); virSysinfoDefFree(libxl_driver->hostsysinfo); @@ -590,6 +592,14 @@ libxlStateInitialize(bool privileged, goto error; } + if (!(libxl_driver->lockManager = + virLockManagerPluginNew(cfg->lockManagerName ? + cfg->lockManagerName : "nop", + "libxl", + cfg->configBaseDir, + 0))) + goto error; + /* read the host sysinfo */ libxl_driver->hostsysinfo = virSysinfoRead(); @@ -2860,11 +2870,21 @@ libxlDomainAttachDeviceDiskLive(virDomainObjPtr vm, virDomainDeviceDefPtr dev) if (libxlMakeDisk(l_disk, &x_disk) < 0) goto cleanup; + if (virDomainLockDiskAttach(libxl_driver->lockManager, + "xen:///system", + vm, l_disk) < 0) + goto cleanup; + if ((ret = libxl_device_disk_add(cfg->ctx, vm->def->id, &x_disk, NULL)) < 0) { virReportError(VIR_ERR_INTERNAL_ERROR, _("libxenlight failed to attach disk '%s'"), l_disk->dst); + if (virDomainLockDiskDetach(libxl_driver->lockManager, + vm, l_disk) < 0) { + VIR_WARN("Unable to release lock on %s", + virDomainDiskGetSource(l_disk)); + } goto cleanup; } @@ -3005,6 +3025,11 @@ libxlDomainDetachDeviceDiskLive(virDomainObjPtr vm, virDomainDeviceDefPtr dev) goto cleanup; } + if (virDomainLockDiskDetach(libxl_driver->lockManager, + vm, l_disk) < 0) + VIR_WARN("Unable to release lock on %s", + virDomainDiskGetSource(l_disk)); + virDomainDiskRemove(vm->def, idx); virDomainDiskDefFree(l_disk); diff --git a/src/libxl/libxl_migration.c b/src/libxl/libxl_migration.c index 51993c6..1efd98f 100644 --- a/src/libxl/libxl_migration.c +++ b/src/libxl/libxl_migration.c @@ -41,6 +41,7 @@ #include "libxl_driver.h" #include "libxl_conf.h" #include "libxl_migration.h" +#include "locking/domain_lock.h" #define VIR_FROM_THIS VIR_FROM_LIBXL @@ -469,6 +470,7 @@ libxlDomainMigrationPerform(libxlDriverPrivatePtr driver, const char *dname ATTRIBUTE_UNUSED, unsigned int flags) { + libxlDomainObjPrivatePtr priv = vm->privateData; char *hostname = NULL; unsigned short port = 0; char portstr[100]; @@ -503,6 +505,10 @@ libxlDomainMigrationPerform(libxlDriverPrivatePtr driver, sockfd = virNetSocketDupFD(sock, true); virObjectUnref(sock); + if (virDomainLockProcessPause(driver->lockManager, vm, &priv->lockState) < 0) + VIR_WARN("Unable to release lease on %s", vm->def->name); + VIR_DEBUG("Preserving lock state '%s'", NULLSTR(priv->lockState)); + /* suspend vm and send saved data to dst through socket fd */ virObjectUnlock(vm); ret = libxlDoMigrateSend(driver, vm, flags, sockfd); diff --git a/src/libxl/test_libvirtd_libxl.aug.in b/src/libxl/test_libvirtd_libxl.aug.in index 23e667c..baa8c79 100644 --- a/src/libxl/test_libvirtd_libxl.aug.in +++ b/src/libxl/test_libvirtd_libxl.aug.in @@ -3,3 +3,4 @@ module Test_libvirtd_libxl = test Libvirtd_libxl.lns get conf = { "autoballoon" = "1" } +{ "lock_manager" = "lockd" } -- 1.8.4.5

3 6

[libvirt] [PATCH] qemu: Fix bus and lun checks when scsi-disk.channel not present
by John Ferlan 01 May '15

01 May '15

Found by Laine and discussed a bit on internal IRC. Commit id c56fe7f1d6 added support for creating a command line to support scsi-disk.channel. Series was here: http://www.redhat.com/archives/libvir-list/2012-February/msg01052.html Which pointed to a design proposal here: http://permalink.gmane.org/gmane.comp.emulators.libvirt/50428 Which states (in part): Libvirt should check for the QEMU "scsi-disk.channel" property. If it is unavailable, QEMU will only support channel=lun=0 and 0<=target<=7. However, the check added was ensuring that bus != lun *and* bus != 0. So if bus == lun and both were non zero, we'd never make the second check. Changing this to an *or* check fixes the check, but still is less readable than the just checking each for 0 Signed-off-by: John Ferlan <jferlan(a)redhat.com> --- src/qemu/qemu_command.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/src/qemu/qemu_command.c b/src/qemu/qemu_command.c index 0dc8628..583bc11 100644 --- a/src/qemu/qemu_command.c +++ b/src/qemu/qemu_command.c @@ -4127,8 +4127,8 @@ qemuBuildDriveDevStr(virDomainDefPtr def, goto error; } - if ((disk->info.addr.drive.bus != disk->info.addr.drive.unit) && - (disk->info.addr.drive.bus != 0)) { + if (disk->info.addr.drive.bus != 0 && + disk->info.addr.drive.unit != 0) { virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s", _("This QEMU only supports both bus and " "unit equal to 0")); -- 2.1.0

2 1

[libvirt] [PATCH] rpm-build: update %files section for libxl
by Pavel Hrdina 30 Apr '15

30 Apr '15

Recent commit 198cc1d3 introduced integration of lockd and sanlock into libxl, but forget to update libvirt.spec.in to also list new files distributed via package. Signed-off-by: Pavel Hrdina <phrdina(a)redhat.com> --- Pushed under build-breaker rule. libvirt.spec.in | 2 ++ 1 file changed, 2 insertions(+) diff --git a/libvirt.spec.in b/libvirt.spec.in index 20af502..f95e31f 100644 --- a/libvirt.spec.in +++ b/libvirt.spec.in @@ -2129,6 +2129,8 @@ exit 0 %files daemon-driver-libxl %defattr(-, root, root) %config(noreplace) %{_sysconfdir}/libvirt/libxl.conf +%config(noreplace) %{_sysconfdir}/libvirt/libxl-lockd.conf +%config(noreplace) %{_sysconfdir}/libvirt/libxl-sanlock.conf %{_datadir}/augeas/lenses/libvirtd_libxl.aug %{_datadir}/augeas/lenses/tests/test_libvirtd_libxl.aug %dir %attr(0700, root, root) %{_localstatedir}/log/libvirt/libxl/ -- 2.0.5

3 2

[libvirt] [PATCH] main: add new generated files to .gitignore
by Pavel Hrdina 30 Apr '15

30 Apr '15

This means new libxl-lockd.conf and libxl-sanlock.conf Signed-off-by: Pavel Hrdina <phrdina(a)redhat.com> --- Pushed under trivial rule. .gitignore | 2 ++ 1 file changed, 2 insertions(+) diff --git a/.gitignore b/.gitignore index 9d09709..7bcf359 100644 --- a/.gitignore +++ b/.gitignore @@ -124,6 +124,8 @@ /src/libvirt_*helper /src/libvirt_*probes.h /src/libvirt_lxc +/src/locking/libxl-lockd.conf +/src/locking/libxl-sanlock.conf /src/locking/lock_daemon_dispatch_stubs.h /src/locking/lock_protocol.[ch] /src/locking/qemu-lockd.conf -- 2.0.5

1 0

[libvirt] [PATCH] qemu: blockjob: Call qemuDomainSupportsBlockJobs only on online VMs
by Peter Krempa 30 Apr '15

30 Apr '15

Since the qemu capabilities are not initialized for offline VMs the caller might get suboptimal error message: $ virsh blockjob VM PATH --bandwidth 1 error: unsupported configuration: block jobs not supported with this QEMU binary Move the checks after we make sure that the VM is alive. --- src/qemu/qemu_driver.c | 36 ++++++++++++++++++------------------ 1 file changed, 18 insertions(+), 18 deletions(-) diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index 3695b26..d6e7570 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -16555,34 +16555,34 @@ qemuDomainBlockPullCommon(virQEMUDriverPtr driver, goto cleanup; } - if (qemuDomainSupportsBlockJobs(vm, &modern) < 0) + if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0) goto cleanup; + if (!virDomainObjIsActive(vm)) { + virReportError(VIR_ERR_OPERATION_INVALID, "%s", + _("domain is not running")); + goto endjob; + } + + if (qemuDomainSupportsBlockJobs(vm, &modern) < 0) + goto endjob; + if (!modern) { if (base) { virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s", _("partial block pull not supported with this " "QEMU binary")); - goto cleanup; + goto endjob; } if (bandwidth) { virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s", _("setting bandwidth at start of block pull not " "supported with this QEMU binary")); - goto cleanup; + goto endjob; } } - if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0) - goto cleanup; - - if (!virDomainObjIsActive(vm)) { - virReportError(VIR_ERR_OPERATION_INVALID, "%s", - _("domain is not running")); - goto endjob; - } - if (!(device = qemuDiskPathToAlias(vm, path, &idx))) goto endjob; disk = vm->def->disks[idx]; @@ -16682,9 +16682,6 @@ qemuDomainBlockJobAbort(virDomainPtr dom, if (virDomainBlockJobAbortEnsureACL(dom->conn, vm->def) < 0) goto cleanup; - if (qemuDomainSupportsBlockJobs(vm, &modern) < 0) - goto cleanup; - if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0) goto cleanup; @@ -16694,6 +16691,9 @@ qemuDomainBlockJobAbort(virDomainPtr dom, goto endjob; } + if (qemuDomainSupportsBlockJobs(vm, &modern) < 0) + goto endjob; + if (!(device = qemuDiskPathToAlias(vm, path, &idx))) goto endjob; disk = vm->def->disks[idx]; @@ -16896,9 +16896,6 @@ qemuDomainBlockJobSetSpeed(virDomainPtr dom, if (virDomainBlockJobSetSpeedEnsureACL(dom->conn, vm->def) < 0) goto cleanup; - if (qemuDomainSupportsBlockJobs(vm, &modern) < 0) - goto cleanup; - if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0) goto cleanup; @@ -16908,6 +16905,9 @@ qemuDomainBlockJobSetSpeed(virDomainPtr dom, goto endjob; } + if (qemuDomainSupportsBlockJobs(vm, &modern) < 0) + goto endjob; + if (!(device = qemuDiskPathToAlias(vm, path, NULL))) goto endjob; -- 2.3.5

2 2

[libvirt] [PATCH] iscsi: do not fail to stop a stopped pool
by Ján Tomko 30 Apr '15

30 Apr '15

Just as we allow stopping filesystem pools when they were unmounted externally, do not fail to stop an iscsi pool when someone else closed the session externally. Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1171984 --- src/storage/storage_backend_iscsi.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/src/storage/storage_backend_iscsi.c b/src/storage/storage_backend_iscsi.c index 197d333..bea6758 100644 --- a/src/storage/storage_backend_iscsi.c +++ b/src/storage/storage_backend_iscsi.c @@ -449,8 +449,13 @@ virStorageBackendISCSIStopPool(virConnectPtr conn ATTRIBUTE_UNUSED, virStoragePoolObjPtr pool) { char *portal; + char *session; int ret = -1; + if ((session = virStorageBackendISCSISession(pool, false)) == NULL) + return 0; + VIR_FREE(session); + if ((portal = virStorageBackendISCSIPortal(&pool->def->source)) == NULL) return -1; -- 2.0.5

3 8

[libvirt] RFC: a couple ideas regarding selinux/DAC labeling
by Cole Robinson 30 Apr '15

30 Apr '15

Hi all, libvirt's selinux/DAC labeling behavior has been a repeated source of frustration for desktop virt users. Granted much of the frustration comes from the fact that virt-manager runs as $user, talks to libvirtd running as root, which launches VMs running as qemu, and we are mixing it all with selinux :) But the flip side is that I've lost count of how many times I've heard someone complain 'why is libvirt chowning/relabeling my .iso? it's world readable' I spent some time looking at bug reports and looking at the relevant libvirt code, and I have a couple ideas for reducing user complaints in this area. The patches should be pretty simple but I haven't written them yet since I'd like other opinions on this stuff, given it could impact security. Here's my two suggestions: 1) Don't attempt to change selinux label for readonly/shareable images, if the image already has a label that svirt_t can access. Background: svirt is about giving r/w disk images a unique selinux label tied to one running VM, so that if other VMs are compromised, they can't access another VMs disk image. Maintaining this guarantee is important, which is why libvirt will raise an explicit error if attempting to change the disk image label errors with EPERM and selinux==Enforcing. Makes sense. The problem is with readonly media. We don't give a unique label to RO media, we just give it plain virt_content_t. RO media labeling isn't about VM isolation, it's about ensuring the svirt VM can even access the media. However, if labeling fails with EPERM, we still fail in the same way mentioned above. Here's a situation where that is problematic: - Running as qemu:///session (unprivileged) - Point a VM at /path/to/my/isos/foo.iso - foo.iso is: owner=root, mode=644, label=public_content_t. We are sharing this media for qemu:///system usage as well Without any libvirt label changing, the VM should be able to access that ISO: permissions are fine, and svirt_t can access public_content_t. However the unprivileged user does not have permissions to change the label to virt_content_t, libvirt fails to set the label, raises the error, VM startup fails. I suggest that in the case of readonly (and maybe shareable which uses the plain svirt_image_t label) disk, we first check to see if svirt_t can access the existing label, and if so, don't do any explicit labeling. This makes things 'just work' in the above example, and doesn't seem to have a detrimental effect on security. I don't know exactly how to do this with the selinux API but I assume there's something that reproduces what sesearch does on the CLI. Another case this helps: qemu:///system VM wants to use /dev/sr0, changes the label to virt_content_t, which can generate selinux violations from udev when media is ejected/inserted and the VM is running (I don't remember specifics here but we've had bug reports in the past). Instead it just leaves /dev/sr0 as removable_device_t, which svirt_t can already read from. (Libvirt already does something along these lines: If the labeling fails, it checks to see if the disk image _already_ has the label we just requested, and if so, it's not an error condition. But this suggestion goes a step further than that) 2) DAC shouldn't change owner if qemu will be able to access the disk. Background: My understanding of the DAC driver is it's _just_ to facilitate running qemu as non-root; we chown the image to ensure the emulator has permissions to access it so that VM startup doesn't fail. So I suggest that if libvirt determines the disk is already accessible by the qemu user, we skip the chown step. Coupled with the above selinux erroring mentioned above, unnecessary chowning can cause issues when sharing media between qemu:///system and qemu:///session; once the file is owned by qemu, qemu:///session can't change the selinux label ever again, and labeling failures may mean the VM fails to start. Thankfully we have libvirt infrastructure for checking whether a file can be read by arbitrary uid:gid: virStorageFileInitAs and friends. We already use that to check the file is searchable at qemu startup time, via: qemuProcessState-> qemuDomainCheckDiskPresence-> qemuDomainDetermineDiskChain-> virStorageFileGetMetadata-> virStorageFileGetMetadataRecurse So I suggest we add a check along those lines in security_dac.c, and if the file is already accessible for the disk's needs, we skip the chown'ing. Thoughts/comments appreciated. Thanks, Cole

3 3