February 2018 - Devel - libvirt List Archives

[libvirt] ANNOUNCE: libguestfs 1.38 released
by Richard W.M. Jones 09 Feb '18

09 Feb '18

I'm pleased to announce libguestfs 1.38, a library and a set of tools for accessing and modifying virtual machine disk images. This release represents about a year of work by many contributors. I'd like to call out in particular substantial contributions from: Cédric Bosdonnat, Pavel Butsykin, Matteo Cafasso, Tomáš Golembiovský, Nikos Skalkotos, and Pino Toscano. Virt-builder-repository is a new tool for creating virt-builder repositories. Virt-rescue has been rewritten, implementing implementing job control, -m and -i options, escape keys. Virt-v2v has several new methods to pull VMs out of VMware faster. The inspection code was rewritten and placed inside the daemon making it much faster and more robust. Of course dozens of other features have been added, and many bugs fixed. See the release notes below for full details. You can get libguestfs 1.38 from here: Main website: http://libguestfs.org/ [not updated yet] Source: http://libguestfs.org/download/1.38-stable/ Fedora: https://koji.fedoraproject.org/koji/packageinfo?packageID=8391 Debian/experimental: https://packages.debian.org/libguestfs0 Note for distro packagers: ocaml >= 4.01, ocaml-hivex are now mandatory build dependencies. If using glibc >= 2.27 which dropped support for SunRPC and crypt(3), you will need rpcgen, libtirpc and libxcrypt. Rich. ---------------------------------------------------------------------- Release notes for libguestfs 1.38 These are also available online at: http://libguestfs.org/guestfs-release-notes.1.html New features New tools Virt-builder-repository is a new tool allowing end users to create and update virt-builder repositories (Cédric Bosdonnat). Virt-rescue (while not a new tool) has been substantially rewritten, implementing job control, -m and -i options, escape keys, etc. New features in existing tools Virt-builder planner has been improved so that faster and more efficient build plans are chosen for complex cases, especially when either the tmpdir or output is on networked storage. New virt-builder Fedora templates (starting with Fedora 26) will have plain partition layout and use GPT for partitions. Virt-customize "firstboot" scripts in guests using systemd are now installed under the "multi-user.target" instead of "default.target" so they will only run when the system is booted normally. Virt-customize now sets a random /etc/machine-id for Linux guests, if one is not already set. Virt-df now works correctly on filesystems with block sizes smaller than 1K (Nikolay Ivanets). Virt-dib has further compatibility enhancements with diskimage-builder (Pino Toscano). Virt-sysprep removes "DHCP_HOSTNAME" from ifcfg-* files. Virt-sysprep now works on Oracle Linux (Jamie Iles). Virt-resize now correctly copies GPT partition attributes from the source to the destination (Cédric Bosdonnat). Bash tab completion implemented or enhanced for: virt-win-reg, virt-v2v-copy-to-local. virt-v2v and virt-p2v Virt-v2v can now read VMware VMX files directly, either from local disk, NFS storage, or over SSH from an ESXi hypervisor. Virt-v2v can now use VDDK as an input source. Both virt-v2v and virt-p2v are now able to pass through the source CPU vendor, model and topology. However unfortunately not all source and target hypervisors are able to provide or consume this data at present (Tomáš Golembiovský). Virt-v2v now supports encrypted guests (Pino Toscano). Virt-v2v can now handle VMware snapshots. Note that the snapshots are collapsed — it does not convert the chain of snapshots into a chain of snapshots. Virt-v2v now installs Windows 10 / Windows Server 2016 virtio block drivers correctly (Pavel Butsykin, Kun Wei). Virt-v2v now installs virtio-rng, balloon and pvpanic drivers, and correctly sets this in the target hypervisor metadata for hypervisors which support that (Tomáš Golembiovský). Virt-v2v now installs both legacy and modern virtio keys in the Windows registry (Ladi Prosek). Virt-p2v can now preserve (in some cases) the offset of the Real Time Clock from UTC. Virt-p2v now combines several scp commands to the conversion server into a single command, improving conversion times. Virt-v2v now detects the special Linux Xen PV-only kernels correctly (Laszlo Ersek). Virt-v2v -o glance now generates the right properties for UEFI guests (Pino Toscano). Virt-v2v -o null now avoids spooling the guest to a temporary file, instead it writes to the qemu "null block device". This makes it faster and use almost no disk space. Virt-v2v -o rhv now supports Windows 2016 Server guest type. Virt-v2v -i libvirtxml can now open network disks over http or https. Virt-v2v will now give a warning about host passthrough devices (Pino Toscano). The virt-v2v --machine-readable output has been enhanced so it includes "vcenter-https", "xen-ssh" and "in-place" facts (Pino Toscano). Language bindings Fix multiple memory leaks and other data corruption problems in the Java bindings (Pino Toscano). Perl %guestfs_introspection has been dropped. Inspection Inspection support was rewritten in OCaml and included inside the daemon. This makes inspection considerably faster, more robust and more easily extensible in future. Better icon support for ALT Linux guests (Pino Toscano). Better support for NeoKylin (Qingzheng Zhang). Can handle OSes like Void Linux which do not include "VERSION_ID" in /etc/os-release (Pino Toscano). Add support for Microsoft MS-DOS (Daniel Berrangé). Architectures and platforms Multiple fixes for S/390 architecture. Libguestfs and all the tools should now compile and run on this architecture. Other The libguestfs API is now thread-safe (although not parallel). You can call APIs on the same handle from multiple threads without needing to take a lock. Security There were multiple vulnerabilities in the icoutils "wrestool" program which is run by libguestfs to create icons for Windows guests. Using the latest "wrestool" is recommended. API New APIs "hivex_value_string" This replaces the deprecated "hivex_value_utf8" API, but does the same thing. "part_get_gpt_attributes" "part_set_gpt_attributes" Read and write GPT partition attribute flags (Cédric Bosdonnat). "part_resize" Enlarge or shrink an existing partition (Nikos Skalkotos). "yara_destroy" "yara_load" "yara_scan" Support for the Yara malware scanning engine (Matteo Cafasso). Other API changes APIs implemented in the daemon can now be written in either C or OCaml. Several APIs were rewritten in OCaml, although we are not planning to rewrite all of them. You will now get a clear error message if you try to add too many disks to the appliance, instead of getting a peculiar failure from qemu. Certain APIs accidentally allowed you to use "/dev/urandom" as an input "device", eg. "g.copy_device_to_device("/dev/urandom", "/dev/sda")". The code has been modified to forbid this usage. All APIs for inspecting installer CDs have been deprecated. Use libosinfo for this task. Build changes A working OCaml compiler ≥ 4.01 is now required for building libguestfs. The "./configure --disable-ocaml" option remains but is only used to disable the OCaml language bindings. Add "RELEASES" file which lists release dates for each version of libguestfs. You must update this file when making a new release. Documentation generated by "gtk-doc" has been removed. "./configure --enable-gtk-doc" now does nothing. Libtirpc is now used for XDR functions and rpcgen. Note that glibc has deprecated and in most Linux distros dropped these, so for most people this will be an extra dependency (Martin Kletzander). Libxcrypt is now used for crypt(3). This is required if using glibc ≥ 2.27. "ocaml-hivex" is now required. Libvirt ≥ 1.2.20 is now required. There is now a "make check-root" target for tests which need to be run as root (analogous to "make check-slow"). "./configure"-time check for "__attribute__((cleanup))" now works in the cross-compilation case (Yann E. Morin). The "AUTHORS" and "p2v/about-authors.c" files are now generated from a single place. Either GnuPG v1 or v2 can be used. "./configure --with-guestfs-path" may be used to set the default "LIBGUESTFS_PATH". In addition the way that the path is searched has changed slightly so that all types of appliances are searched in each path element separately (Pavel Butsykin). "GUESTFSD_EXT_CMD" which was used to mark external commands in the daemon has been removed. It was originally used by SUSE builds, but they have not been using it for a while. The output from "./configure" is now visually grouped under headings related to what it is doing, making it much easier to scan (Pino Toscano). OCaml dependencies are now generated from a single script instead of multiple not-quite-the-same Makefile fragments. "./configure --with-distro=ID" can be used to override automatic Linux distro detection at build time (Pino Toscano). qemu ≥ 2.10 is supported (but not required). This adds mandatory locking to disks and libguestfs turns this off in certain circumstances when it is known to be safe (Lars Seipel, Peter Krempa, Daniel Berrangé, Pino Toscano, Fam Zheng, Yongkui Guo, Václav Kadlčík). Internals Most common code has been moved to the common/ subdirectory, with OCaml common code being in common/ml* directories (eg. common/visit and common/mlvisit contain the visitor library in C and OCaml respectively). The mllib directory has been deleted and replaced by common/mltools. There is now a lightweight OCaml binding for PCRE, see common/mlpcre. Use of OCaml "Str" library has been mostly replaced with PCRE. Add more calls to "udev_settle" to improve stability of partition code (Dawid Zamirski). Run "udev_settle" with --exit-if-exists option, which improves the speed of this command (Pavel Butsykin). Detect new locations of major(3), minor(3), makedev(3). Actions can now be deprecated with no suggested replacement, for APIs such as "guestfs_wait_ready" that should simply be removed from client code. Use gnulib "set_nonblocking_flag" wrapper instead of calling fcntl(2) with "O_NONBLOCK" (Eric Blake). Similarly "set_cloexec_flag". Fix memory leak in XFS version of "guestfs_vfs_minimum_size" (Pino Toscano). Valgrind checks now run on the virt-p2v binary. Unicode single quotes ("‘’") and now used in place of '' or `' throughout the code and documentation. Similarly for "’s" instead of "'s". The "is_zero" function has been reimplemented for greater speed (Eric Blake). In the direct backend, virtio-blk support has been removed. Virtio- scsi is now the only supported way to add disks. Generator string parameter and return types have been rationalised so there are only two types ("String", "StringList") with many subtypes eg. "FileIn" becomes "String (FileIn, ...)". The appliance disk image can now be in formats other than raw (Pavel Butsykin). Multiple improvements to how we automatically build Debian templates for virt-builder (Pino Toscano). Enable serial console for these templates (Florian Klink). In the daemon, instead of making a private copy of lvm.conf and modifying it (eg for filters), start with an empty file since LVM understands that to mean "all defaults" (Alasdair Kergon, Zdenek Kabelac). The "direct" backend can now run QMP queries against the QEMU binary, enhancing the kinds of information we can detect. In addition the code to query QEMU has been made more robust for handling multiple parallel queries of different versions of QEMU. OCaml Augeas bindings are bundled under common/mlaugeas. The long term plan is to remove this and use system ocaml-augeas when it is more widely available in distros (Pino Toscano). All OCaml modules ("*.ml" files) are now required to have an interface file ("*.mli"). If they don't export anything then the interface will be empty except for comments. Certain OCaml features in OCaml ≥ 4.01 are used throughout the code, including replacing ‘{ field = field }’ with ‘{ field }’. Virt-builder "make-template" utility now uses the "virt-install --transient" option so that we should never need to clean up left over domains after a crash. It also saves kickstarts and virt-install commands, which are committed to git for future reference. /dev/shm is now created in the appliance (Nicolas Hicher). In verbose mode on Fedora guests, virt-customize will now use "dnf --verbose" enabling better debugging output. Virt-v2v input and output classes now contain a "#precheck" method which is used to perform environmental checks before conversion starts. Virt-p2v enables miniexpect debugging. It is written to stderr (of virt-p2v). Virt-v2v free space checks are more liberal especially for smaller guests (Pino Toscano). Bugs fixed https://bugzilla.redhat.com/1540535 Example URI of "Convert from ESXi hypervisor over SSH to local libvirt" is incorrect in v2v man page https://bugzilla.redhat.com/1539395 virt-customize segfaults after upgrading to 1.37.35-3 https://bugzilla.redhat.com/1536765 Libguestfs Perl bindings can leak a small amount of memory on error https://bugzilla.redhat.com/1536763 libguestfs Lua bindings use strerror(), which isn’t thread safe https://bugzilla.redhat.com/1536603 man page makes no mention of using '--' when trying to change exit on error behavior https://bugzilla.redhat.com/1525241 virt-df displays zeros for filesystems with block size =512 https://bugzilla.redhat.com/1519204 v2v should improve the result when convert a rhel7.4 guest with no available kernels found in the bootloader https://bugzilla.redhat.com/1518517 virt-v2v fails with "unsupported configuration: shared access for disk 'sdb' requires use of supported storage format" https://bugzilla.redhat.com/1516094 Mere presence of QEMU file locking options breaks NBD (Block protocol 'nbd' doesn't support the option 'locking') https://bugzilla.redhat.com/1514756 ./configure --disable-ocaml breaks building common/mlpcre which breaks building daemon https://bugzilla.redhat.com/1513884 [RFE]Should update some vddk info in v2v man page https://bugzilla.redhat.com/1508874 virt-v2v: warning: ova disk has an unknown VMware controller type (20) https://bugzilla.redhat.com/1506572 virt-v2v '-i ova' is not parsing the MAC address from the source OVF https://bugzilla.redhat.com/1506511 virt-builder fails to parse repo file if it has blank space after the repository identifier https://bugzilla.redhat.com/1503958 Failed to convert the rhel5 guest with kmod-xenpv installed from xen server by virt-v2v https://bugzilla.redhat.com/1503497 qemu-kvm fails to open qcow2 files in read-only mode with qemu-kvm 1.5.3 https://bugzilla.redhat.com/1500673 Error info shows wrong sometimes when ssh to conversion server using non-root user with sudo on p2v client https://bugzilla.redhat.com/1500537 /dev/shm does not exist in the appliance environment https://bugzilla.redhat.com/1497475 guestfish cannot list commands from interactive mode https://bugzilla.redhat.com/1493048 Unbound constructor Hivex.OPEN_UNSAFE https://bugzilla.redhat.com/1484957 bump debian images to use single-partition layout https://bugzilla.redhat.com/1482737 virt-resize failed to expand swap partition for RHEL5.11 guest image with "parsing UUID failed" https://bugzilla.redhat.com/1477623 Running file API on a special chardev may hang forever https://bugzilla.redhat.com/1476081 inspect-os report error: could not parse integer in version number: V7Update2 https://bugzilla.redhat.com/1472719 [RFE]Add warning in process of v2v converting guest which has pci passthrough device https://bugzilla.redhat.com/1472208 virt-v2v fails on opensuse 13.2 guest with error: statns: statns_stub: path must start with a / character https://bugzilla.redhat.com/1469655 firstboot scripts are not correctly installed in Fedora 26 https://bugzilla.redhat.com/1466563 Libguestfs should pass copyonread flag through to the libvirt XML https://bugzilla.redhat.com/1465665 1.36.x build failure: gtkdocize fails using newer autotools due to missing GTK_DOC_CHECK in configure.ac https://bugzilla.redhat.com/1460338 guestfs_shutdown hangs if main process sets signal handlers https://bugzilla.redhat.com/1459979 guestfs_add_domain_argv fails with readonly option when vdi/vhd disk is attached to libvirt domain https://bugzilla.redhat.com/1451665 RFE: Virt-v2v can't convert the guest which has encrypted partition https://bugzilla.redhat.com/1450325 document URI format for -a parameters of tools https://bugzilla.redhat.com/1448739 RFE: Support multicore decompression for OVA files using pigz and pxz https://bugzilla.redhat.com/1447202 Win 2016 guest is described as Win 10 after imported to RHEVM https://bugzilla.redhat.com/1441197 RFE: ability to convert VMware virtual machines via vmx https://bugzilla.redhat.com/1438939 Please drop or update GnuPG (1.4.x) dependency https://bugzilla.redhat.com/1438794 [RFE] Install Windows virtio-rng drivers on VMs imported https://bugzilla.redhat.com/1433937 virt-inspector can't get icon info from altlinux-centaurus https://bugzilla.redhat.com/1433577 policycoreutils setfiles >= 2.6 does .. nothing https://bugzilla.redhat.com/1431579 Windows 8 UEFI from VMware to KVM fails to boot after conversion https://bugzilla.redhat.com/1430680 There is error info about "No such file or directory" when convert a guest from ova file by v2v https://bugzilla.redhat.com/1430184 virt-dib should generate sha256 checksum instead of sha512 https://bugzilla.redhat.com/1429506 RFE: OVMF should be detected on conversion server to prevent failed conversion https://bugzilla.redhat.com/1429491 Should rename network name of rhv in virt-v2v man page https://bugzilla.redhat.com/1427529 virt-sysprep should remove DHCP_HOSTNAME https://bugzilla.redhat.com/1417306 QEMU image file locking (libguestfs) https://bugzilla.redhat.com/1409024 [Debian] Missing db_dump abort inspection https://bugzilla.redhat.com/1406906 Segmentation fault when reading corrupted path with Python 3 bindings https://bugzilla.redhat.com/1379289 RFE: virt-p2v should support mnemonic operations https://bugzilla.redhat.com/1378022 There is virt-v2v warning about <listen type='none'> during converting a guest which has listen type='none' in XML https://bugzilla.redhat.com/1376547 qemu-system-s390x: -device isa-serial,chardev=charserial0,id=serial0: 'isa-serial' is not a valid device model name https://bugzilla.redhat.com/1374232 selinux relabel fails on RHEL 6.2 guests with "libguestfs error: selinux_relabel: : Success" https://bugzilla.redhat.com/1367738 Missing bash completion scripts for: virt-diff guestunmount virt- copy-in virt-copy-out virt-customize virt-get-kernel virt-p2v-make-disk virt-p2v-make-kickstart virt-tar-in virt-tar-out virt-v2v-copy-to-local virt-win-reg https://bugzilla.redhat.com/1362649 RFE: virt-sysprep does not utilize libguestfs encryption support https://bugzilla.redhat.com/1172425 [RFE]virt-v2v failed to convert VMware ESX VM with snapshot https://bugzilla.redhat.com/1171654 Modify a file in virt-rescue with vi on some linux terminal such as yakuake, can lead to abnormal display in virt-rescue shell https://bugzilla.redhat.com/1167623 Remove "If reporting bugs, run virt-v2v with debugging enabled .." message when running virt-p2v https://bugzilla.redhat.com/1152819 Can not end a running command in virt-rescue by press ^C or other keys, the only way is to exit virt-rescue -- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones Read my programming and virtualization blog: http://rwmj.wordpress.com virt-top is 'top' for virtual machines. Tiny program with many powerful monitoring features, net stats, disk stats, logging, etc. http://people.redhat.com/~rjones/virt-top

1 0

[libvirt] [PATCH] qemu: Fix indentation in qemuDomainDeviceDefValidateControllerPCI()
by Andrea Bolognani 09 Feb '18

09 Feb '18

Signed-off-by: Andrea Bolognani <abologna(a)redhat.com> --- Pushed as trivial. src/qemu/qemu_domain.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c index 84207db16..16833474a 100644 --- a/src/qemu/qemu_domain.c +++ b/src/qemu/qemu_domain.c @@ -4458,10 +4458,10 @@ qemuDomainDeviceDefValidateControllerPCI(const virDomainControllerDef *controlle if (pciopts->numaNode != -1 && !virQEMUCapsGet(qemuCaps, QEMU_CAPS_SPAPR_PCI_HOST_BRIDGE_NUMA_NODE)) { - virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s", - _("the spapr-pci-host-bridge controller doesn't " - "support numa_node in this QEMU binary")); - return -1; + virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s", + _("the spapr-pci-host-bridge controller doesn't " + "support numa_node in this QEMU binary")); + return -1; } break; -- 2.14.3

1 0

[libvirt] [PATCH] qemu: fix migration with local and VIR_STORAGE_TYPE_NETWORK disks
by Chris Friesen 09 Feb '18

09 Feb '18

In the current implementation of qemuMigrateDisk() the value of the "nmigrate_disks" parameter wrongly impacts the decision whether or not to migrate a disk that is not a member of "migrate_disks": 1) If "nmigrate_disks" is zero, "disk" is migrated if it's non-shared non-readonly with source. 2) If "nmigrate_disks" is non-zero and "disk" is not a member of "migrate_disks" then "disk" is not migrated. This should instead proceed with checking conditions as per 1) and allow migration of non-shared non-readonly disks with source. Fixing 2) breaks migration of VMs with a mix of rbd and local disks because now libvirt tries to migrate the rbd root disk and it fails. This new problem is solved by updating 1) to factor in disk source type and migrate only 'local' non-shared non-readonly disks with source. The end result is that disks not in "migrate_disks" are treated uniformly regardless of the value of "nmigrate_disks". Signed-off-by: Chris Friesen <chris.friesen(a)windriver.com> --- src/qemu/qemu_migration.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/src/qemu/qemu_migration.c b/src/qemu/qemu_migration.c index 5ee9e5c..77fafc6 100644 --- a/src/qemu/qemu_migration.c +++ b/src/qemu/qemu_migration.c @@ -409,12 +409,12 @@ qemuMigrateDisk(virDomainDiskDef const *disk, if (STREQ(disk->dst, migrate_disks[i])) return true; } - return false; } - /* Default is to migrate only non-shared non-readonly disks + /* Default is to migrate only non-shared non-readonly local disks * with source */ return !disk->src->shared && !disk->src->readonly && + (disk->src->type != VIR_STORAGE_TYPE_NETWORK) && !virStorageSourceIsEmpty(disk->src); } -- 1.8.3.1

3 8

[libvirt] [PATCH] qemu: command: Extract formatting of floppy related stuff into a helper
by Peter Krempa 09 Feb '18

09 Feb '18

The floppy command formatting is special-cased since it does not directly translate to a single '-device' argument. Move the code from qemuBuildDiskDriveCommandLine to a new helper function so that all the related code is together. --- src/qemu/qemu_command.c | 91 +++++++++++++++++++++++++++++-------------------- 1 file changed, 54 insertions(+), 37 deletions(-) diff --git a/src/qemu/qemu_command.c b/src/qemu/qemu_command.c index 24b434a458..3857a5ef5b 100644 --- a/src/qemu/qemu_command.c +++ b/src/qemu/qemu_command.c @@ -2211,6 +2211,58 @@ qemuBuildDriveDevStr(const virDomainDef *def, } +static int +qemuBulildFloppyCommandLineOptions(virCommandPtr cmd, + const virDomainDef *def, + virDomainDiskDefPtr disk, + unsigned int bootindex) + +{ + virBuffer fdc_opts = VIR_BUFFER_INITIALIZER; + char *fdc_opts_str = NULL; + char *optstr; + + if (virAsprintf(&optstr, "drive%c=drive-%s", + disk->info.addr.drive.unit ? 'B' : 'A', + disk->info.alias) < 0) + return -1; + + if (!qemuDomainNeedsFDC(def)) { + virCommandAddArg(cmd, "-global"); + virCommandAddArgFormat(cmd, "isa-fdc.%s", optstr); + } else { + virBufferAsprintf(&fdc_opts, "%s,", optstr); + } + VIR_FREE(optstr); + + if (bootindex) { + if (virAsprintf(&optstr, "bootindex%c=%u", + disk->info.addr.drive.unit + ? 'B' : 'A', + bootindex) < 0) + return -1; + + if (!qemuDomainNeedsFDC(def)) { + virCommandAddArg(cmd, "-global"); + virCommandAddArgFormat(cmd, "isa-fdc.%s", optstr); + } else { + virBufferAsprintf(&fdc_opts, "%s,", optstr); + } + VIR_FREE(optstr); + } + + /* Newer Q35 machine types require an explicit FDC controller */ + virBufferTrim(&fdc_opts, ",", -1); + if ((fdc_opts_str = virBufferContentAndReset(&fdc_opts))) { + virCommandAddArg(cmd, "-device"); + virCommandAddArgFormat(cmd, "isa-fdc,%s", fdc_opts_str); + VIR_FREE(fdc_opts_str); + } + + return 0; +} + + static int qemuBuildDiskDriveCommandLine(virCommandPtr cmd, const virDomainDef *def, @@ -2220,8 +2272,6 @@ qemuBuildDiskDriveCommandLine(virCommandPtr cmd, unsigned int bootCD = 0; unsigned int bootFloppy = 0; unsigned int bootDisk = 0; - virBuffer fdc_opts = VIR_BUFFER_INITIALIZER; - char *fdc_opts_str = NULL; if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_DRIVE_BOOT) || virQEMUCapsGet(qemuCaps, QEMU_CAPS_BOOTINDEX)) { @@ -2300,34 +2350,9 @@ qemuBuildDiskDriveCommandLine(virCommandPtr cmd, if (qemuDiskBusNeedsDeviceArg(disk->bus)) { if (disk->bus == VIR_DOMAIN_DISK_BUS_FDC) { - if (virAsprintf(&optstr, "drive%c=drive-%s", - disk->info.addr.drive.unit ? 'B' : 'A', - disk->info.alias) < 0) + if (qemuBulildFloppyCommandLineOptions(cmd, def, disk, + bootindex) < 0) return -1; - - if (!qemuDomainNeedsFDC(def)) { - virCommandAddArg(cmd, "-global"); - virCommandAddArgFormat(cmd, "isa-fdc.%s", optstr); - } else { - virBufferAsprintf(&fdc_opts, "%s,", optstr); - } - VIR_FREE(optstr); - - if (bootindex) { - if (virAsprintf(&optstr, "bootindex%c=%u", - disk->info.addr.drive.unit - ? 'B' : 'A', - bootindex) < 0) - return -1; - - if (!qemuDomainNeedsFDC(def)) { - virCommandAddArg(cmd, "-global"); - virCommandAddArgFormat(cmd, "isa-fdc.%s", optstr); - } else { - virBufferAsprintf(&fdc_opts, "%s,", optstr); - } - VIR_FREE(optstr); - } } else { virCommandAddArg(cmd, "-device"); @@ -2339,14 +2364,6 @@ qemuBuildDiskDriveCommandLine(virCommandPtr cmd, } } } - /* Newer Q35 machine types require an explicit FDC controller */ - virBufferTrim(&fdc_opts, ",", -1); - if ((fdc_opts_str = virBufferContentAndReset(&fdc_opts))) { - virCommandAddArg(cmd, "-device"); - virCommandAddArgFormat(cmd, "isa-fdc,%s", fdc_opts_str); - VIR_FREE(fdc_opts_str); - } - return 0; } -- 2.15.0

2 1

[libvirt] [tck PATCH v2 0/5] Fix libvirt-tck nwfilter tests
by Laine Stump 09 Feb '18

09 Feb '18

I've pushed the patches that passed review. These patches split a single patch from V1 into 2 patches, repost one patch that had no response, calls out a single line that had sneaked into one of the ACKed patches, and remove all tabs from perl and sh scripts. Laine Stump (5): parameterize filterref in generic_domain default to no nwfilter in test domains filter for proper IP address in tcpdump looking for broadcast packet kill dhclient before attempting to change guest IP in no-ip-spoofing test untabify all perl and sh files lib/Sys/Virt/TCK.pm | 549 ++++++++++++----------- lib/Sys/Virt/TCK/Capabilities.pm | 84 ++-- lib/Sys/Virt/TCK/DomainBuilder.pm | 312 ++++++------- lib/Sys/Virt/TCK/NetworkBuilder.pm | 44 +- lib/Sys/Virt/TCK/NetworkHelpers.pm | 6 +- lib/Sys/Virt/TCK/SELinux.pm | 6 +- lib/Sys/Virt/TCK/StoragePoolBuilder.pm | 40 +- lib/Sys/Virt/TCK/StorageVolBuilder.pm | 48 +- lib/Sys/Virt/TCK/TAP/XMLFormatter.pm | 18 +- lib/Sys/Virt/TCK/TAP/XMLFormatterSession.pm | 66 +-- scripts/nwfilter/100-ping-still-working.t | 3 +- scripts/nwfilter/210-no-mac-spoofing.t | 3 +- scripts/nwfilter/220-no-ip-spoofing.t | 4 +- scripts/nwfilter/230-no-mac-broadcast.t | 5 +- scripts/nwfilter/240-no-arp-spoofing.t | 3 +- scripts/nwfilter/300-vsitype.t | 2 +- scripts/nwfilter/concurrency/chg-vm-filter.sh | 18 +- scripts/nwfilter/concurrency/start-destroy-vm.sh | 20 +- scripts/nwfilter/nwfilter2vmtest.sh | 42 +- 19 files changed, 641 insertions(+), 632 deletions(-) -- 2.14.3

4 19

[libvirt] [PATCH] rpc: export xdr_virNetMessageError for virnetmessagetest on Win32
by Daniel P. Berrangé 09 Feb '18

09 Feb '18

The Win32 symbol export file format can't do wildcards, so none of the 'xdr_*' symbols are exported from the libvirt DLL. This doesn't matter generally since the RPC client is built into the DLL and we don't build libvirtd on Win32. The virnetmessagetest, however, does require xdr_virNetMessageError to be exported, so just do a hack for that. Signed-off-by: Daniel P. Berrangé <berrange(a)redhat.com> --- Pushed as win32 CI build fix src/libvirt_remote.syms | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/src/libvirt_remote.syms b/src/libvirt_remote.syms index 736848273a..97e22275b9 100644 --- a/src/libvirt_remote.syms +++ b/src/libvirt_remote.syms @@ -7,6 +7,12 @@ # Generated files xdr_*; +# Hack only required for Win32 to make tests link properly, +# as the wildcard above doesn't work there. Fortunately +# Win32 builds don't require the other xdr_* syms exported. +# since there's no libvirtd built. +xdr_virNetMessageError; + # rpc/virnetclient.h virNetClientAddProgram; -- 2.14.3

1 0

[libvirt] [PATCH 0/2] Support network stats for hostdev(SR-IOV) in Switchdev mode
by Jai Singh Rana 09 Feb '18

09 Feb '18

With availability of switchdev model in linux, it is possible to capture stats for hostdev SR-IOV VFs using its VF representor interface name on host for nics supporting switchdev model. These stats are supported by adding helper APIs for getting VF Representor name based on BDF info in 'hostdev' and querying required net sysfs entries on host. These helper APIs are then used in qemu_driver to get the hostdev interface stats for pci SR-IOV device. [1] https://www.kernel.org/doc/Documentation/networking/switchdev.txt Jai Singh Rana (2): util: Add helper APIs to get/verify VF Representor name qemu: conf: Network stats support for hostdev VF Representor po/POTFILES.in | 1 + src/Makefile.am | 1 + src/conf/domain_conf.c | 7 ++ src/libvirt_private.syms | 5 + src/qemu/qemu_driver.c | 34 +++++- src/util/virhostdev.c | 10 ++ src/util/virhostdev.h | 6 + src/util/virnetdevhostdev.c | 284 ++++++++++++++++++++++++++++++++++++++++++++ src/util/virnetdevhostdev.h | 33 +++++ 9 files changed, 377 insertions(+), 4 deletions(-) create mode 100644 src/util/virnetdevhostdev.c create mode 100644 src/util/virnetdevhostdev.h -- 2.13.6

1 2

[libvirt] [PATCH python] tests: add special case for virStoragePoolLookupByTargetPath
by Daniel P. Berrangé 09 Feb '18

09 Feb '18

This new API trips up the sanity test for checking mapping of all C APIs into Python. Signed-off-by: Daniel P. Berrangé <berrange(a)redhat.com> --- Pushed as build fix. sanitytest.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/sanitytest.py b/sanitytest.py index 1d65e8d..190d32b 100644 --- a/sanitytest.py +++ b/sanitytest.py @@ -258,7 +258,8 @@ for name in sorted(basicklassmap): "LookupByUUIDString", "LookupByVolume" "LookupByName", "LookupByID", "LookupByName", "LookupByKey", "LookupByPath", "LookupByMACString", "LookupByUsage", "LookupByVolume", - "LookupSCSIHostByWWN", "Restore", "RestoreFlags", + "LookupByTargetPath","LookupSCSIHostByWWN", + "Restore", "RestoreFlags", "SaveImageDefineXML", "SaveImageGetXMLDesc", "DefineXMLFlags"]: if klass != "virDomain": func = klass[3:] + func -- 2.14.3

1 0

[libvirt] [PATCH 0/3] lock manager plugin dlm-corosync implementation
by river 09 Feb '18

09 Feb '18

This is the implementation of the RFC from https://www.redhat.com/archives/libvir-list/2017-December/msg00689.html The following is the design of this iplementation. Environmental requirements: DLM is a symmetric general-purpose distributed lock manager, consisting of a kernel modules named 'dlm.ko', userspace application named 'libdlm', and cluster infrastructure environment which could be 'corosync' responsible for communication. For more usage about libdlm and corosync, `man corosync.conf` and `man dlm_controld` Implementation details: In order to use dlm in cluster, daemons, 'dlm_controld' and 'corosync', must have been runned before. Because dlm is asynchronous, there is something receiving notification from kernel. 'libdlm' provides `dlm_pthread_init` and `dlm_ls_pthread_init` for this purpose. Here is lockspace concept in dlm. Lockspace is different from sanlock, it's just to provide a private namespaceit's one namespace for locks that are part of a single application. One lock must belong to one lockspace, and is associated with one lock resource, owned by one process in one node. Lock/unlock need specific flag. Lock flags `LKF_PERSISTENT` specifies a lock that wouldn't be unlocked if the process was dead or lockspace was closed (`dlm_close_lockspace`). lockspace is cluster-wide, lock with `LKF_PERSISTENT` flag won't be automatically released except that the owner node reboot, but it will become an orphan lock once process disappears. New process in the same node want to adopt orphan lock, just specific `LKF_ORPHAN` when re-lock using the right lock-name and lock-mode, new lock id would be assigned for adopted lock. lkid(lock id) is a mark uniquely identifing a lock in a process, which is returned after locking sucessfully or adopting sucessfully. Lock could be only released by self-process applied using lkid. The following means: locks won't be automatically released when the process(libvirt) that acquired them dies, daemon likes virtlockd isn't necessary. In order to adopt orphan locks, there must be some lock information left after the process died. A file would store the lock information in this design. However, I find that the existence of "leaked" fd in vm instance process inherited from libvirt would tell lock daemon to release their locks instead of libvirt. So destroy or shutdown qemu instance won't make dlm release locks. I refer to the action of libxl(`libxlDomainCleanup` function in 'src/libxl/libxl_domain.c'), add some code after unlink pid and xml files in `qemuProcessStop` ('src/qemu/qemu_process.c') to proactively tell lock manager release locks. Adhere to the simple, enough principle, lock information using by running libvirt process is stored in list instead of hash table. I have tested in some cases: start, migrate, destroy... It works well. *** BLURB HERE *** river (3): Add a plugin dlm for lock manager Fix release of locks with dlm plugin Add configuration files for dlm plugin .gitignore | 2 + configure.ac | 6 + m4/virt-cpg.m4 | 37 ++ m4/virt-dlm.m4 | 36 ++ src/Makefile.am | 53 ++ src/libxl/libxl.conf | 8 +- src/locking/dlm.conf | 52 ++ src/locking/libvirt_dlm.aug | 35 ++ src/locking/lock_driver_dlm.c | 1056 +++++++++++++++++++++++++++++++++++ src/locking/test_libvirt_dlm.aug.in | 9 + src/qemu/qemu.conf | 8 +- src/qemu/qemu_process.c | 5 + src/util/virlist.h | 110 ++++ 13 files changed, 1410 insertions(+), 7 deletions(-) create mode 100644 m4/virt-cpg.m4 create mode 100644 m4/virt-dlm.m4 create mode 100644 src/locking/dlm.conf create mode 100644 src/locking/libvirt_dlm.aug create mode 100644 src/locking/lock_driver_dlm.c create mode 100644 src/locking/test_libvirt_dlm.aug.in create mode 100644 src/util/virlist.h -- 2.15.1

3 6

[libvirt] [PATCH] fix regex to check CN from server certificate
by Tiago M. Vieira 08 Feb '18

08 Feb '18

Currently when the script validates the PKI files and the certificate 'Subject:' field contains RDNs after the Common Name (CN), these values are also included, creating a false result that the CN is not correct. A small change to the sed regex fixes this issue, by extracting only the value for CN and nothing else. The regex is replaced with the exact same regex used to extract the CN value from the client certificate. --- tools/virt-pki-validate.in | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/virt-pki-validate.in b/tools/virt-pki-validate.in index 206637abf..b04680dde 100755 --- a/tools/virt-pki-validate.in +++ b/tools/virt-pki-validate.in @@ -255,7 +255,7 @@ then echo CA organization: $ORG echo Server organization: $S_ORG fi - S_HOST=`"$CERTOOL" -i --infile "$LIBVIRT/servercert.pem" | grep Subject: | sed 's+.*CN=$[a-zA-Z\. _-]*$+\1+'` + S_HOST=`"$CERTOOL" -i --infile "$LIBVIRT/servercert.pem" | grep Subject: | sed 's+.*CN=$.[a-zA-Z \._-]*$.*+\1+'` if test "$S_HOST" != "`hostname -s`" && test "$S_HOST" != "`hostname`" then echo The server certificate does not seem to match the host name -- 2.14.3

2 1