June 2017 - Devel - libvirt List Archives

[libvirt] [PATCH RFC V2 0/2] Implement l3 CAT
by Eli Qiao 03 Jul '17

03 Jul '17

Allow user to define cachetune in domain xml. V2 -> V1 changes: * Redefine cachetune xml in domain xml. * Create a struct for driver to talk with util/virresctrl.* * Nit fixes Eli Qiao (2): Resctrl: Add new xml element to support cache tune Resctrl: Do cache allocation while boot a qemu VM docs/schemas/domaincommon.rng | 29 ++ include/libvirt/virterror.h | 1 + src/Makefile.am | 1 + src/conf/domain_conf.c | 112 ++++ src/conf/domain_conf.h | 19 + src/libvirt_private.syms | 9 + src/qemu/qemu_process.c | 81 +++ src/util/virerror.c | 1 + src/util/virresctrl.c | 822 ++++++++++++++++++++++++++++++ src/util/virresctrl.h | 88 ++++ tests/Makefile.am | 8 +- tests/virresctrldata/L3-free.schemata | 1 + tests/virresctrldata/L3CODE-free.schemata | 1 + tests/virresctrldata/L3DATA-free.schemata | 1 + tests/virresctrldata/linux-resctrl | 1 + tests/virresctrldata/linux-resctrl-cdp | 1 + tests/virresctrltest.c | 119 +++++ 17 files changed, 1294 insertions(+), 1 deletion(-) create mode 100644 src/util/virresctrl.c create mode 100644 src/util/virresctrl.h create mode 100644 tests/virresctrldata/L3-free.schemata create mode 100644 tests/virresctrldata/L3CODE-free.schemata create mode 100644 tests/virresctrldata/L3DATA-free.schemata create mode 120000 tests/virresctrldata/linux-resctrl create mode 120000 tests/virresctrldata/linux-resctrl-cdp create mode 100644 tests/virresctrltest.c -- 1.9.1

3 4

[libvirt] [PATCH] pci: add support for VMD domains
by Jon Derrick 30 Jun '17

30 Jun '17

VMD domains start at 0x10000, so expand dev->name to fit at least this many characters. Signed-off-by: Jon Derrick <jonathan.derrick(a)intel.com> --- src/util/virpci.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/util/virpci.c b/src/util/virpci.c index 2c1b758..b3afefe 100644 --- a/src/util/virpci.c +++ b/src/util/virpci.c @@ -50,7 +50,7 @@ VIR_LOG_INIT("util.pci"); #define PCI_SYSFS "/sys/bus/pci/" #define PCI_ID_LEN 10 /* "XXXX XXXX" */ -#define PCI_ADDR_LEN 13 /* "XXXX:XX:XX.X" */ +#define PCI_ADDR_LEN 14 /* "XXXXX:XX:XX.X" */ VIR_ENUM_IMPL(virPCIELinkSpeed, VIR_PCIE_LINK_SPEED_LAST, "", "2.5", "5", "8", "16") -- 2.9.3

1 0

[libvirt] [PATCH 00/10] conf: Clean up virDomain*Def creation
by Andrea Bolognani 30 Jun '17

30 Jun '17

This series is required to solve a known limitation in the current incarnation of device isolation support[1], namely the inability to isolate host devices coming from IOMMU group 0. The issue lies in the fact that virDomainDeviceInfo, and all virDomain*Def that embed it, are usually allocated through VIR_ALLOC(), which result in all their fields being initialized to zero. That's worked out just fine so far, because zero was a sensible default value for all existing fields; however, when implementing isolation groups, we add a new virDomainDeviceInfo::isolationGroup field which we need to be initialized to -1 instead so that it doesn't overlap with IOMMU group 0 mentioned above. Solving the issue involves creating twenty or so virDomain*DefNew() functions and using them instead of VIR_ALLOC() every time a virDomain*Def needs to be created, which is arguably a pretty good idea regardless. The series could probably be organized so that eg. the patch order makes more sense, but honestly I'm pretty tired right now and I just don't have it in me. Moreover, the goal of the series and its implementation are both straighforward enough that I feel reviewers will have no problem following along. My main concern with this series is that, while converting from VIR_ALLOC() to virDomain*DefNew(), I might simply have missed some instances. That would not cause any issue right away, but once we introduce isolation groups it would lead to suboptimal PCI addresses being allocated at best, and to perfectly legitimate hotplug requests being denied at worst. The fact that the test suite[2] passes gives me a lot of confidence that I haven't overlooked anything, but if anyone has clever ideas on how to be actually sure rather than merely confident please do let me know. [1] https://www.redhat.com/archives/libvir-list/2017-June/msg01018.html [2] That is, the version of it where the default isolation group being initialized correctly actually matters Andrea Bolognani (10): conf: Make virDomainDeviceGetInfo() private conf: Rename virDomainHostdevDefAlloc() to virDomainHostdevDefNew() conf: Clean up virDomainHostdevDefNew() conf: Clean up virDomain{Memory,Shmem}DefParseXML() conf: Move some virDomainDeviceInfo functions conf: Introduce virDomainDeviceInfoNew() conf: Clean up virDomainDeviceInfo functions conf, qemu: Use virDomainDeviceInfoNew() conf: Provide missing virDomain*DefNew() functions conf: Call virDomainDeviceInfoClear() in virDomain*DefNew() src/bhyve/bhyve_parse_command.c | 4 +- src/conf/device_conf.c | 146 ++++++++++++ src/conf/device_conf.h | 9 + src/conf/domain_addr.c | 6 +- src/conf/domain_conf.c | 486 ++++++++++++++++++++++++---------------- src/conf/domain_conf.h | 25 ++- src/libvirt_private.syms | 19 +- src/lxc/lxc_native.c | 2 +- src/openvz/openvz_conf.c | 2 +- src/qemu/qemu_command.c | 12 +- src/qemu/qemu_domain.c | 11 +- src/qemu/qemu_domain_address.c | 31 +-- src/qemu/qemu_hotplug.c | 5 +- src/qemu/qemu_parse_command.c | 31 +-- src/vbox/vbox_common.c | 14 +- src/vmx/vmx.c | 2 +- src/vz/vz_sdk.c | 6 +- src/xen/xen_driver.c | 2 +- src/xenapi/xenapi_driver.c | 2 +- src/xenconfig/xen_common.c | 4 +- src/xenconfig/xen_sxpr.c | 10 +- src/xenconfig/xen_xl.c | 4 +- src/xenconfig/xen_xm.c | 2 +- tests/virhostdevtest.c | 2 +- 24 files changed, 561 insertions(+), 276 deletions(-) -- 2.7.5

3 30

[libvirt] [PATCH] qemu: Don't update CPU when checking ABI stability
by Jiri Denemark 30 Jun '17

30 Jun '17

When checking ABI stability between two domain definitions, we first make migratable copies of them. However, we also asked for the guest CPU to be updated, even though the updated CPU is supposed to be already included in the original definitions. Moreover, if we do this on the destination host during migration, we're potentially updating the definition with according to an incompatible host CPU. While updating the CPU when checking ABI stability doesn't make any sense, it actually just worked because updating the CPU doesn't do anything for custom CPUs (only host-model CPUs are affected) and we updated both definitions in the same way. Less then a year ago commit v2.3.0-rc1~42 stopped updating the CPU in the definition we got internally and only the user supplied definition was updated. However, the same commit started updating host-model CPUs to custom CPUs which are not affected by the request to update the CPU. So it still seemed to work right, unless a user upgraded libvirt 2.2.0 to a newer version while there were some domains with host-model CPUs running on the host. Such domains couldn't be migrated with a user supplied XML since libvirt would complain: Target CPU mode custom does not match source host-model The fix is pretty straightforward, we just need to stop updating the CPU when checking ABI stability. https://bugzilla.redhat.com/show_bug.cgi?id=1463957 Signed-off-by: Jiri Denemark <jdenemar(a)redhat.com> --- src/qemu/qemu_domain.c | 1 - 1 file changed, 1 deletion(-) diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c index 8e7404da6..d9357f20c 100644 --- a/src/qemu/qemu_domain.c +++ b/src/qemu/qemu_domain.c @@ -5926,7 +5926,6 @@ qemuDomainMigratableDefCheckABIStability(virQEMUDriverPtr driver, #define COPY_FLAGS (VIR_DOMAIN_XML_SECURE | \ - VIR_DOMAIN_XML_UPDATE_CPU | \ VIR_DOMAIN_XML_MIGRATABLE) bool -- 2.13.2

2 1

[libvirt] [PATCH 0/3] Add setting CPU features (CPUID) with libxenlight driver.
by Marek Marczykowski-Górecki 30 Jun '17

30 Jun '17

Marek Marczykowski-Górecki (3): cpu: define sub-leaf 0 for leaf 7 in cpu_map.xml libxl: add support for CPUID features policy tests: check domain XML to libxl structures conversion src/cpu/cpu_map.xml | 58 +++++------ src/libxl/libxl_conf.c | 139 ++++++++++++++++++++----- src/libxl/libxl_conf.h | 1 +- tests/Makefile.am | 8 +- tests/libxlxmltest.c | 232 ++++++++++++++++++++++++++++++++++++++++++- 5 files changed, 383 insertions(+), 55 deletions(-) create mode 100644 tests/libxlxmltest.c base-commit: f914b3f2d24d05615095e5ab5b0c91560c4dc903 -- git-series 0.9.1

4 15

[libvirt] [RFC PATCH 0/5] hotplug: fix premature rebinding of VFIO devices to host
by Michael Roth 30 Jun '17

30 Jun '17

Hi everyone. Hoping to get some feedback on this approach, or some alternatives proposed below, to the following issue: Currently libvirt immediately attempts to rebind a managed device back to the host driver when it receives a DEVICE_DELETED event from QEMU. This is problematic for 2 reasons: 1) If multiple devices from a group are attached to a guest, this can move the group into a "non-viable" state where some devices are assigned to the host and some to the guest. 2) When QEMU emits the DEVICE_DELETED event, there's still a "finalize" phase where additional cleanup occurs. In most cases libvirt can ignore this cleanup, but in the case of VFIO devices this is where closing of a VFIO group FD occurs, and failing to wait before rebinding the device to the host driver can result in unexpected behavior. In the case of powernv hosts at least, this can lead to a host driver crashing due to the default DMA windows not having been fully-restored yet. The window between this is and the initial DEVICE_DELETED seems to be ~6 seconds in practice. We've seen host dumps with Mellanox CX4 VFs being rebound to host driver during this period (on powernv hosts). Patches 1-4 address 1) by deferring rebinding of a hostdev to the host driver until all the devices in the group have been detached, at which point all the hostdevs are rebound as a group. Until that point, the devices are traced by the drvManager's inactiveList in a similar manner to hostdevs that are assigned to VFIO via the nodedev-detach interface. Patch 5 addresses 2) by adding an additional check that, when the last device from a group is detached, polls /proc for open FDs referencing the VFIO group path in /dev/vfio/<iommu_group> and waiting for the FD to be closed. If we time out, we abandon rebinding the hostdevs back to the host. There are a couple alternatives to Patch 5 that might be worth considering: a) Add a DEVICE_FINALIZED event to QEMU and wait for that instead of DEVICE_DELETED. Paired with patches 1-4 this would let us drop patch 5 in favor of minimal changes to libvirt's event handlers. The downsides are: - that we'd incur some latency for all device-detach calls, but it's not apparent to me whether this delay is significant for anything outside of VFIO. - there may be cases where finalization after DEVICE_DELETE/unparent are is not guaranteed, and I'm not sure QOM would encourage such expectations even if that's currently the case. b) Add a GROUP_DELETED event to VFIO's finalize callback. This is the most direct solution. With this we could completely separate out the handling of rebinding to host driver based on receival of this event. The downsides are: - this would only work for newer versions of QEMU, though we could use the poll-wait in patch 5 as a fallback. - synchronizing sync/async device-detach threads with sync/async handlers for this would be a bit hairy, but I have a WIP in progress that seems *fairly reasonable* c) Take the approach in Patch 5, either as a precursor to implementing b) or something else, or just sticking with that for now. d) ??? Personally I'm leaning toward c), but I'm wondering if that's "good enough" for now, or if we should pursue something more robust from the start, or perhaps something else entirely? Any feedback is greatly appreciated! src/libvirt_private.syms | 5 ++ src/qemu/qemu_hostdev.c | 16 +++++ src/qemu/qemu_hostdev.h | 4 ++ src/qemu/qemu_hotplug.c | 56 ++++++++++++++---- src/util/virfile.c | 52 +++++++++++++++++ src/util/virfile.h | 1 + src/util/virhostdev.c | 173 ++++++++++++++++++++++++++++++++++++++++++++++++++----- src/util/virhostdev.h | 16 +++++ src/util/virpci.c | 69 ++++++++++++++++++---- src/util/virpci.h | 4 ++ 10 files changed, 360 insertions(+), 36 deletions(-)

5 15

[libvirt] [PATCH] conf: Move virDomainPCIAddressBusIsFullyReserved()
by Andrea Bolognani 29 Jun '17

29 Jun '17

This function was private to the QEMU driver and was, accordingly, called qemuDomainPCIBusFullyReserved(). However, since it's really not QEMU-specific at all, it makes a lot of sense to have it closer to the virDomainPCIAddressBus struct it operates on; even more so considering we're going to introduce a complementary virDomainPCIAddressBusIsEmpty() function later on. Signed-off-by: Andrea Bolognani <abologna(a)redhat.com> --- src/conf/domain_addr.c | 14 ++++++++++++++ src/conf/domain_addr.h | 3 +++ src/libvirt_private.syms | 1 + src/qemu/qemu_domain_address.c | 15 +-------------- 4 files changed, 19 insertions(+), 14 deletions(-) diff --git a/src/conf/domain_addr.c b/src/conf/domain_addr.c index 639168e..0ae1ed9 100644 --- a/src/conf/domain_addr.c +++ b/src/conf/domain_addr.c @@ -355,6 +355,20 @@ virDomainPCIAddressBusSetModel(virDomainPCIAddressBusPtr bus, } +bool +virDomainPCIAddressBusIsFullyReserved(virDomainPCIAddressBusPtr bus) +{ + size_t i; + + for (i = bus->minSlot; i <= bus->maxSlot; i++) { + if (!bus->slot[i].functions) + return false; + } + + return true; +} + + /* Ensure addr fits in the address set, by expanding it if needed * * Return value: diff --git a/src/conf/domain_addr.h b/src/conf/domain_addr.h index f884b8a..c90a335 100644 --- a/src/conf/domain_addr.h +++ b/src/conf/domain_addr.h @@ -141,6 +141,9 @@ int virDomainPCIAddressBusSetModel(virDomainPCIAddressBusPtr bus, virDomainControllerModelPCI model) ATTRIBUTE_NONNULL(1); +bool virDomainPCIAddressBusIsFullyReserved(virDomainPCIAddressBusPtr bus) + ATTRIBUTE_NONNULL(1); + bool virDomainPCIAddressSlotInUse(virDomainPCIAddressSetPtr addrs, virPCIDeviceAddressPtr addr) ATTRIBUTE_NONNULL(1) ATTRIBUTE_NONNULL(2); diff --git a/src/libvirt_private.syms b/src/libvirt_private.syms index 888412a..d76c1a7 100644 --- a/src/libvirt_private.syms +++ b/src/libvirt_private.syms @@ -106,6 +106,7 @@ virDomainCCWAddressSetFree; virDomainCCWAddressValidate; virDomainGetBlkioParametersAssignFromDef; virDomainPCIAddressAsString; +virDomainPCIAddressBusIsFullyReserved; virDomainPCIAddressBusSetModel; virDomainPCIAddressEnsureAddr; virDomainPCIAddressFlagsCompatible; diff --git a/src/qemu/qemu_domain_address.c b/src/qemu/qemu_domain_address.c index b5b863f..efc698d 100644 --- a/src/qemu/qemu_domain_address.c +++ b/src/qemu/qemu_domain_address.c @@ -1499,19 +1499,6 @@ qemuDomainValidateDevicePCISlotsChipsets(virDomainDefPtr def, } -static bool -qemuDomainPCIBusFullyReserved(virDomainPCIAddressBusPtr bus) -{ - size_t i; - - for (i = bus->minSlot; i <= bus->maxSlot; i++) - if (!bus->slot[i].functions) - return false; - - return true; -} - - /* * This assigns static PCI slots to all configured devices. * The ordering here is chosen to match the ordering used @@ -2043,7 +2030,7 @@ qemuDomainAssignPCIAddresses(virDomainDefPtr def, bool buses_reserved = true; for (i = 0; i < addrs->nbuses; i++) { - if (!qemuDomainPCIBusFullyReserved(&addrs->buses[i])) { + if (!virDomainPCIAddressBusIsFullyReserved(&addrs->buses[i])) { buses_reserved = false; break; } -- 2.7.5

2 1

[libvirt] libvirt-2.0.0 build error (hidden symbol libvirt_event_poll_update_handle_semaphore)
by longguang.yue 29 Jun '17

29 Jun '17

Hi, all: i back port a patch which makes qemu depends on util directory. so i correct its dependency by applying a patch. but another error occur. the patch is http://libvirt.org/git/?p=libvirt.git;a=commit;h=57b5e27d3d74f8dbb77ac909b4… the patch fix dependency : --- libvirt-2.0.0/src/Makefile.am 2016-06-27 22:12:20.523191076 +0800 +++ libvirt-2.0.0-ok/src/Makefile.am 2017-06-22 12:25:17.512000000 +0800 @@ -1362,6 +1362,7 @@ -I$(srcdir)/access \ -I$(srcdir)/conf \ -I$(srcdir)/secret \ + -I$(srcdir)/util \ $(AM_CFLAGS) libvirt_driver_qemu_impl_la_LDFLAGS = $(AM_LDFLAGS) libvirt_driver_qemu_impl_la_LIBADD = $(CAPNG_LIBS) \ @@ -1369,6 +1370,7 @@ $(LIBNL_LIBS) \ $(LIBXML_LIBS) \ libvirt_secret.la \ + libvirt_util.la \ $(NULL) libvirt_driver_qemu_impl_la_SOURCES = $(QEMU_DRIVER_SOURCES) --------- BUILD ERROR: CCLD qemucapsprobe /bin/ld: .libs/qemucapsprobe: hidden symbol `libvirt_event_poll_update_handle_semaphore' in ../src/libvirt_probes.o is referenced by DSO /bin/ld: final link failed: Bad value collect2: error: ld returned 1 exit status make[1]: *** [qemucapsprobe] Error 1 make[1]: Leaving directory `/root/libvirt-2.0/BUILD/libvirt-2.0.0/tests' make: *** [check-am] Error 2 + cat test-suite.log cat: test-suite.log: No such file or directory + true + exit 1 error: Bad exit status from /var/tmp/rpm-tmp.ZVbGSv (%check) RPM build errors: Bad exit status from /var/tmp/rpm-tmp.ZVbGSv (%check) -------------

2 2

[libvirt] [PATCH v2 0/2] vz: fix vz driver compilation
by Mikhail Feoktistov 29 Jun '17

29 Jun '17

Mikhail Feoktistov (2): vz: add argument xmlopt for virDomainDefCheckABIStability call vz: nseclabels member is moved to virDomainChrSourceDef struct src/vz/vz_driver.c | 2 +- src/vz/vz_sdk.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) -- 1.8.3.1

2 3

[libvirt] [PATCH] docs: Document bhyve's vgaconf support
by Roman Bogorodskiy 29 Jun '17

29 Jun '17

- Add a news entry - Update driver's page with information about the new vgaconf attribute and provide usage example; while here, fix a grammar mistake Signed-off-by: Roman Bogorodskiy <bogorodskiy(a)gmail.com> --- docs/drvbhyve.html.in | 28 +++++++++++++++++++++++++++- docs/news.xml | 12 ++++++++++++ 2 files changed, 39 insertions(+), 1 deletion(-) diff --git a/docs/drvbhyve.html.in b/docs/drvbhyve.html.in index dd6620d..f083db9 100644 --- a/docs/drvbhyve.html.in +++ b/docs/drvbhyve.html.in @@ -351,10 +351,36 @@ FreeBSD port.</p> <p>This way, VNC will be accessible on <code>127.0.0.1:5904</code>.</p> -<p>Please note that the tablet device requires to have an USB controller +<p>Please note that the tablet device requires to have a USB controller of the <code>nec-xhci</code> model. Currently, only a single controller of this type and a single tablet are supported per domain.</p> +<p><span class="since">Since 3.5.0</span>, it's possible to configure how the video device is exposed +to the guest using the <code>vgaconf</code> attribute:</p> + +<pre> +<domain type='bhyve'> + <devices> + ... + <graphics type='vnc' port='5904'> + <listen type='address' address='127.0.0.1'/> + </graphics> + <video> + <driver vgaconf='on'/> + <model type='gop' heads='1' primary='yes'/> + </video> + ... + </devices> + ... +</domain> +</pre> + +<p>If not specified, bhyve's default mode for <code>vgaconf</code> +will be used. Please refer to the +<a href="https://www.freebsd.org/cgi/man.cgi?query=bhyve&sektion=8&manpath=F…">bhyve(8)</a> +manual page and the <a href="https://wiki.freebsd.org/bhyve">bhyve wiki</a> for more details on using +the <code>vgaconf</code> option.</p> + <h3><a name="clockconfig">Clock configuration</a></h3> <p>Originally bhyve supported only localtime for RTC. Support for UTC time was introduced in diff --git a/docs/news.xml b/docs/news.xml index 27562e1..d320965 100644 --- a/docs/news.xml +++ b/docs/news.xml @@ -46,6 +46,18 @@ select a boot entry. </description> </change> + <change> + <summary> + Support controlling how video devices are exposed to the bhyve guests + </summary> + <description> + The <code>vgaconf</code> attribute was added to <code>video</code>'s + <code>driver</code> element. Possible values are: <code>on</code>, + <code>off</code>, and <code>io</code>. It controlls the way how + bhyve exposes video devices to its guests; various guest OSes might + require different settings to boot properly. + </description> + </change> </section> <section title="Improvements"> </section> -- 2.7.4

2 2