- Devel - libvirt List Archives

[libvirt] [PATCH] virtio-blk: deprecate SCSI passthrough
by Stefan Hajnoczi 17 Dec '19

17 Dec '19

The Linux virtio_blk.ko guest driver is removing legacy SCSI passthrough support. Deprecate this feature in QEMU too. Signed-off-by: Stefan Hajnoczi <stefanha(a)redhat.com> --- qemu-deprecated.texi | 11 +++++++++++ 1 file changed, 11 insertions(+) diff --git a/qemu-deprecated.texi b/qemu-deprecated.texi index 4b4b7425ac..ef94d497da 100644 --- a/qemu-deprecated.texi +++ b/qemu-deprecated.texi @@ -285,6 +285,17 @@ spec you can use the ``-cpu rv64gcsu,priv_spec=v1.9.1`` command line argument. @section Device options +@subsection Emulated device options + +@subsubsection -device virtio-blk,scsi=on|off (since 5.0.0) + +The virtio-blk SCSI passthrough feature is a legacy VIRTIO feature. VIRTIO 1.0 +and later do not support it because the virtio-scsi device was introduced for +full SCSI support. Use virtio-scsi instead when SCSI passthrough is required. + +Note this also applies to ``-device virtio-blk-pci,scsi=on|off'', which is an +alias. + @subsection Block device options @subsubsection "backing": "" (since 2.12.0) -- 2.23.0

6 7

[libvirt] [PATCH 0/5] Introduce the support of Intel RDT-MBM
by Wang Huaqiang 17 Dec '19

17 Dec '19

RDT is the short for Intel Resource Director Technology, consists of four sub-technologies until now: -. CAT for cache allocation -. CMT for cache usage monitoring -. MBA for memory bandwidth allocation -. MBM for memory bandwidth usage monitoring The Linux kernel interface is 'resctrl' file system, and we have already implemented the support of CAT, CMT and MBA, to accomplish the tasks such as allocating a part of shared CPU last level cache to particular domain vcpu or a list of vcpus and monitoring the usage of cache, or the task of allocating a mount of memory bandwidth to specify domain vcpu(s). This series is to introduce the support of MBM. Basically the interfaces are: ** Identify host capability ** Similar to identify the host capability of CMT, it could be gotten through the result of 'virsh capabilities', if following elements are found, then MBM is supported: <memory_bandwidth> <monitor maxMonitors='176'> <feature name='mbm_total_bytes'/> <feature name='mbm_local_bytes'/> </monitor> </memory_bandwidth> 'mbm_total_bytes' means supporting to report the memory bandwidth used by the vcpu(s) of specific monitor on all CPU sockets. 'mbm_local_bytes' means supporting to report the memory bandwidth used by vcpu(s) that is passing through local CPU socket. ** Create monitor group** The monitor group for specific domain vcpus, for example vcpu 0-4, is defined in domain configuration file, in such kind of way: <cputune> <memorytune vcpus='0-4'> <monitor vcpus='0-4'/> </memorytune> </cputune> ** Report memory usage ** Introduced an option '--memory' against 'virsh domstats' command to show the memory bandwidth usage in such way: (also very similar to the format of CMT result.) # virsh domstats --memory Domain: 'libvirt-vm' memory.bandwidth.monitor.count=4 memory.bandwidth.monitor.0.name=vcpus_0-4 memory.bandwidth.monitor.0.vcpus=0-4 memory.bandwidth.monitor.0.node.count=2 memory.bandwidth.monitor.0.node.0.id=0 memory.bandwidth.monitor.0.node.0.bytes.total=14201651200 memory.bandwidth.monitor.0.node.0.bytes.local=7369809920 memory.bandwidth.monitor.0.node.1.id=1 memory.bandwidth.monitor.0.node.1.bytes.total=188897640448 memory.bandwidth.monitor.0.node.1.bytes.local=170044047360 Huaqiang (5): util, resctrl: using 64bit interface instead of 32bit for counters conf: showing cache/memoryBW monitor features in capabilities cachetune schema: a looser check for the order of <cache> and <monitor> element conf: Parse dommon configure file for memorytune monitors virsh: show memoryBW info in 'virsh domstats' command docs/schemas/domaincommon.rng | 91 +++++++++--------- include/libvirt/libvirt-domain.h | 1 + src/conf/capabilities.c | 4 +- src/conf/domain_conf.c | 44 +++++++-- src/libvirt-domain.c | 21 +++++ src/qemu/qemu_driver.c | 103 ++++++++++++++++++++- src/util/virfile.c | 40 ++++++++ src/util/virfile.h | 2 + src/util/virresctrl.c | 6 +- src/util/virresctrl.h | 2 +- tests/genericxml2xmlindata/cachetune.xml | 1 + tests/genericxml2xmlindata/memorytune.xml | 5 + tests/genericxml2xmloutdata/cachetune.xml | 34 +++++++ tests/genericxml2xmloutdata/memorytune.xml | 42 +++++++++ tests/genericxml2xmltest.c | 4 +- tools/virsh-domain-monitor.c | 7 ++ tools/virsh.pod | 23 ++++- 17 files changed, 367 insertions(+), 63 deletions(-) create mode 100644 tests/genericxml2xmloutdata/cachetune.xml create mode 100644 tests/genericxml2xmloutdata/memorytune.xml -- 2.23.0

4 17

[libvirt] [PATCH v1 00/26] move qemucaps validations from qemu_command to qemu_domain
by Daniel Henrique Barboza 16 Dec '19

16 Dec '19

(series based on master commit 97cafa610ecf5) This work was proposed by Cole in [1]. This is Cole's reasoning for it, copy/pasted from [1]: ------- The benefits of moving to validate time is that XML is rejected by 'virsh define' rather than at 'virsh start' time. It also makes it easier to follow the cli building code, and makes it easier to verify qemu_command.c test suite code coverage for the important stuff like covering every CLI option. It's also a good intermediate step for sharing validation with domain capabilities building, like is done presently for rng models. ------- Cole also mentioned that the machine features validation was a good place to start. I took it a step further and did it across the board on all qemu_command.c. This series didn't create any new validation, just moved them to domain define time. Any other outcome is unintended. Not all cases where covered in this work. For a first version I decided to do only the most trivial cases. These are the other cases I left out for another day: - all verifications contained in functions that are also called by qemu_hotplug.c. This means that the hotplug process will also use the validation code, and we can't just move it to qemu_domain.c. Besides, not all validation done in domain define time applies to hotplug, so it's not simply a matter of calling the same function from qemu_domain in qemu_hotplug. I have patches that moved the verification of all virtio options to qemu_domain.c, while also considering qemu_hotplug validation. In the end I decided to leave it away from this work for now because (1) it took 8 patches just for virtio case alone and (2) there are a lot of these cases in qemu_command.c and it would be too much to do it all at in this same series. - moving CPU Model validation is trickier than the rest because there are code in DefPostParse() that makes CPU Model changes that are then validated in qemu_command.c. Moving the validation to define time doesn't cut in this case - the validation is considering PostParse changes in the CPU Model and some of the will fail if done by qemuDomainDefValidate time. I didn't think too much about how to handle this case, but given that the approach would be different from the other cases handled here, I left it out too. - SHMem: part of SHMem validation is being used by hotplug code. I could've moved the non-hotplug validation to define time, instead I decided it's better to to leave it to do it all at once in the future. - DBus-VMState: validation of this fella must be done by first querying if the hash vm->priv->dbusVMState has elements. qemuDomainDefValidate does not have 'vm->priv' in it's API, meaning that I would need to either change the API to include it (which means changing domainValidateCallback), or do the validation by another branch where I can get access to vm->priv. - vmcoreinfo: there are no challenges in moving vmcoreinfo validation to qemuDomainDefValidate. The problem were the unit tests. Moving this validation to domain define time breaks 925 tests on qemuxml2xmltest.c, including tests labelled as 'minimal' which should pass under any circurstances, as far as I understand. It is possible to just shoehorn QEMU_CAPS_DEVICE_VMCOREINFO in the qemuCaps of all tests, but this would tamper the idea of having to deal with NONE or LATEST or a specific set of capabilities for certain tests. Another case I would rather discuss separately. [1] https://www.redhat.com/archives/libvir-list/2019-October/msg00568.html Daniel Henrique Barboza (26): qemu_command.c: move PSeries features validation to qemu_domain.c qemu_command.c: move mem.nosharepages validation to qemu_domain.c qemu_command.c: move validation of vmport to qemu_domain.c qemu_command.c: move nvdimm validation to qemu_domain.c qemu_command.c: move I/O APIC validation to qemu_domain.c numa_conf: add virDomainNumaNodesDistancesAreBeingSet() helper qemu_command.c move NUMA validation to qemu_domain.c qemu_command.c: move NVRAM validation to qemu_domain.c qemu_command: move qemuBuildSoundDevStr caps validation to qemu_domain qemu_command.c: move sound codec validation to qemu_domain.c qemu_command: move qemuBuildHubDevStr caps validation to qemu_domain qemu_command: move qemuBuildChrChardevStr caps validation to qemu_domain qemu: move qemuBuildHostdevCommandLine caps validation to qemu_domain qemu_command.c: move vmGenID validation to qemu_domain.c qemu: move qemuBuildSgaCommandLine validation to qemu_domain.c qemu: move virDomainClockDef validation to qemu_domain.c qemu: move qemuBuildPMCommandLine validation to qemu_domain.c qemu: move qemuBuildBootCommandLine validation to qemu_domain.c qemu_command.c: move pcihole64 validation to qemu_domain.c qemu: move qemuBuildGraphicsSDLCommandLine validation to qemu_domain.c qemu: move qemuBuildGraphicsVNCCommandLine validation to qemu_domain.c qemu: move qemuBuildGraphicsSPICECommandLine validation to qemu_domain.c qemu: move qemuBuildGraphicsEGLHeadlessCommandLine validation to qemu_domain.c qemu: move qemuBuildSmartcardCommandLine validation to qemu_domain.c qemu: move qemuBuildConsoleCommandLine validation to qemu_domain.c qemu: move qemuBuildTPMDevStr TPM validation to qemu_domain.c src/conf/numa_conf.c | 19 + src/conf/numa_conf.h | 2 + src/libvirt_private.syms | 1 + src/qemu/qemu_command.c | 575 +--------- src/qemu/qemu_command.h | 1 + src/qemu/qemu_domain.c | 983 ++++++++++++++++-- tests/qemuhotplugtest.c | 10 + tests/qemumemlocktest.c | 16 +- .../default-video-type-aarch64.xml | 2 +- .../default-video-type-ppc64.xml | 2 +- .../default-video-type-s390x.xml | 2 +- .../graphics-egl-headless.args | 2 +- .../graphics-spice-egl-headless.args | 2 +- .../graphics-vnc-egl-headless.args | 2 +- tests/qemuxml2argvtest.c | 80 +- ...ault-video-type-aarch64.aarch64-latest.xml | 4 +- .../default-video-type-ppc64.ppc64-latest.xml | 4 +- .../default-video-type-s390x.s390x-latest.xml | 4 +- tests/qemuxml2xmltest.c | 244 +++-- 19 files changed, 1246 insertions(+), 709 deletions(-) -- 2.23.0

4 42

[libvirt] [PATCH v6 0/4] scripts: convert most perl scripts to python
by Daniel P. Berrangé 16 Dec '19

16 Dec '19

This series is an effort to reduce the number of different languages we use by eliminating most use of perl in favour of python. This aligns with fact that the likely future build system we'll use (meson) is written in python, and that python is much more commonly used/understood by developers these days than perl. With this applied we use perl in a handful of places only: - src/rpc/gendispatch.pl - this is a horrendously large script and very hard to understand/follow. A straight syntax conversion to Python would still leave a hgue and hard to understand/follow script. It really needs a clean room rewrite from scratch, with better structure. - src/rpc/genprotocol.pl - fairly easy to convert, but might be obsolete depending on approach for rewriting gendispatch.pl, so ignored for now - tests/oomtrace.pl - will be purge by the patches that drop OOM handling anyway - tools/wireshark/util/genxdrstub.pl - a very large script, which I haven't got the courage to tackle yet. - cfg.mk/maint.mk - many syntax rules involve regexes which are fed to perl. Decision on what to do with syntax-check rules punted to another time. - build-aux/gitlog-to-changelog - build-aux/useless-if-before-free - Both pulled in from gnulib. Could be rewritten quite easily if desired, but given that we aren't maintainers of them right now, they're ignored as they don't really impact our developers. Note that the check-spacing.py script is significantly slower in Python than in Perl. After researching this it appears there is nothing that can be done. The Perl regex engine is simply much better optimized than the Python one. As previously discussed we need to loook at uncrustify or clang-format or some other tool to validate whitespace formatting. This is ongoing. We can either accept the slow down in the short term or keep the Perl version in the short term. In v6: - Fix various bugs in check-file-access.py so it now reports same output as original perl script. - Merged the already-ACKd scripts In v5: - Rebased to cope with changes to require VPATH build - Merged the already-ACKd scripts In v4: - Moved all scripts into the scripts/ directory instead of having them scattered around source tree - Use re.search instead of re.match - Don't bother re-compiling regexes In v3: - All scripts comply with all flake8 style rules with exception of E129 visually indented line with same indent as next logical line In v2: - Pulled in patch to hacking file - Converted many more scripts - Forced UTF-8 character set to avoid ascii codec on py3 < 3.7 Daniel P. Berrangé (4): tests: rewrite qemu capability grouper in Python tests: rewrite file access checker in Python docs: rewrite hvsupport.html page generator in python docs: rewrite polkit docs generator in Python Makefile.am | 4 + build-aux/syntax-check.mk | 3 +- docs/Makefile.am | 12 +- docs/genaclperms.pl | 125 -------- docs/hvsupport.pl | 459 ----------------------------- scripts/check-file-access.py | 125 ++++++++ scripts/genaclperms.py | 121 ++++++++ scripts/group-qemu-caps.py | 121 ++++++++ scripts/hvsupport.py | 502 ++++++++++++++++++++++++++++++++ tests/Makefile.am | 3 +- tests/check-file-access.pl | 130 --------- tests/file_access_whitelist.txt | 2 +- tests/group-qemu-caps.pl | 124 -------- 13 files changed, 882 insertions(+), 849 deletions(-) delete mode 100755 docs/genaclperms.pl delete mode 100755 docs/hvsupport.pl create mode 100755 scripts/check-file-access.py create mode 100755 scripts/genaclperms.py create mode 100755 scripts/group-qemu-caps.py create mode 100755 scripts/hvsupport.py delete mode 100755 tests/check-file-access.pl delete mode 100755 tests/group-qemu-caps.pl -- 2.21.0

2 5

[libvirt] [tck PATCH] Update x86_64 image & virt-builder image to F31
by Daniel P. Berrangé 16 Dec '19

16 Dec '19

i386 was dropped even as a secondary arch in F31 so must remain on F30 Signed-off-by: Daniel P. Berrangé <berrange(a)redhat.com> --- conf/default.cfg | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/conf/default.cfg b/conf/default.cfg index 39d9b04..3493ad7 100644 --- a/conf/default.cfg +++ b/conf/default.cfg @@ -49,7 +49,7 @@ images = ( hvm xen ) - osname = fedora-30 + osname = fedora-31 } ) @@ -76,6 +76,7 @@ images = ( # kernels = ( # Fedora 30 i686 PAE has pv_ops, so one kernel can do both Xen and KVM guests here + # No i386 releases since F30 { arch = i686 ostype = ( @@ -85,15 +86,15 @@ kernels = ( kernel = http://dl.fedoraproject.org/pub/fedora-secondary/releases/30/Everything/i38… initrd = http://dl.fedoraproject.org/pub/fedora-secondary/releases/30/Everything/i38… } - # Fedora 30 x86_64 has pv_ops, so one kernel can do both Xen and KVM guests here + # Fedora 31 x86_64 has pv_ops, so one kernel can do both Xen and KVM guests here { arch = x86_64 ostype = ( hvm xen ) - kernel = http://dl.fedoraproject.org/pub/fedora/linux/releases/30/Everything/x86_64/… - initrd = http://dl.fedoraproject.org/pub/fedora/linux/releases/30/Everything/x86_64/… + kernel = http://dl.fedoraproject.org/pub/fedora/linux/releases/31/Everything/x86_64/… + initrd = http://dl.fedoraproject.org/pub/fedora/linux/releases/31/Everything/x86_64/… } # LXC containers need a virtual container filesystem somewhere # { -- 2.23.0

2 1

[libvirt] ongoing code transitions section on the wiki
by Cole Robinson 16 Dec '19

16 Dec '19

Hi all, I added a 'Ongoing code transitions' section to the BiteSizedTasks page on the wiki: https://wiki.libvirt.org/page/BiteSizedTasks#Ongoing_code_transitions BiteSizedTasks might not be the best location for it, but I wanted to start a list and that page already exists. I added 3 sections: * g_autofree/g_autoptr, with some example conversion commits * GLib in general, listing APIs that can replace custom libvirt code. Currently the list is only g_key_file for virkeyfile, g_user_*_dir for virGetUser*Dir, some virFile APIs for g_path_* APIs. * gnulib removal. I filled in a few examples of possible gnulib module replacements, and listed all the others as 'Unsorted/replacments unclear'. If you know any GLib replacements for existing code or gnulib modules, please fill them in there. Or any other incomplete code conversions internally Thanks, Cole

1 0

[libvirt] [rust PATCH] Domain: Implement scheduler information related APIs
by Vineeth Remanan Pillai 16 Dec '19

16 Dec '19

Implement the following: - virDomainGetSchedulerType - virDomainSetSchedulerParameters - virDomainSetSchedulerParametersFlags - virDomainGetSchedulerParameters - virDomainGetSchedulerParametersFlags Signed-off-by: Vineeth Remanan Pillai <vpilllai(a)digitalocean.com> --- examples/hello.rs | 38 +++++++ src/domain.rs | 260 +++++++++++++++++++++++++++++++++++++++++++--- 2 files changed, 283 insertions(+), 15 deletions(-) diff --git a/examples/hello.rs b/examples/hello.rs index e5ec6e4..1f336a9 100644 --- a/examples/hello.rs +++ b/examples/hello.rs @@ -89,6 +89,44 @@ fn show_domains(conn: &Connect) -> Result<(), Error> { numa.node_set.unwrap_or(String::from(""))); println!(" Mode: {}", numa.mode.unwrap_or(0)); } + + if let Ok((sched_type, nparams)) = dom.get_scheduler_type() { + println!("SchedType: {}, nparams: {}", + sched_type, nparams); + } + + if let Ok(sched_info) = dom.get_scheduler_parameters() { + println!("Schedule Information:"); + println!("\tScheduler\t: {}", sched_info.scheduler_type); + if let Some(shares) = sched_info.cpu_shares { + println!("\tcpu_shares\t: {}", shares); + } + if let Some(period) = sched_info.vcpu_bw.period { + println!("\tvcpu_period\t: {}", period); + } + if let Some(quota) = sched_info.vcpu_bw.quota { + println!("\tvcpu_quota\t: {}", quota); + } + if let Some(period) = sched_info.emulator_bw.period { + println!("\temulator_period\t: {}", period); + } + if let Some(quota) = sched_info.emulator_bw.quota { + println!("\temulator_quota\t: {}", quota); + } + if let Some(period) = sched_info.global_bw.period { + println!("\tglobal_period\t: {}", period); + } + if let Some(quota) = sched_info.global_bw.quota { + println!("\tglobal_quota\t: {}", quota); + } + if let Some(period) = sched_info.global_bw.period { + println!("\tiothread_period\t: {}", period); + } + if let Some(quota) = sched_info.global_bw.quota { + println!("\tiothread_quota\t: {}", quota); + } + } + } } return Ok(()); diff --git a/src/domain.rs b/src/domain.rs index acb9e6e..40a18d8 100644 --- a/src/domain.rs +++ b/src/domain.rs @@ -363,6 +363,29 @@ extern "C" { snaps: *mut *mut virDomainSnapshotPtr, flags: libc::c_uint) -> libc::c_int; + + fn virDomainGetSchedulerType(ptr: sys::virDomainPtr, + nparams: *mut libc::c_int) + -> *mut libc::c_char; + fn virDomainGetSchedulerParameters(ptr: sys::virDomainPtr, + params: virTypedParameterPtr, + nparams: *mut libc::c_int) + -> libc::c_int; + fn virDomainSetSchedulerParameters(ptr: sys::virDomainPtr, + params: virTypedParameterPtr, + nparams: libc::c_int) + -> libc::c_int; + fn virDomainGetSchedulerParametersFlags(ptr: sys::virDomainPtr, + params: virTypedParameterPtr, + nparams: *mut libc::c_int, + flags: libc::c_uint) + -> libc::c_int; + fn virDomainSetSchedulerParametersFlags(ptr: sys::virDomainPtr, + params: virTypedParameterPtr, + nparams: libc::c_int, + flags: libc::c_uint) + -> libc::c_int; + } pub type DomainXMLFlags = self::libc::c_uint; @@ -607,6 +630,124 @@ impl MemoryStats { } } +/// Structure representing the CFS scheduler cpu bandwidth parameters +/// see https://www.kernel.org/doc/html/latest/scheduler/sched-bwc.html +#[derive(Clone, Debug, Default)] +pub struct SchedBandwidth { + pub period: Option<u64>, + pub quota: Option<i64>, +} + +#[derive(Clone, Debug, Default)] +pub struct SchedulerInfo { + pub scheduler_type: String, + // cpu shares for the domain. + pub cpu_shares: Option<u64>, + // Bandwidth allocated for the vcpu threads. + pub vcpu_bw: SchedBandwidth, + // Bandwidth allocated for the emulator threads. + pub emulator_bw: SchedBandwidth, + // Bandwidth allocated for the Domain. + pub global_bw: SchedBandwidth, + // Bandwidth allocated for the io threads.. + pub iothread_bw: SchedBandwidth, +} + +impl SchedulerInfo { + pub fn from_vec(vec: Vec<virTypedParameter>, scheduler_type: String) -> SchedulerInfo { + unsafe { + let mut ret = SchedulerInfo::default(); + ret.scheduler_type = scheduler_type; + for param in vec { + match str::from_utf8(CStr::from_ptr(param.field.as_ptr()).to_bytes()).unwrap() { + "cpu_shares" => ret.cpu_shares = Some(param.value as u64), + "vcpu_period" => ret.vcpu_bw.period = Some(param.value as u64), + "vcpu_quota" => ret.vcpu_bw.quota = Some(param.value as i64), + "emulator_period" => ret.emulator_bw.period = Some(param.value as u64), + "emulator_quota" => ret.emulator_bw.quota = Some(param.value as i64), + "global_period" => ret.global_bw.period = Some(param.value as u64), + "global_quota" => ret.global_bw.quota = Some(param.value as i64), + "iothread_period" => ret.iothread_bw.period = Some(param.value as u64), + "iothread_quota" => ret.iothread_bw.quota = Some(param.value as i64), + unknow => panic!("Field not implemented for SchedulerInfo, {:?}", unknow), + } + } + ret + } + } + + pub fn to_vec(&self) -> Vec<virTypedParameter> { + let mut cparams: Vec<virTypedParameter> = Vec::new(); + + if let Some(shares) = self.cpu_shares { + cparams.push(virTypedParameter { + field: to_arr("cpu_shares\0"), + typed: ::typedparam::VIR_TYPED_PARAM_ULLONG, + value: shares + }); + } + + if let Some(period) = self.vcpu_bw.period { + cparams.push(virTypedParameter { + field: to_arr("vcpu_period\0"), + typed: ::typedparam::VIR_TYPED_PARAM_ULLONG, + value: period, + }); + } + if let Some(quota) = self.vcpu_bw.quota { + cparams.push(virTypedParameter { + field: to_arr("vcpu_quota\0"), + typed: ::typedparam::VIR_TYPED_PARAM_LLONG, + value: quota as u64, + }); + } + if let Some(period) = self.emulator_bw.period { + cparams.push(virTypedParameter { + field: to_arr("emulator_period\0"), + typed: ::typedparam::VIR_TYPED_PARAM_ULLONG, + value: period, + }); + } + if let Some(quota) = self.emulator_bw.quota { + cparams.push(virTypedParameter { + field: to_arr("emulator_quota\0"), + typed: ::typedparam::VIR_TYPED_PARAM_LLONG, + value: quota as u64, + }); + } + if let Some(period) = self.global_bw.period { + cparams.push(virTypedParameter { + field: to_arr("global_period\0"), + typed: ::typedparam::VIR_TYPED_PARAM_ULLONG, + value: period, + }); + } + if let Some(quota) = self.global_bw.quota { + cparams.push(virTypedParameter { + field: to_arr("global_quota\0"), + typed: ::typedparam::VIR_TYPED_PARAM_LLONG, + value: quota as u64, + }); + } + if let Some(period) = self.iothread_bw.period { + cparams.push(virTypedParameter { + field: to_arr("iothread_period\0"), + typed: ::typedparam::VIR_TYPED_PARAM_ULLONG, + value: period, + }); + } + if let Some(quota) = self.iothread_bw.quota { + cparams.push(virTypedParameter { + field: to_arr("iothread_quota\0"), + typed: ::typedparam::VIR_TYPED_PARAM_LLONG, + value: quota as u64, + }); + } + + cparams + } +} + /// Provides APIs for the management of domains. /// /// See http://libvirt.org/html/libvirt-libvirt-domain.html @@ -627,6 +768,14 @@ impl Drop for Domain { } } +fn to_arr(name: &str) -> [libc::c_char; 80] { + let mut field: [libc::c_char; 80] = [0; 80]; + for (a, c) in field.iter_mut().zip(name.as_bytes()) { + *a = *c as i8 + } + field +} + impl Domain { pub fn new(ptr: sys::virDomainPtr) -> Domain { return Domain { ptr: Some(ptr) }; @@ -1656,14 +1805,6 @@ impl Domain { flags: u32) -> Result<u32, Error> { unsafe { - fn to_arr(name: &str) -> [libc::c_char; 80] { - let mut field: [libc::c_char; 80] = [0; 80]; - for (a, c) in field.iter_mut().zip(name.as_bytes()) { - *a = *c as i8 - } - field - } - let mut cparams: Vec<virTypedParameter> = Vec::new(); if params.hard_limit.is_some() { cparams.push(virTypedParameter { @@ -1809,13 +1950,6 @@ impl Domain { pub fn set_numa_parameters(&self, params: NUMAParameters, flags: u32) -> Result<u32, Error> { unsafe { - fn to_arr(name: &str) -> [libc::c_char; 80] { - let mut field: [libc::c_char; 80] = [0; 80]; - for (a, c) in field.iter_mut().zip(name.as_bytes()) { - *a = *c as i8 - } - field - } let mut cparams: Vec<virTypedParameter> = Vec::new(); if params.node_set.is_some() { @@ -1863,4 +1997,100 @@ impl Domain { return Ok(array); } } + + /// Get the cpu scheduler type for the domain + pub fn get_scheduler_type(&self) -> Result<(String, i32), Error> + { + unsafe { + let mut nparams: libc::c_int = -1; + let sched_type = virDomainGetSchedulerType(self.as_ptr(), + &mut nparams); + if sched_type.is_null() { + return Err(Error::new()); + } + + return Ok((c_chars_to_string!(sched_type), nparams)); + } + } + + /// Get the scheduler parameters for the domain. + pub fn get_scheduler_parameters(&self) -> Result<SchedulerInfo, Error> { + + let (sched_type, mut nparams) = self.get_scheduler_type()?; + unsafe { + let mut params: Vec<virTypedParameter> = vec![ + virTypedParameter::default(); 9]; + let ret = virDomainGetSchedulerParameters(self.as_ptr(), + &mut params[0], + &mut nparams); + if ret == -1 { + return Err(Error::new()); + } + Ok(SchedulerInfo::from_vec(params, sched_type)) + } + + } + + /// Get the scheduler parameters for the domain for the configuration + /// as specified by the flags. + /// # Arguments + /// + /// * `flags` - Specifies the Domain Impact: CONFIG, LIVE or CURRENT. + pub fn get_scheduler_parameters_flags(&self, + flags: DomainModImpactFlags) + -> Result<SchedulerInfo, Error> { + + let (sched_type, mut nparams) = self.get_scheduler_type()?; + unsafe { + let mut params: Vec<virTypedParameter> = vec![ + virTypedParameter::default(); 9]; + let ret = virDomainGetSchedulerParametersFlags(self.as_ptr(), + &mut params[0], + &mut nparams, + flags as libc::c_uint); + if ret == -1 { + return Err(Error::new()); + } + Ok(SchedulerInfo::from_vec(params, sched_type)) + } + + } + + /// Set the scheduler parameters for the domain. + pub fn set_scheduler_parameters(&self, + sched_info: &SchedulerInfo) + -> Result<i32, Error> { + unsafe { + let mut params = sched_info.to_vec(); + let ret = virDomainSetSchedulerParameters(self.as_ptr(), + &mut params[0], + params.len() as libc::c_int); + if ret == -1 { + return Err(Error::new()); + } + Ok(ret) + } + } + + /// Set the scheduler parameters for the domain for the configuration + /// as specified by the flags. + /// # Arguments + /// + /// * `flags` - Specifies the Domain Impact: CONFIG, LIVE or CURRENT. + pub fn set_scheduler_parameters_flags(&self, + sched_info: &SchedulerInfo, + flags: DomainModImpactFlags) + -> Result<i32, Error> { + unsafe { + let mut params = sched_info.to_vec(); + let ret = virDomainSetSchedulerParametersFlags(self.as_ptr(), + &mut params[0], + params.len() as libc::c_int, + flags as libc::c_uint); + if ret == -1 { + return Err(Error::new()); + } + Ok(ret) + } + } } -- 2.17.1

2 2

[libvirt] [PATCH] src: warn against virNodeGetInfo() API call due to inaccurate info
by Daniel P. Berrangé 16 Dec '19

16 Dec '19

Signed-off-by: Daniel P. Berrangé <berrange(a)redhat.com> --- include/libvirt/libvirt-host.h | 4 ++++ src/libvirt-host.c | 16 ++++++++++++++++ 2 files changed, 20 insertions(+) diff --git a/include/libvirt/libvirt-host.h b/include/libvirt/libvirt-host.h index be65b4686b..b87d634813 100644 --- a/include/libvirt/libvirt-host.h +++ b/include/libvirt/libvirt-host.h @@ -147,6 +147,10 @@ typedef virSecurityModel *virSecurityModelPtr; * * a virNodeInfo is a structure filled by virNodeGetInfo() and providing * the information for the Node. + * + * Note that the information in this struct is not guaranteed to be an + * accurate relection of the system hardware. See the virNodeGetInfo() + * API documentation for further guidance. */ typedef struct _virNodeInfo virNodeInfo; diff --git a/src/libvirt-host.c b/src/libvirt-host.c index 94ba5a8e80..bc3d1d2803 100644 --- a/src/libvirt-host.c +++ b/src/libvirt-host.c @@ -399,6 +399,22 @@ virConnectGetMaxVcpus(virConnectPtr conn, * * Extract hardware information about the node. * + * Use of this API is strongly discouraged as the information provided + * is not guaranteed to be accurate on all hardware platforms. + * + * The mHZ value merely reflects the speed that the first CPU in the + * machine is currently running at. This speed may vary across CPUs + * and changes continually as the host OS throttles. + * + * The nodes/sockets/cores/threads data is potentially inaccurate as + * it assumes a symmetric installation. If one NUMA node has more + * sockets populated that another NUMA node this information will be + * wrong. It is also not able to report about CPU dies. + * + * Applications are recommended to use the virConnectGetCapabilities() + * call instead, which provides all the information except CPU mHZ, + * in a more accurate representation. + * * Returns 0 in case of success and -1 in case of failure. */ int -- 2.23.0

2 1

[libvirt] [PATCH] tools: skip libvirt-guests fast if libvirtd is not active
by Christian Ehrhardt 16 Dec '19

16 Dec '19

The most common operation of libvirt-guests is to manage the local libvirtd. But users might have disabled that and while we are After=libvirtd for ordering we are not Requiring it.. OTOH adding that or any harder dependency might affect our ordering. But if people have disabled libvirt they will do a full retry loop until timeout. Lets check if the local service is active at all and skip fast if it is not. Fixes: https://bugs.launchpad.net/ubuntu/+source/libvirt/+bug/1854653 Reported-by: Doug Smythies <dsmythies(a)telus.net> Signed-off-by: Christian Ehrhardt <christian.ehrhardt(a)canonical.com> --- tools/libvirt-guests.sh.in | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/tools/libvirt-guests.sh.in b/tools/libvirt-guests.sh.in index 4bc6e866f0..5a9930ee2f 100644 --- a/tools/libvirt-guests.sh.in +++ b/tools/libvirt-guests.sh.in @@ -90,6 +90,14 @@ test_connect() { uri=$1 + if [ "x$uri" = xdefault ]; then + # Default config is most common and for the local libvirtd + # Check if it is active before wasting time in connect loop + if ! systemctl -q is-active libvirtd; then + return 1 + fi + fi + i=${CONNECT_RETRIES} while [ $i -gt 0 ]; do run_virsh "$uri" connect 2>/dev/null -- 2.24.0

4 8

[libvirt] [PATCH] virsh.pod: Mention iscsi-direct is supported in find-storage-pool-sources
by Han Han 13 Dec '19

13 Dec '19

Signed-off-by: Han Han <hhan(a)redhat.com> --- tools/virsh.pod | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/tools/virsh.pod b/tools/virsh.pod index a8331154e1..14f4121787 100644 --- a/tools/virsh.pod +++ b/tools/virsh.pod @@ -3994,7 +3994,8 @@ any other provided XML elements in I<srcSpec>. For a "logical" pool, the contents of the I<srcSpec> file are ignored, although if provided the file must at least exist. -For an "iscsi" pool, the minimal expect XML required is the <host> element +For an "iscsi" or "iscsi-direct" pool, +the minimal expect XML required is the <host> element with a "name" attribute describing the IP address or hostname to be used to find the pool (the iSCSI server address). Optionally, the "port" attribute may be provided, although it will default to 3260. Optionally, an <initiator> -- 2.24.0.rc1

2 1