January 2019 - Devel - libvirt List Archives

[libvirt] [PATCH 00/19] implement cgroups v2 devices support
by Pavel Hrdina 07 Jan '19

07 Jan '19

In cgroups v2 there is no devices controller, BPF should be used instead. Patches 3 - 12 will be squashed into single commit and they need to be compiled together, I've separated them to make review easier. Pavel Hrdina (19): util: introduce virbpf helpers vircgroup: introduce virCgroupV2DevicesAvailable vircgroup: introduce virCgroupV2DeviceLoadProg vircgroup: introduce virCgroupV2DeviceAttachProg vircgroup: introduce virCgroupV2DeviceDetectProg vircgroup: introduce virCgroupV2DeviceCreateProg vircgroup: introduce virCgroupV2DeviceReallocMap vircgroup: introduce virCgroupV2DevicePrepareProg vircgroup: introduce virCgroupV2DeviceRemoveProg vircgroup: introduce virCgroupV2DeviceGetPerms vircgroup: introduce virCgroupV2DeviceGetKey vircgroup: introduce virCgroupV2AllowDevice vircgroup: introduce virCgroupV2DenyDevice vircgroup: introduce virCgroupV2AllowAllDevices vircgroup: introduce virCgroupV2DenyAllDevices vircgroup: workaround devices in hybrid mode vircgroupv2: detech BPF program before removing cgroup vircgroupv2: use dummy process to workaround kernel bug with systemd vircgroupmock: mock virBPFQueryProg include/libvirt/virterror.h | 1 + src/Makefile.am | 1 + src/libvirt_private.syms | 17 + src/lxc/lxc_cgroup.c | 1 + src/qemu/qemu_cgroup.c | 6 +- src/util/Makefile.inc.am | 2 + src/util/virbpf.c | 263 ++++++++++++ src/util/virbpf.h | 249 ++++++++++++ src/util/vircgroup.c | 18 +- src/util/vircgroup.h | 1 + src/util/vircgroupbackend.h | 3 +- src/util/vircgrouppriv.h | 12 + src/util/vircgroupv1.c | 9 +- src/util/vircgroupv2.c | 638 +++++++++++++++++++++++++++++- src/util/virerror.c | 1 + src/util/virsystemd.c | 2 +- src/util/virsystemd.h | 2 + tests/vircgroupdata/hybrid.parsed | 2 +- tests/vircgroupmock.c | 11 + tests/vircgrouptest.c | 4 +- 20 files changed, 1233 insertions(+), 10 deletions(-) create mode 100644 src/util/virbpf.c create mode 100644 src/util/virbpf.h -- 2.20.1

2 20

[libvirt] [PATCH 0/3] Rebase LXC network definition to support version 3.0
by Julio Faracco 07 Jan '19

07 Jan '19

The series propose a new way to define NICs inside LXC native. It is needed because LXC version 3.X uses indexes to define NICs and the current algorithm is not able to support them. At least, if you consider settings defined using a random logic. Julio Faracco (3): lxc: refactoring LXC network definition with a sparse array. tests: Change legacy network configs to version 3.0 with indexes. tests: Include a random network testcase to test indexes. src/lxc/lxc_native.c | 191 ++++++++++++------ .../lxcconf2xml-ethernet-v3.config | 16 +- .../lxcconf2xml-fstab-v3.config | 10 +- .../lxcconf2xml-macvlannetwork-v3.config | 10 +- .../lxcconf2xml-miscnetwork-v3.config | 38 ++-- .../lxcconf2xml-physnetwork-v3.config | 14 +- .../lxcconf2xml-randomnetwork-v3.config | 21 ++ .../lxcconf2xml-randomnetwork.xml | 45 +++++ .../lxcconf2xml-simple-v3.config | 18 +- .../lxcconf2xml-vlannetwork-v3.config | 10 +- tests/lxcconf2xmltest.c | 1 + 11 files changed, 249 insertions(+), 125 deletions(-) create mode 100644 tests/lxcconf2xmldata/lxcconf2xml-randomnetwork-v3.config create mode 100644 tests/lxcconf2xmldata/lxcconf2xml-randomnetwork.xml -- 2.19.1

2 4

[libvirt] [PATCH 0/2] News update for libvirt-5.0
by Han Han 07 Jan '19

07 Jan '19

Han Han (2): news: Add support for "stibp" x86_64 feature news: Add support for postcopy-requests migration statistics docs/news.xml | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+) -- 2.20.1

2 3

[libvirt] [jenkins-ci PATCH 0/2] Stop building Go projects on CentOS 7
by Andrea Bolognani 07 Jan '19

07 Jan '19

Found while creating a new CentOS 7 guest from scratch using lcitool. See the commit messages for more details. Andrea Bolognani (2): Stop building Go projects on CentOS 7 guests: Don't prepare CentOS 7 for Go projects guests/host_vars/libvirt-centos-7/main.yml | 2 -- guests/playbooks/build/projects/libvirt-go-xml.yml | 13 ++++++++++++- guests/playbooks/build/projects/libvirt-go.yml | 13 ++++++++++++- projects/libvirt-go-xml.yaml | 9 ++++++++- projects/libvirt-go.yaml | 9 ++++++++- 5 files changed, 40 insertions(+), 6 deletions(-) -- 2.19.2

3 5

[libvirt] [PATCH] lxc: allow empty path in URI for historical compatibility
by Daniel P. Berrangé 07 Jan '19

07 Jan '19

The use of 'lxc://' was mistakenly broken in: commit 4c8574c85c554e68de0d8bf9b85727954a5bea91 Author: Daniel P. Berrangé <berrange(a)redhat.com> Date: Wed Mar 28 12:49:29 2018 +0100 driver: ensure NULL URI isn't passed to drivers with whitelisted URIs Allow it again for historical compatibility. Signed-off-by: Daniel P. Berrangé <berrange(a)redhat.com> --- src/lxc/lxc_driver.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/src/lxc/lxc_driver.c b/src/lxc/lxc_driver.c index de045c80bb..df15a0da50 100644 --- a/src/lxc/lxc_driver.c +++ b/src/lxc/lxc_driver.c @@ -140,7 +140,8 @@ static virDrvOpenStatus lxcConnectOpen(virConnectPtr conn, virCheckFlags(VIR_CONNECT_RO, VIR_DRV_OPEN_ERROR); /* If path isn't '/' then they typoed, tell them correct path */ - if (STRNEQ(conn->uri->path, "/") && + if (STRNEQ(conn->uri->path, "") && + STRNEQ(conn->uri->path, "/") && STRNEQ(conn->uri->path, "/system")) { virReportError(VIR_ERR_INTERNAL_ERROR, _("Unexpected LXC URI path '%s', try lxc:///system"), -- 2.19.2

2 1

[libvirt] [GSoC] Update ideas page
by Michal Prívozník 07 Jan '19

07 Jan '19

Dear list, Google announced another round of GSoC [1]. The organization applications can be filled mid January next year (~1.5 month to go). Please update our Ideas page [2] if you have any idea for student projects (you do not have to mentor it if you don't feel like it). Similarly, if you want to mentor certain idea and there is no mentor assigned yet, do not hesitate to add yourself there. Michal 1: https://developers.google.com/open-source/gsoc/timeline 2: https://wiki.libvirt.org/page/Google_Summer_of_Code_Ideas

2 1

[libvirt] [RFC PATCH] maint: let MIN/MAX evaluate only once, on modern compilers
by Eric Blake 04 Jan '19

04 Jan '19

We might as well take advantage of gcc's extensions for a safer MIN()/MAX() macro. Signed-off-by: Eric Blake <eblake(a)redhat.com> --- RFC because we could also try to fall back to older gcc's typeof(a)_a=(a) when the newer __auto_type _a=(a) is not present, and because I don't know how to properly probe for __auto_type and/or typeof support in clang (I know that clang added __auto_type in 2015, per https://reviews.llvm.org/D12686) I also don't know if we have officially stated that we require gcc/clang (because of attribute((cleanup)), which really cannot be emulated without extensions), or if we still try to allow compilation with other compilers; if we have, I don't know if we have declared a minimum compiler version that we demand (qemu has recently demanded gcc 4.8 or clang 3.4 [branded 5.1 on Apple]). src/util/virutil.h | 20 ++++++++++++++++++-- 1 file changed, 18 insertions(+), 2 deletions(-) diff --git a/src/util/virutil.h b/src/util/virutil.h index 908d8920ec..364e354bcb 100644 --- a/src/util/virutil.h +++ b/src/util/virutil.h @@ -28,10 +28,26 @@ # include <sys/types.h> # ifndef MIN -# define MIN(a, b) ((a) < (b) ? (a) : (b)) +# if __GNUC_PREREQ(4, 9) +# define MIN(a, b) ({ \ + __auto_type _a = (a); \ + __auto_type _b = (b); \ + _a < _b ? _a : _b; \ + }) +# else +# define MIN(a, b) ((a) < (b) ? (a) : (b)) +# endif # endif # ifndef MAX -# define MAX(a, b) ((a) > (b) ? (a) : (b)) +# if __GNUC_PREREQ(4, 9) +# define MAX(a, b) ({ \ + __auto_type _a = (a); \ + __auto_type _b = (b); \ + _a > _b ? _a : _b; \ + }) +# else +# define MAX(a, b) ((a) > (b) ? (a) : (b)) +# endif # endif -- 2.20.1

3 4

[libvirt] [PATCH] maint: cfg.mk typo fix
by Eric Blake 04 Jan '19

04 Jan '19

Signed-off-by: Eric Blake <eblake(a)redhat.com> --- Pushing as trivial. cfg.mk | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/cfg.mk b/cfg.mk index 90ec8dfc60..e00213b7dd 100644 --- a/cfg.mk +++ b/cfg.mk @@ -1066,7 +1066,7 @@ sc_prohibit_backslash_alignment: # Some syntax rules pertaining to the usage of cleanup macros # implementing GNU C's cleanup attribute -# Rule to ensure that varibales declared using a cleanup macro are +# Rule to ensure that variables declared using a cleanup macro are # always initialized. sc_require_attribute_cleanup_initialization: @prohibit='VIR_AUTO((FREE|PTR)\(.+\)|CLOSE) *[^=]+;' \ -- 2.20.1

1 0

[libvirt] [PATCH RFC] qemu: caps: invalidate kvm capable qemu binaries on every restart
by Nikolay Shirokovskiy 04 Jan '19

04 Jan '19

If qemu binary exposes kvm then caps depends heavily on host state. We are already check explicitly some host features - kernel version, microcode version, nested flag state. This won't nesserily help if somebody change CPU on host. Let's instead invalidate such caps on every daemon restart. This makes kernel version and microcode version check excessive but nested flag is still need to be checked. Cache invalidating on restart is archived by keeping list of binaries that whose caps cache was validated at least once. If binary in not yet validated and supports kvm then we need to invalidate it's caps cache. Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy(a)virtuozzo.com> --- If the patch will be accepted then I'll send a patch which removes mentioned excessive checks also. src/qemu/qemu_capabilities.c | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+) diff --git a/src/qemu/qemu_capabilities.c b/src/qemu/qemu_capabilities.c index 51bf97b..649aba2 100644 --- a/src/qemu/qemu_capabilities.c +++ b/src/qemu/qemu_capabilities.c @@ -3264,6 +3264,7 @@ struct _virQEMUCapsCachePriv { virArch hostArch; unsigned int microcodeVersion; char *kernelVersion; + virHashTablePtr validated; /* cache whether /dev/kvm is usable as runUid:runGuid */ virTristateBool kvmUsable; @@ -3280,6 +3281,7 @@ virQEMUCapsCachePrivFree(void *privData) VIR_FREE(priv->libDir); VIR_FREE(priv->kernelVersion); + virHashFree(priv->validated); VIR_FREE(priv); } @@ -3952,10 +3954,19 @@ virQEMUCapsIsValid(void *data, bool kvmUsable; struct stat sb; bool kvmSupportsNesting; + bool validated = true; if (!qemuCaps->binary) return true; + if (!virHashLookup(priv->validated, qemuCaps->binary)) { + validated = false; + + /* If we fail to remember this binary is validated it is not problem, + * it will be validated again next time once again */ + virHashAddEntry(priv->validated, qemuCaps->binary, (void*)1); + } + if (qemuCaps->libvirtCtime != virGetSelfLastChanged() || qemuCaps->libvirtVersion != LIBVIR_VERSION_NUMBER) { VIR_DEBUG("Outdated capabilities for '%s': libvirt changed " @@ -4011,6 +4022,12 @@ virQEMUCapsIsValid(void *data, } if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_KVM)) { + if (!validated) { + VIR_DEBUG("Capabilities for QEMU that supports KVM need to be " + "updated after daemon restart"); + return false; + } + if (priv->microcodeVersion != qemuCaps->microcodeVersion) { VIR_DEBUG("Outdated capabilities for '%s': microcode version " "changed (%u vs %u)", @@ -4809,6 +4826,9 @@ virQEMUCapsCacheNew(const char *libDir, if (VIR_STRDUP(priv->libDir, libDir) < 0) goto error; + if (!(priv->validated = virHashCreate(1, NULL))) + goto error; + priv->hostArch = virArchFromHost(); priv->runUid = runUid; -- 1.8.3.1

2 1

[libvirt] [PATCH] docs: schemas: Decouple the virtio options from each other
by Erik Skultety 04 Jan '19

04 Jan '19

Currently, all of the VirtioOptions are under a single <optional> element, however, neither our parser/formatter or QEMU driver requires the presence of all the options if only a single one from the set has been specified, so fix it and silence the schema validator. Signed-off-by: Erik Skultety <eskultet(a)redhat.com> --- docs/schemas/domaincommon.rng | 2 ++ 1 file changed, 2 insertions(+) diff --git a/docs/schemas/domaincommon.rng b/docs/schemas/domaincommon.rng index 6518aff73a..aa50eac424 100644 --- a/docs/schemas/domaincommon.rng +++ b/docs/schemas/domaincommon.rng @@ -5498,6 +5498,8 @@ <attribute name="iommu"> <ref name="virOnOff"/> </attribute> + </optional> + <optional> <attribute name="ats"> <ref name="virOnOff"/> </attribute> -- 2.20.1

2 1