March 2019 - Devel - libvirt List Archives

[libvirt] AMD SEV's /dev/sev permissions and probing QEMU for capabilities
by Erik Skultety 15 Mar '19

15 Mar '19

Hi, this is a summary of a private discussion I've had with guys CC'd on this email about finding a solution to [1] - basically, the default permissions on /dev/sev (below) make it impossible to query for SEV platform capabilities, since by default we run QEMU as qemu:qemu when probing for capabilities. It's worth noting is that this is only relevant to probing, since for a proper QEMU VM we create a mount namespace for the process and chown all the nodes (needs a SEV fix though). # ll /dev/sev crw-------. 1 root root I suggested either force running QEMU as root for probing (despite the obvious security implications) or using namespaces for probing too. Dan argued that this would have a significant perf impact and suggested we ask systemd to add a global udev rule. I proceeded with cloning [1] to systemd and creating an udev rule that I planned on submitting to systemd upstream - the initial idea was to mimic /dev/kvm and make it world accessible to which Brijesh from AMD expressed a concern that regular users might deplete the resources (limit on the number of guests allowed by the platform). But since the limit is claimed to be around 4, Dan discouraged me to continue with restricting the udev rule to only the 'kvm' group which Laszlo suggested earlier as the limit is so small that a malicious QEMU could easily deplete this during probing. This fact also ruled out any kind of ACL we could create dynamically. Instead, he suggested that we filter out the kvm-capable QEMU and put only that one in the namespace without a significant perf impact. - my take on this is that there could potentially be more than a single kvm-enabled QEMU and therefore we'd need to create more than just a single namespace. - I also argued that I can image that the same kind of DOS attack might be possible from within the namespace, even if we created the /dev/sev node only in SEV-enabled guests (which we currently don't). All of us have agreed that allowing /dev/sev in the namespace for only SEV-enabled guests is worth doing nonetheless. In the meantime, Christophe went through the kernel code to verify how the SEV resources are managed and what protection is currently in place to mitigate the chance of a process easily depleting the limit on SEV guests. He found that ASID, which determines the encryption key, is allocated from a single ASID bitmap and essentially guarded by a single 'sev->active' flag. So, in conclusion, we absolutely need input from Brijesh (AMD) whether there was something more than the low limit on number of guests behind the default permissions. Also, we'd like to get some details on how the limit is managed, helping to assess the approaches mentioned above. Thanks and please do share your ideas, Erik [1] https://bugzilla.redhat.com/show_bug.cgi?id=1665400 [2] https://bugzilla.redhat.com/show_bug.cgi?id=1561113

4 25

[libvirt] [dockerfiles PATCH 0/3] refresh: Minor Python style tweaks
by Andrea Bolognani 15 Mar '19

15 Mar '19

These are basically the same tweaks that have recently been applied to lcitool. Andrea Bolognani (3): refresh: Remove double space refresh: Drop Error class refresh: Move exception handling to Application.run() refresh | 40 +++++++++++++++++----------------------- 1 file changed, 17 insertions(+), 23 deletions(-) -- 2.20.1

2 4

[libvirt] [PATCH 0/4] qemu: hotplug: Clean up and fix disk device unplug (blockdev-add saga)
by Peter Krempa 15 Mar '19

15 Mar '19

Refactor checking of disk bus and type for disk unplug and fix checking whether a virtio disk has a running block job by deleting the almost-identical implementation specific to virtio disks. Peter Krempa (4): qemu: hotplug: Remove 'ret' variable in qemuDomainDetachDeviceDiskLive qemu: hotplug: Use typecasted enum in qemuDomainDetachDeviceDiskLive qemu: hotplug: Use switch statement for selecting disk bus function qemu: hotplug: Merge virtio and non-virtio disk unplug code src/qemu/qemu_hotplug.c | 87 ++++++++++++++++++----------------------- 1 file changed, 39 insertions(+), 48 deletions(-) -- 2.20.1

2 8

[libvirt] [PATCH] snapshots: More clarification about REDEFINE
by Eric Blake 15 Mar '19

15 Mar '19

Based on recent list questions about the proposed addition of virDomainCheckpointCreateXML(REDEFINE), it is worth adding some clarification to the existing snapshot redefine documentation that is serving as the basis for checkpoints. Normal snapshot creation requires very few elements from the user XML (libvirt can pick sane defaults for items that are omitted, and many fields, including <domain> are documented as readonly output fields ignored on input). But during REDEFINE, the API wants the complete XML produced by an earlier virDomainSnapshotGetXMLDesc, which includes a <domain> sub-element capturing the state a the time the snapshot was first created. As the domain state has likely changed in the meantime, omitting <domain> during REDEFINE is extremely risky (to the point that virDomainSnapshotRevert() requires the use of FORCE flag to use such a snapshot) - we only support it because of backwards-compatibility to snapshots created before 0.9.5 started capturing <domain>. When checkpoints are introduced, the <domain> element will be mandatory in REDEFINE. Note that because <domain> can be potentially huge, we also are unable to include a bulk listing or redefine of all <domainsnapshot>s for a single domain in one XML dump (for example, a 1M <domain> XML * 16 snapshots explodes into 16M in a bulk form, which gets difficult to send over RPC). Perhaps we could add a flag to request that the <domain> sub-element be omitted on output, but such output is no longer suitable for sane REDEFINE input. Signed-off-by: Eric Blake <eblake(a)redhat.com> --- Question: 0.9.5 is so long ago, is it worth reworking the qemu snapshot code to declare that REDEFINE will no longer import a snapshot definition that lacks a <domain> subelement? (Note: the vbox driver also supports snapshots and the REDEFINE flag, but does not output <domain> and thus should still allow it to be omitted on redefine). The current API wording DOES have the disclaimer regarding REDEFINE that "the hypervisor may validate that reverting to the snapshot appears to be possible", which we can use as our escape clause whereby the modern qemu driver is merely adding a validation that <domain> must be present for the redefine to be viable, even though it is at odds with our usual efforts to always parse XML that used to work in earlier versions. If we make my proposed followup change, and someone has kept a VM definition with snapshots alive since libvirt pre-0.9.5, updating to modern libvirt would lose their old snapshots. Then again, if your guest has been around that long, do you REALLY want to revert to a snapshot from that many years ago? docs/formatsnapshot.html.in | 3 ++- src/libvirt-domain-snapshot.c | 24 +++++++++++++++--------- 2 files changed, 17 insertions(+), 10 deletions(-) diff --git a/docs/formatsnapshot.html.in b/docs/formatsnapshot.html.in index c60b4fb7c9..99addc6675 100644 --- a/docs/formatsnapshot.html.in +++ b/docs/formatsnapshot.html.in @@ -79,7 +79,8 @@ redefining a snapshot (<span class="since">since 0.9.5</span>), with the <code>VIR_DOMAIN_SNAPSHOT_CREATE_REDEFINE</code> flag of <code>virDomainSnapshotCreateXML()</code>, all of the XML - described here is relevant. + described here is relevant on input, even the fields that are + normally described as readonly for output. </p> <p> Snapshots are maintained in a hierarchy. A domain can have a diff --git a/src/libvirt-domain-snapshot.c b/src/libvirt-domain-snapshot.c index 947547627a..be9bf71af9 100644 --- a/src/libvirt-domain-snapshot.c +++ b/src/libvirt-domain-snapshot.c @@ -117,15 +117,21 @@ virDomainSnapshotGetConnect(virDomainSnapshotPtr snapshot) * * If @flags includes VIR_DOMAIN_SNAPSHOT_CREATE_REDEFINE, then this * is a request to reinstate snapshot metadata that was previously - * discarded, rather than creating a new snapshot. This can be used - * to recreate a snapshot hierarchy on a destination, then remove it - * on the source, in order to allow migration (since migration - * normally fails if snapshot metadata still remains on the source - * machine). When redefining snapshot metadata, the current snapshot - * will not be altered unless the VIR_DOMAIN_SNAPSHOT_CREATE_CURRENT - * flag is also present. It is an error to request the - * VIR_DOMAIN_SNAPSHOT_CREATE_CURRENT flag without - * VIR_DOMAIN_SNAPSHOT_CREATE_REDEFINE. On some hypervisors, + * captured from virDomainSnapshotGetXMLDesc() before removing that + * metadata, rather than creating a new snapshot. This can be used to + * recreate a snapshot hierarchy on a destination, then remove it on + * the source, in order to allow migration (since migration normally + * fails if snapshot metadata still remains on the source machine). + * Note that while original creation can omit a number of elements + * from @xmlDesc (and libvirt will supply sane defaults based on the + * domain state at that point in time), a redefinition must supply + * more elements (as the domain may have changed in the meantime, so + * that libvirt no longer has a way to resupply correct + * defaults). When redefining snapshot metadata, the domain's current + * snapshot will not be altered unless the + * VIR_DOMAIN_SNAPSHOT_CREATE_CURRENT flag is also present. It is an + * error to request the VIR_DOMAIN_SNAPSHOT_CREATE_CURRENT flag + * without VIR_DOMAIN_SNAPSHOT_CREATE_REDEFINE. On some hypervisors, * redefining an existing snapshot can be used to alter host-specific * portions of the domain XML to be used during revert (such as * backing filenames associated with disk devices), but must not alter -- 2.20.1

2 2

[libvirt] [PATCH] virobject: Improve documentation
by Eric Blake 15 Mar '19

15 Mar '19

I had to inspect the code to learn whether a final virObjectUnref() calls ALL dispose callbacks in child-to-parent order (akin to C++ destructors), or whether I manually had to call a parent-class dispose when writing a child class dispose method. The answer is the former. (Thankfully, since VIR_FREE wipes out pointers for safety, even if I had guessed wrong, I probably would not have tripped over a double-free fault when the parent dispose ran for the second time). While at it, the VIR_CLASS_NEW macro requires that the virObject component at offset 0 be reached through the name 'parent', not 'object'. Signed-off-by: Eric Blake <eblake(a)redhat.com> --- src/util/virobject.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/src/util/virobject.c b/src/util/virobject.c index f08c18ce44..462b3d7d3f 100644 --- a/src/util/virobject.c +++ b/src/util/virobject.c @@ -147,10 +147,11 @@ virClassForObjectRWLockable(void) * * Register a new object class with @name. The @objectSize * should give the total size of the object struct, which - * is expected to have a 'virObject object;' field as its - * first member. When the last reference on the object is - * released, the @dispose callback will be invoked to free - * memory of the object fields + * is expected to have a 'virObject parent;' field as (or + * contained in) its first member. When the last reference + * on the object is released, the @dispose callback will be + * invoked to free memory of the local object fields, as + * well as dispose callbacks of the parent classes * * Returns a new class instance */ -- 2.20.1

2 2

[libvirt] [jenkins-ci PATCH] lcitool: add special case for Debian i386/i686 mixup
by Daniel P. Berrangé 15 Mar '19

15 Mar '19

Although the toolchain binaries for 32-bit x86 are using the ABI i686-linux-gnu, the library install directory is strangely named with the different i386-linux-gnu. Signed-off-by: Daniel P. Berrangé <berrange(a)redhat.com> --- guests/lcitool | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) diff --git a/guests/lcitool b/guests/lcitool index 76b2dc7..0f60704 100755 --- a/guests/lcitool +++ b/guests/lcitool @@ -101,6 +101,13 @@ class Util: raise Exception("Unsupported architecture {}".format(native_arch)) return archmap[native_arch] + @staticmethod + def native_arch_to_lib(native_arch): + arch = Util.native_arch_to_abi(native_arch) + if arch == "i686-linux-gnu": + arch = "i386-linux-gnu" + return arch + @staticmethod def native_arch_to_deb_arch(native_arch): archmap = { @@ -640,12 +647,14 @@ class Application: if args.cross_arch: deb_arch = Util.native_arch_to_deb_arch(args.cross_arch) abi = Util.native_arch_to_abi(args.cross_arch) + lib = Util.native_arch_to_lib(args.cross_arch) gcc = "gcc-" + abi varmap["cross_arch"] = deb_arch pkg_names = [p + ":" + deb_arch for p in cross_pkgs.values()] pkg_names.append(gcc) varmap["cross_pkgs"] = " \\\n ".join(sorted(set(pkg_names))) varmap["cross_abi"] = abi + varmap["cross_lib"] = lib sys.stdout.write(textwrap.dedent(""" RUN export DEBIAN_FRONTEND=noninteractive && \\ @@ -673,7 +682,7 @@ class Application: ENV ABI "{cross_abi}" ENV CONFIGURE_OPTS "--host={cross_abi} \\ --target={cross_abi}" - ENV PKG_CONFIG_LIBDIR "/usr/lib/{cross_abi}/pkgconfig" + ENV PKG_CONFIG_LIBDIR "/usr/lib/{cross_lib}/pkgconfig" """).format(**varmap)) elif package_format == "rpm": if os_name == "Fedora" and os_version == "Rawhide": -- 2.20.1

3 5

[libvirt] [PATCH v3 0/5] qemu_hotplug: Fix a rare race condition when detaching a device twice
by Michal Privoznik 15 Mar '19

15 Mar '19

v3 of: https://www.redhat.com/archives/libvir-list/2019-March/msg00924.html diff to v2: - Dropped @enterMonitor from qemuDomainDeleteDevice, - Simplified event checking since @vm is always locked after @enterMonitor is dropped, - Switched to plain qemuMonitorDelDevice for zpci device - Added comment about error reporting to 3/5 Note that 3/5 was ACKed in v2. Michal Prívozník (5): qemu_hotplug: Introduce and use qemuDomainDeleteDevice DO NOT APPLY: Simple reproducer qemuMonitorJSONDelDevice: Return -2 on DeviceNotFound error qemu_hotplug: Fix a rare race condition when detaching a device twice DO NOT APPLY: Revert simple reproducer src/qemu/qemu_domain.h | 1 + src/qemu/qemu_hotplug.c | 278 +++++++++++++++-------------------- src/qemu/qemu_monitor.c | 10 ++ src/qemu/qemu_monitor_json.c | 5 + 4 files changed, 134 insertions(+), 160 deletions(-) -- 2.19.2

2 7

[libvirt] [PATCH 0/3] Drop Upstart support
by Andrea Bolognani 15 Mar '19

15 Mar '19

Andrea Bolognani (3): travis: Use 'redhat' init script on CentOS 7 Drop Upstart support news: Document Upstart support being dropped .travis.yml | 2 +- docs/news.xml | 11 +++++++++ m4/virt-init-script.m4 | 7 +----- src/Makefile.am | 29 ----------------------- src/remote/Makefile.inc.am | 2 -- src/remote/libvirtd.upstart | 47 ------------------------------------- 6 files changed, 13 insertions(+), 85 deletions(-) delete mode 100644 src/remote/libvirtd.upstart -- 2.20.1

2 9

[libvirt] [PATCH v2 0/4] tools: console: a fix and improvment
by Nikolay Shirokovskiy 15 Mar '19

15 Mar '19

Jonh, I split the 1st patch the other way to avoid changing same lines of code in subsequent patches. Diff to v1[1] - split 1st patch into 3 - change wordings of commit messages/error messages - add minor changes to code as suggested by John [1] https://www.redhat.com/archives/libvir-list/2019-February/msg00920.html Nikolay Shirokovskiy (4): tools: console: cleanup console on errors in main thread tools: console: add missing locks in callbacks tools: console: check if console was shutdown in callbacks tools: console: pass stream/fd errors to user tools/virsh-console.c | 224 ++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 171 insertions(+), 53 deletions(-) -- 1.8.3.1

2 5

[libvirt] fix method delete() in src/main/java/org/libvirt/DomainSnapshot.java
by 傅姜晨 15 Mar '19

15 Mar '19

if VDSP is set to null during the method delete(), Object snapshot can't be freed in method free() in the same class. Showing 1 changed file with 1 addition and 7 deletions. View file 8 src/main/java/org/libvirt/DomainSnapshot.java | | @@ -35,13 +35,7 @@ public DomainSnapshot(Connect virConnect, DomainSnapshotPointer VDSP) { | | | | | * @throws LibvirtException | | | | | */ | | | | | publicintdelete(intflags) throwsLibvirtException { | | | | | int success =0; | | | | | if (VDSP!=null) { | | | | | success = processError(libvirt.virDomainSnapshotDelete(VDSP, flags)); | | | | | VDSP=null; | | | | | } | | | | | | | | | | return success; | | | | | return processError(libvirt.virDomainSnapshotDelete(VDSP, flags)); | | | | | } | | | | | | | | | | @Override |

2 1