December 2015 - Devel - libvirt List Archives

[libvirt] ANNOUNCE: libvirt 1.2.18.2 maintenance release
by Cole Robinson 24 Dec '15

24 Dec '15

libvirt 1.2.18.2 is now available. This is a maintenance release of libvirt 1.2.18 with additional bugfixes that have accumulated upstream since the initial release. This release can be downloaded at: http://libvirt.org/sources/stable_updates/libvirt-1.2.18.2.tar.gz Changes in this version: * Fix a trailing space in spec file * virsh: report errors for empty strings * bridge: check for invalid MAC in networkGetDHCPLeases * Enhance documentation of virDomainDetachDevice * apparmor: add missing qemu binaries * qemu: Use live autoNodeset when numatune placement is auto * Close the source fd if the destination qemu exits during tunnelled migration * storage: Fix incorrect format for <disk> <auth> XML * virt-host-validate: distinguish exists vs accessible for devices * spec: Delete .git after applying patches * apparmor: differentiate between error and unconfined profiles * storage: Adjust calculation of alloc/capacity for disk * qemu: Add conditions for qemu-kvm use on ppc64 * rpc: libssh2: Fix regression in ssh host key verification * rpc: libssh2: Add more debugging info * Update pool allocation with new values on volume creation * Use daemon log facility for journald * virDomainCreateXML: Make domain definition transient * virDomainCreateXML: Don't remove persistent domains on error * qemu: Refresh memory size only on fresh starts * domain: Fix migratable XML with graphics/@listen * tpm: adapt sysfs cancel path for new TPM driver * libvirt-guests: Disable shutdown timeout * systemd: Escape only needed characters for machined * systemd: Escape machine name for machined * CVE-2015-5313: storage: don't allow '/' in filesystem volume names * docs: event impl. registration before hypervisor connection * spec: Fix some warnings with latest rpmbuild * qemu: Fix dynamic_ownership qemu.conf setting For info about past maintenance releases, see: http://wiki.libvirt.org/page/Maintenance_Releases Thanks, Cole

1 0

[libvirt] ANNOUNCE: libvirt 1.2.13.2 maintenance release
by Cole Robinson 24 Dec '15

24 Dec '15

libvirt 1.2.13.2 is now available. This is a maintenance release of libvirt 1.2.13 with additional bugfixes that have accumulated upstream since the initial release. This release can be downloaded at: http://libvirt.org/sources/stable_updates/libvirt-1.2.13.2.tar.gz Changes in this version: * spec: Delete .git after applying patches * qemu: block-commit: Mark disk in block jobs only on successful command * qemu: Disallow concurrent block jobs on a single disk * qemu: event: Don't fiddle with disk backing trees without a job * qemu: process: Export qemuProcessFindDomainDiskByAlias * spec: Fix polkit dep on F23 * domain: Fix migratable XML with graphics/@listen * qemu: hotplug: Properly clean up drive backend if frontend hotplug fails * tpm: adapt sysfs cancel path for new TPM driver * libvirt-guests: Disable shutdown timeout * systemd: Escape only needed characters for machined * systemd: Escape machine name for machined * cgroup: Drop resource partition from virSystemdMakeScopeName * CVE-2015-5313: storage: don't allow '/' in filesystem volume names * remoteClientCloseFunc: Don't mangle connection object refcount * Revert "LXC: show used memory as 0 when domain is not active" * lxc: Don't pass a local variable address randomly * lxc: set nosuid+nodev+noexec flags on /proc/sys mount * virnetdev: fix moving of 802.11 phys * interface: don't error out if a bond has no interfaces * lxc: don't up the veth interfaces unless explicitly asked to * tests: Add virnetdevtestdata to EXTRA_DIST * lxc: move wireless PHYs to a network namespace * Cleanup "/sys/class/net" usage * Introduce virnetdevtest * build: provide virNetDevSysfsFile on non-Linux For info about past maintenance releases, see: http://wiki.libvirt.org/page/Maintenance_Releases Thanks, Cole

1 0

[libvirt] [PATCH 0/3] 'Append' mode for serial ports output to file
by Dmitry Mishin 23 Dec '15

23 Dec '15

By default, QEMU truncates serial file on open. Sometimes, it could be weird - for example, when we are trying to investigate some event, which occured several restarts ago. This patch setadds an ability to preserve previous content. Dmitry Mishin (3): tests: add qemu 2.6 caps test conf: Add mode 'append' for chardevs with file source qemu: Process new mode='append' for char dev with output to a file src/conf/domain_conf.c | 18 +- src/conf/domain_conf.h | 1 + src/qemu/qemu_capabilities.c | 2 + src/qemu/qemu_capabilities.h | 1 + src/qemu/qemu_command.c | 10 + tests/qemucapabilitiesdata/caps_2.6.0-1.caps | 173 + tests/qemucapabilitiesdata/caps_2.6.0-1.replies | 4043 +++++++++++++++++++++++ tests/qemucapabilitiestest.c | 1 + 8 files changed, 4245 insertions(+), 4 deletions(-) create mode 100644 tests/qemucapabilitiesdata/caps_2.6.0-1.caps create mode 100644 tests/qemucapabilitiesdata/caps_2.6.0-1.replies -- 1.8.3.1

2 4

[libvirt] [PATCHv2 0/3] 'autodeflate' attribute for mememory balloon
by Dmitry Andreev 23 Dec '15

23 Dec '15

This patch-set adds support for the new feature 'deflate-on-oom' that has been added to the QEMU virtio memory balloon a month ago. This feature lets a guest OS deflate balloon on OOM. With 'deflate-on-oom' enabled virtio balloon will release it's memory if possible right before an OOM termination of processes. QEMU commit with the complete feature description: http://git.qemu.org/?p=qemu.git;a=commit;h=e3816255bf4b6377bb405331e2ee0dc1… v2: * attribute renamed to 'autodeflate' [1/3][3/3] * capability check added [2/3][2/3] * ability to change 'autodeflate' for running domain wasn't added because QEMU doesn't allow to on/off 'deflate-on-oom' in runtime * more tests added [3/3] *** BLURB HERE *** Dmitry Andreev (3): conf: introduce 'autodeflate' attribute for memballoon device qemu: add capability check for memballoon 'autodeflate' feature qemu: add support of optional 'autodeflate' attribute docs/formatdomain.html.in | 10 ++ docs/schemas/domaincommon.rng | 5 + src/conf/domain_conf.c | 23 +++ src/conf/domain_conf.h | 1 + src/qemu/qemu_capabilities.c | 11 ++ src/qemu/qemu_capabilities.h | 2 + src/qemu/qemu_command.c | 11 ++ tests/qemucapabilitiesdata/caps_1.2.2-1.replies | 68 ++++++++- tests/qemucapabilitiesdata/caps_1.3.1-1.replies | 68 ++++++++- tests/qemucapabilitiesdata/caps_1.4.2-1.replies | 68 ++++++++- tests/qemucapabilitiesdata/caps_1.5.3-1.replies | 66 ++++++++- tests/qemucapabilitiesdata/caps_1.6.0-1.replies | 66 ++++++++- tests/qemucapabilitiesdata/caps_1.6.50-1.replies | 66 ++++++++- tests/qemucapabilitiesdata/caps_2.1.1-1.replies | 78 +++++++++- tests/qemucapabilitiesdata/caps_2.4.0-1.caps | 1 + tests/qemucapabilitiesdata/caps_2.4.0-1.replies | 146 ++++++++++++++++++- tests/qemucapabilitiesdata/caps_2.5.0-1.caps | 1 + tests/qemucapabilitiesdata/caps_2.5.0-1.replies | 161 ++++++++++++++++++++- .../qemuxml2argv-balloon-ccw-deflate.args | 20 +++ .../qemuxml2argv-balloon-ccw-deflate.xml | 21 +++ .../qemuxml2argv-balloon-device-deflate-off.args | 23 +++ .../qemuxml2argv-balloon-device-deflate-off.xml | 25 ++++ .../qemuxml2argv-balloon-device-deflate.args | 23 +++ .../qemuxml2argv-balloon-device-deflate.xml | 25 ++++ .../qemuxml2argv-balloon-mmio-deflate.args | 25 ++++ .../qemuxml2argv-balloon-mmio-deflate.xml | 30 ++++ tests/qemuxml2argvtest.c | 9 ++ 27 files changed, 990 insertions(+), 63 deletions(-) create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-balloon-ccw-deflate.args create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-balloon-ccw-deflate.xml create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-balloon-device-deflate-off.args create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-balloon-device-deflate-off.xml create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-balloon-device-deflate.args create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-balloon-device-deflate.xml create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-balloon-mmio-deflate.args create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-balloon-mmio-deflate.xml -- 1.8.3.1

1 0

[libvirt] [PATCH 0/3] vz: allow to create CT from template and boot VM from cdrom
by Mikhail Feoktistov 22 Dec '15

22 Dec '15

Mikhail Feoktistov (3): vz: allow to create container based on template vz: allow to boot VM from cdrom vz: implementation of boot device order src/vz/vz_sdk.c | 174 ++++++++++++++++++++++++++++++++++++++++---------------- 1 file changed, 125 insertions(+), 49 deletions(-) -- 1.8.3.1

2 6

[libvirt] [PATCH] qemu: Connect to guest agent iff needed
by Michal Privoznik 22 Dec '15

22 Dec '15

https://bugzilla.redhat.com/show_bug.cgi?id=1293351 I've came across a bit of a silly scenario, but long story short: after domain was resumed, the virDomainSetTime() API hung for 5 seconds after which it failed with an error. Problem was, that there is no guest agent installed in the domain. But this got me thinking and digging into the code. So even though we do listen to VSERPORT_CHANGE events and connect and disconnect ourselves to guest agent, we still do connect to the guest agent at both domain startup and resume. This is a bit silly as it produces the delay - basically, because we connect to the guest agent, priv->agent is not NULL. Therefore qemuDomainAgentAvailable() will return true. And the place where we hang? Well, historically, when there was no VSERPORT_CHANGE event, we used a dummy ping command to the guest which has 5 seconds timeout. And it's still there and effective. So there's where the delay comes from. What's the resolution? Well, I'm introducing new capability QEMU_CAPS_VSERPORT_CHANGE that is enabled iff qemu is capable of the event. And, during domain startup, reconnect after resume and attach we do connect to the agent socket iff the capability is NOT set. Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com> --- src/qemu/qemu_capabilities.c | 2 + src/qemu/qemu_capabilities.h | 1 + src/qemu/qemu_process.c | 66 ++++++++++++++++++---------- tests/qemucapabilitiesdata/caps_2.1.1-1.caps | 1 + tests/qemucapabilitiesdata/caps_2.4.0-1.caps | 1 + tests/qemucapabilitiesdata/caps_2.5.0-1.caps | 1 + 6 files changed, 48 insertions(+), 24 deletions(-) diff --git a/src/qemu/qemu_capabilities.c b/src/qemu/qemu_capabilities.c index 6e5d203..9e11af9 100644 --- a/src/qemu/qemu_capabilities.c +++ b/src/qemu/qemu_capabilities.c @@ -308,6 +308,7 @@ VIR_ENUM_IMPL(virQEMUCaps, QEMU_CAPS_LAST, "virtio-tablet", /* 205 */ "virtio-input-host", + "vserport-change-event", ); @@ -1479,6 +1480,7 @@ struct virQEMUCapsStringFlags virQEMUCapsEvents[] = { { "SPICE_MIGRATE_COMPLETED", QEMU_CAPS_SEAMLESS_MIGRATION }, { "DEVICE_DELETED", QEMU_CAPS_DEVICE_DEL_EVENT }, { "MIGRATION", QEMU_CAPS_MIGRATION_EVENT }, + { "VSERPORT_CHANGE", QEMU_CAPS_VSERPORT_CHANGE }, }; struct virQEMUCapsStringFlags virQEMUCapsObjectTypes[] = { diff --git a/src/qemu/qemu_capabilities.h b/src/qemu/qemu_capabilities.h index 61d6379..983faf6 100644 --- a/src/qemu/qemu_capabilities.h +++ b/src/qemu/qemu_capabilities.h @@ -335,6 +335,7 @@ typedef enum { /* 205 */ QEMU_CAPS_VIRTIO_TABLET, /* -device virtio-tablet-{device,pci} */ QEMU_CAPS_VIRTIO_INPUT_HOST, /* -device virtio-input-host-{device,pci} */ + QEMU_CAPS_VSERPORT_CHANGE, /* VSERPORT_CHANGE event */ QEMU_CAPS_LAST /* this must always be the last item */ } virQEMUCapsFlags; diff --git a/src/qemu/qemu_process.c b/src/qemu/qemu_process.c index f274068..95a795c 100644 --- a/src/qemu/qemu_process.c +++ b/src/qemu/qemu_process.c @@ -3524,15 +3524,21 @@ qemuProcessReconnect(void *opaque) if (qemuConnectMonitor(driver, obj, QEMU_ASYNC_JOB_NONE, NULL) < 0) goto error; - /* Failure to connect to agent shouldn't be fatal */ - if ((ret = qemuConnectAgent(driver, obj)) < 0) { - if (ret == -2) - goto error; + /* With newer qemus capable of VSERPORT_CHANGE event, we are connecting and + * disconnecting to the guest agent as it connects or disconnects to the + * channel within the guest. So, it's important to connect here iff we are + * running qemu not capable of the event. */ + if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_VSERPORT_CHANGE)) { + /* Failure to connect to agent shouldn't be fatal */ + if ((ret = qemuConnectAgent(driver, obj)) < 0) { + if (ret == -2) + goto error; - VIR_WARN("Cannot connect to QEMU guest agent for %s", - obj->def->name); - virResetLastError(); - priv->agentError = true; + VIR_WARN("Cannot connect to QEMU guest agent for %s", + obj->def->name); + virResetLastError(); + priv->agentError = true; + } } if (qemuHostdevUpdateActiveDomainDevices(driver, obj->def) < 0) @@ -4949,15 +4955,21 @@ qemuProcessLaunch(virConnectPtr conn, if (qemuProcessWaitForMonitor(driver, vm, asyncJob, priv->qemuCaps, logCtxt) < 0) goto cleanup; - /* Failure to connect to agent shouldn't be fatal */ - if ((rv = qemuConnectAgent(driver, vm)) < 0) { - if (rv == -2) - goto cleanup; + /* With newer qemus capable of VSERPORT_CHANGE event, we are connecting and + * disconnecting to the guest agent as it connects or disconnects to the + * channel within the guest. So, it's important to connect here iff we are + * running qemu not capable of the event. */ + if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_VSERPORT_CHANGE)) { + /* Failure to connect to agent shouldn't be fatal */ + if ((rv = qemuConnectAgent(driver, vm)) < 0) { + if (rv == -2) + goto cleanup; - VIR_WARN("Cannot connect to QEMU guest agent for %s", - vm->def->name); - virResetLastError(); - priv->agentError = true; + VIR_WARN("Cannot connect to QEMU guest agent for %s", + vm->def->name); + virResetLastError(); + priv->agentError = true; + } } VIR_DEBUG("Detecting if required emulator features are present"); @@ -5664,15 +5676,21 @@ int qemuProcessAttach(virConnectPtr conn ATTRIBUTE_UNUSED, if (qemuProcessWaitForMonitor(driver, vm, QEMU_ASYNC_JOB_NONE, priv->qemuCaps, NULL) < 0) goto error; - /* Failure to connect to agent shouldn't be fatal */ - if ((ret = qemuConnectAgent(driver, vm)) < 0) { - if (ret == -2) - goto error; + /* With newer qemus capable of VSERPORT_CHANGE event, we are connecting and + * disconnecting to the guest agent as it connects or disconnects to the + * channel within the guest. So, it's important to connect here iff we are + * running qemu not capable of the event. */ + if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_VSERPORT_CHANGE)) { + /* Failure to connect to agent shouldn't be fatal */ + if ((ret = qemuConnectAgent(driver, vm)) < 0) { + if (ret == -2) + goto error; - VIR_WARN("Cannot connect to QEMU guest agent for %s", - vm->def->name); - virResetLastError(); - priv->agentError = true; + VIR_WARN("Cannot connect to QEMU guest agent for %s", + vm->def->name); + virResetLastError(); + priv->agentError = true; + } } VIR_DEBUG("Detecting VCPU PIDs"); diff --git a/tests/qemucapabilitiesdata/caps_2.1.1-1.caps b/tests/qemucapabilitiesdata/caps_2.1.1-1.caps index 1098dcf..332b85a 100644 --- a/tests/qemucapabilitiesdata/caps_2.1.1-1.caps +++ b/tests/qemucapabilitiesdata/caps_2.1.1-1.caps @@ -159,4 +159,5 @@ <flag name='rtl8139'/> <flag name='e1000'/> <flag name='virtio-net'/> + <flag name='vserport-change-event'/> </qemuCaps> diff --git a/tests/qemucapabilitiesdata/caps_2.4.0-1.caps b/tests/qemucapabilitiesdata/caps_2.4.0-1.caps index d67a48d..1a5fe81 100644 --- a/tests/qemucapabilitiesdata/caps_2.4.0-1.caps +++ b/tests/qemucapabilitiesdata/caps_2.4.0-1.caps @@ -167,4 +167,5 @@ <flag name='virtio-mouse'/> <flag name='virtio-tablet'/> <flag name='virtio-input-host'/> + <flag name='vserport-change-event'/> </qemuCaps> diff --git a/tests/qemucapabilitiesdata/caps_2.5.0-1.caps b/tests/qemucapabilitiesdata/caps_2.5.0-1.caps index f4f3673..8b3586a 100644 --- a/tests/qemucapabilitiesdata/caps_2.5.0-1.caps +++ b/tests/qemucapabilitiesdata/caps_2.5.0-1.caps @@ -168,4 +168,5 @@ <flag name='virtio-mouse'/> <flag name='virtio-tablet'/> <flag name='virtio-input-host'/> + <flag name='vserport-change-event'/> </qemuCaps> -- 2.4.10

1 1

[libvirt] [PATCH 0/3] Fix mingw build
by Michal Privoznik 22 Dec '15

22 Dec '15

We have a build failure on mingw platform:4 https://ci.centos.org/job/libvirt-mingw/899/console Michal Privoznik (3): sysconf: Include unistd.h tools: Include PIE_LDFLAGS at the correct place tools: Disable virt-login-shell on mingw m4/virt-login-shell.m4 | 18 +++++++++++++++--- src/nodeinfo.c | 1 + src/qemu/qemu_conf.h | 2 ++ src/util/vircgroup.c | 1 + src/util/vircommand.c | 1 + tools/Makefile.am | 4 ++-- tools/virt-login-shell.c | 1 + 7 files changed, 23 insertions(+), 5 deletions(-) -- 2.4.10

2 4

[libvirt] Global domain cpu.cfs_quota_us setup
by Alexander Burluka 21 Dec '15

21 Dec '15

Hello all! I would greatly appreciate your help with some cgroup question. Libvirt xml has cpu tuning section and quota element in it. That parameter sets up cpu.cfs_cgroup_us for every domain's vcpu. I set it to 55555 and has such hierarchy of cgroups: cpu.cfs_quota_us (-1) | |-emulator/cpu.cfs_quota_us (-1) |-vcpu0/cpu.cfs_quota_us (55555) |-vcpu1/cpu.cfs_quota_us (55555) |-vcpu2/cpu.cfs_quota_us (55555) .... \-vcpuN/cpu.cfs_quota_us (55555) Is there any way to set up upper level quota limit via libvirt without setting every vcpu? So I'd like to get next hierarchy: cpu.cfs_quota_us (55555) | |-emulator/cpu.cfs_quota_us (-1) |-vcpu0/cpu.cfs_quota_us (-1) |-vcpu1/cpu.cfs_quota_us (-1) |-vcpu2/cpu.cfs_quota_us (-1) .... \-vcpuN/cpu.cfs_quota_us (-1) Thank you! -- Regards, Alexander Burluka

2 4

[libvirt] [PATCH] virStorageVolWipe: Document that wiping journaled FS is useless
by Michal Privoznik 21 Dec '15

21 Dec '15

So you have a libvirt volume that you want to wipe out. But lets say that the volume is actually a file stored on a journaled filesystem. Overwriting it with zeroes or a pattern does not mean that corresponding physical location on the disk is overwritten too, due to journaling. It's the same story with network based volumes, copy-on-write filesystems, and so on. Since there is no way that an userland application can write onto specific areas on disk, all that we can do is document the fact. Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com> --- src/libvirt-storage.c | 16 +++++++++++++--- 1 file changed, 13 insertions(+), 3 deletions(-) diff --git a/src/libvirt-storage.c b/src/libvirt-storage.c index 66dd9f0..513c72f 100644 --- a/src/libvirt-storage.c +++ b/src/libvirt-storage.c @@ -1725,7 +1725,12 @@ virStorageVolDelete(virStorageVolPtr vol, * @vol: pointer to storage volume * @flags: extra flags; not used yet, so callers should always pass 0 * - * Ensure data previously on a volume is not accessible to future reads + * Ensure data previously on a volume is not accessible to future + * reads. Also note, that depending on the actual volume + * representation, this call may not really overwrite the + * physical location of the volume. For instance, files stored + * journaled, log structured, copy-on-write, versioned, and + * network file systems are known to be problematic. * * Returns 0 on success, or -1 on error */ @@ -1765,8 +1770,13 @@ virStorageVolWipe(virStorageVolPtr vol, * @algorithm: one of virStorageVolWipeAlgorithm * @flags: future flags, use 0 for now * - * Similar to virStorageVolWipe, but one can choose - * between different wiping algorithms. + * Similar to virStorageVolWipe, but one can choose between + * different wiping algorithms. Also note, that depending on the + * actual volume representation, this call may not really + * overwrite the physical location of the volume. For instance, + * files stored journaled, log structured, copy-on-write, + * versioned, and network file systems are known to be + * problematic. * * Returns 0 on success, or -1 on error. */ -- 2.4.10

3 5

[libvirt] [PATCH 0/6] Small PCI hostdev improvements and fixes
by Andrea Bolognani 21 Dec '15

21 Dec '15

While working on v2 of my series[1] intended to fix https://bugzilla.redhat.com/show_bug.cgi?id=1272300 I'm running into some smaller stuff that isn't necessarily strictly related to the issue at hand. The idea is to have this smaller patches merged before the bigger, more invasive changes that series will bring, and hopefully make it easier to review by laying down the groundwork in advance. The series is organized as follows: 1-2: remove function parameters that are either not inspected by the function or never have an interesting value 3: change a string field that can only assume one of three values to a proper enumeration 4: improve bookkeeping when detaching PCI devices from the host, by making sure they always transition through the 'inactive' state (when they're attached to the stub driver) 5: make sure we never try to reattach an unmanaged device to the host, not even when recovering from a previous error 6: add some useful debug messages I expect patches 1, 2 and 6 to be fairly uncontroversial and easy to review; patch 3 is kinda big, but it's IMHO a nice improvement over the current situation; patches 4 and 5 are small but definitely need a long hard stare from a different pair of eyes to ensure they're not breaking anything. Cheers. [1] https://www.redhat.com/archives/libvir-list/2015-December/msg00070.html Andrea Bolognani (6): pci: Remove redundant parameter from virPCIDeviceBindToStub() pci: Remove 'reprobe' parameter from virPCIDeviceUnbind() pci: Introduce virPCIStubDriver enumeration hostdev: Mark PCI devices as inactive as they're detached hostdev: Only rollback detach of managed devices on error hostdev: Emit debug messages while handling PCI hostdevs src/libvirt_private.syms | 2 + src/libxl/libxl_driver.c | 3 +- src/qemu/qemu_driver.c | 6 +-- src/util/virhostdev.c | 107 ++++++++++++++++++++++++++++++----------------- src/util/virpci.c | 102 ++++++++++++++++++++++---------------------- src/util/virpci.h | 18 +++++--- src/xen/xen_driver.c | 3 +- tests/virhostdevtest.c | 5 ++- tests/virpcitest.c | 35 ++++++++++------ 9 files changed, 163 insertions(+), 118 deletions(-) -- 2.5.0

3 20