- Devel - libvirt List Archives

[libvirt] [PATCH] qemu: migration: Refresh device information after transferring state
by Peter Krempa 02 Feb '18

02 Feb '18

In my first approach in 4b480d10768c I overlooked the comment in qemuMigrationRunIncoming stating that during actual migration the qemuMigrationRunIncoming does not wait until the migration is complete but rather offloads that to the Finish phase of migration. This means that during actual migration qemuProcessRefreshState was called prior to qemu actually transferring the full state and thus the queries did not get the correct information. The approach worked only for restore, where we wait for the migration to finish during qemu startup. Fix the issue by calling qemuProcessRefreshState both from qemuProcessStart if there's no incomming migration and from qemuMigrationFinish so that the code actually works as expected. --- src/qemu/qemu_migration.c | 9 +++++++++ src/qemu/qemu_process.c | 17 ++++++++++------- src/qemu/qemu_process.h | 4 ++++ 3 files changed, 23 insertions(+), 7 deletions(-) diff --git a/src/qemu/qemu_migration.c b/src/qemu/qemu_migration.c index 1854900c9a..ea8b275601 100644 --- a/src/qemu/qemu_migration.c +++ b/src/qemu/qemu_migration.c @@ -5373,6 +5373,15 @@ qemuMigrationFinish(virQEMUDriverPtr driver, goto endjob; } + /* Now that the state data was transferred we can refresh the actual state + * of the devices */ + if (qemuProcessRefreshState(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN) < 0) { + /* Similarly to the case above v2 protocol will not be able to recover + * from this. Let's ignore this and perhaps stuff will not break. */ + if (v3proto) + goto endjob; + } + if (priv->job.current->status == QEMU_DOMAIN_JOB_STATUS_POSTCOPY) inPostCopy = true; diff --git a/src/qemu/qemu_process.c b/src/qemu/qemu_process.c index bcd4ac8ad6..fe12d94d37 100644 --- a/src/qemu/qemu_process.c +++ b/src/qemu/qemu_process.c @@ -6084,7 +6084,7 @@ qemuProcessLaunch(virConnectPtr conn, * function is called after a deferred migration finishes so that we can update * state influenced by the migration stream. */ -static int +int qemuProcessRefreshState(virQEMUDriverPtr driver, virDomainObjPtr vm, qemuDomainAsyncJob asyncJob) @@ -6125,9 +6125,6 @@ qemuProcessFinishStartup(virConnectPtr conn, virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver); int ret = -1; - if (qemuProcessRefreshState(driver, vm, asyncJob) < 0) - goto cleanup; - if (startCPUs) { VIR_DEBUG("Starting domain CPUs"); if (qemuProcessStartCPUs(driver, vm, conn, @@ -6231,11 +6228,17 @@ qemuProcessStart(virConnectPtr conn, VIR_DOMAIN_PAUSED_USER) < 0) goto stop; - /* Keep watching qemu log for errors during incoming migration, otherwise - * unset reporting errors from qemu log. */ - if (!incoming) + if (!incoming) { + /* Keep watching qemu log for errors during incoming migration, otherwise + * unset reporting errors from qemu log. */ qemuMonitorSetDomainLog(priv->mon, NULL, NULL, NULL); + /* Refresh state of devices from qemu. During migration this needs to + * happen after the state information is fully transferred. */ + if (qemuProcessRefreshState(driver, vm, asyncJob) < 0) + goto stop; + } + ret = 0; cleanup: diff --git a/src/qemu/qemu_process.h b/src/qemu/qemu_process.h index cd9a720313..b383ff309b 100644 --- a/src/qemu/qemu_process.h +++ b/src/qemu/qemu_process.h @@ -129,6 +129,10 @@ int qemuProcessFinishStartup(virConnectPtr conn, bool startCPUs, virDomainPausedReason pausedReason); +int qemuProcessRefreshState(virQEMUDriverPtr driver, + virDomainObjPtr vm, + qemuDomainAsyncJob asyncJob); + typedef enum { VIR_QEMU_PROCESS_STOP_MIGRATED = 1 << 0, VIR_QEMU_PROCESS_STOP_NO_RELABEL = 1 << 1, -- 2.15.0

2 1

[libvirt] [PATCH v3 00/11] Implement query-dump command
by John Ferlan 02 Feb '18

02 Feb '18

v2: https://www.redhat.com/archives/libvir-list/2018-January/msg00636.html Summary of changes since v2: * Generate a dump stats extraction helper in order to share with the DUMP_COMPLETED event and the query-dump command. Additionally add the error string from Qemu to be processed later * When processing DUMP_COMPLETED event implementation, use the returned stats buffer and copy any error message for the waiting completion to process properly. NB: Was able to test this by altering the guest memory size larger than the space to store the dump memory and got the following error: error: operation failed: memory-only dump failed: dump: failed to save memory * As suggested during review - alter the jobInfo @stats to be part of a union. Took the liberty to rename the fields as well. * The point raised in patch 5 regarding mirrorStats being collect for a non-migration (e.g. save and dump) was handled by creating a new type which will use the migStats only to collect data and avoid the mirrorStats. When converting to a JobInfo, only the migStats will then be used. * Patches 6 and 7 are new. One for the union and one to separate migrate and save/dump style migration jobs. * Former patch 6 (now 8) is altered to handle the union separation * Former patch 8 (now 10) is altered mainly to handle the data in buffers from the DUMP_COMPLETED event (either data or error). NB: Comment from former patch 8: "We should update job.completed at this point." I believe that's handled because qemuDomainObjResetAsyncJob will reset the dumpCompleted back to false. Unless there's a specific reason to do it there... NB: Formerly ACK'd patches 4 and 7 (now 9) did not change. John Ferlan (11): qemu: Add support for DUMP_COMPLETED event qemu: Introduce qemuProcessHandleDumpCompleted qemu: Introduce qemuMonitor[JSON]QueryDump qemu: Add new parameter to qemuMonitorDumpToFd qemu: Introduce qemuDomainGetJobInfoMigrationStats qemu: Convert jobInfo stats into a union qemu: Introduce QEMU_DOMAIN_JOB_STATS_TYPE_SAVEDUMP qemu: Introduce qemuDomainGetJobInfoDumpStats qemu: Add dump completed event to the capabilities qemu: Allow showing the dump progress for memory only dump docs: Add news article for query memory-only dump processing percentage docs/news.xml | 11 ++ src/qemu/qemu_capabilities.c | 2 + src/qemu/qemu_capabilities.h | 1 + src/qemu/qemu_domain.c | 130 +++++++++++++-- src/qemu/qemu_domain.h | 18 +- src/qemu/qemu_driver.c | 183 ++++++++++++++++++--- src/qemu/qemu_migration.c | 13 +- src/qemu/qemu_migration_cookie.c | 4 +- src/qemu/qemu_monitor.c | 45 ++++- src/qemu/qemu_monitor.h | 36 +++- src/qemu/qemu_monitor_json.c | 106 +++++++++++- src/qemu/qemu_monitor_json.h | 6 +- src/qemu/qemu_process.c | 34 +++- .../caps_2.10.0-gicv2.aarch64.xml | 1 + .../caps_2.10.0-gicv3.aarch64.xml | 1 + tests/qemucapabilitiesdata/caps_2.10.0.ppc64.xml | 1 + tests/qemucapabilitiesdata/caps_2.10.0.s390x.xml | 1 + tests/qemucapabilitiesdata/caps_2.10.0.x86_64.xml | 1 + .../caps_2.6.0-gicv2.aarch64.xml | 1 + .../caps_2.6.0-gicv3.aarch64.xml | 1 + tests/qemucapabilitiesdata/caps_2.6.0.ppc64.xml | 1 + tests/qemucapabilitiesdata/caps_2.6.0.x86_64.xml | 1 + tests/qemucapabilitiesdata/caps_2.7.0.s390x.xml | 1 + tests/qemucapabilitiesdata/caps_2.7.0.x86_64.xml | 1 + tests/qemucapabilitiesdata/caps_2.8.0.s390x.xml | 1 + tests/qemucapabilitiesdata/caps_2.8.0.x86_64.xml | 1 + tests/qemucapabilitiesdata/caps_2.9.0.ppc64.xml | 1 + tests/qemucapabilitiesdata/caps_2.9.0.s390x.xml | 1 + tests/qemucapabilitiesdata/caps_2.9.0.x86_64.xml | 1 + tests/qemumonitorjsontest.c | 3 +- 30 files changed, 553 insertions(+), 55 deletions(-) -- 2.13.6

2 17

[libvirt] [PATCH v3 0/5] Introduce VIR_ERR_DEVICE_MISSING
by Chen Hanxiao 01 Feb '18

01 Feb '18

Use Introduce VIR_ERR_DEVICE_MISSING for hotplug and detach device error message. Chen Hanxiao (5): qemu: Add some more details for hotplug errors when device not found qemu: Introduce VIR_ERR_DEVICE_MISSING qemu: Use VIR_ERR_DEVICE_MISSING for various hotplug messages qemu: Use VIR_ERR_DEVICE_MISSING for various DetachDeviceConfig messages news: Add VIR_ERR_DEVICE_MISSING change as improvements docs/news.xml | 5 +++++ include/libvirt/virterror.h | 1 + src/conf/domain_conf.c | 8 ++++---- src/libvirt_private.syms | 2 ++ src/qemu/qemu_driver.c | 20 +++++++++---------- src/qemu/qemu_hotplug.c | 47 ++++++++++++++++++++++++++++----------------- src/util/virerror.c | 6 ++++++ 7 files changed, 57 insertions(+), 32 deletions(-) -- 2.14.3

2 13

[libvirt] [PATCH 0/3] Misc RPC fixes
by Daniel P. Berrangé 01 Feb '18

01 Feb '18

Misc fixes pointed out by John after coverity had a fit :-) Daniel P. Berrangé (3): tests: validate private data / pre / post exec hooks for RPC APIs rpc: assume private data callbacks are always non-NULL rpc: fix non-NULL annotations when GNUTLS is disabled src/rpc/virnetserverclient.c | 47 +++++++------- src/rpc/virnetserverclient.h | 4 ++ .../virnetdaemondata/input-data-admin-nomdns.json | 12 ++-- .../input-data-admin-server-names.json | 12 ++-- .../virnetdaemondata/input-data-anon-clients.json | 6 +- .../input-data-client-auth-pending-failure.json | 3 +- .../input-data-client-auth-pending.json | 6 +- tests/virnetdaemondata/input-data-client-ids.json | 6 +- .../input-data-client-timestamp.json | 6 +- .../input-data-initial-nomdns.json | 6 +- tests/virnetdaemondata/input-data-initial.json | 6 +- .../input-data-no-keepalive-required.json | 12 ++-- .../virnetdaemondata/output-data-admin-nomdns.json | 12 ++-- .../output-data-admin-server-names.json | 12 ++-- .../virnetdaemondata/output-data-anon-clients.json | 6 +- .../output-data-client-auth-pending.json | 6 +- tests/virnetdaemondata/output-data-client-ids.json | 6 +- .../output-data-client-timestamp.json | 6 +- .../output-data-initial-nomdns.json | 6 +- tests/virnetdaemondata/output-data-initial.json | 6 +- .../output-data-no-keepalive-required.json | 12 ++-- tests/virnetdaemontest.c | 71 +++++++++++++++++++--- tests/virnetserverclienttest.c | 22 ++++++- 23 files changed, 208 insertions(+), 83 deletions(-) -- 2.14.3

2 6

[libvirt] [PATCH] qemu: Add USB and memory balloon by default for aarch64/virt guests
by Andrea Bolognani 01 Feb '18

01 Feb '18

Basically all existing guest types, regardless of the architectur, get both a USB controller and a virtio memory balloon by default. s390 guests are an exception, for the very good reason that they don't support USB at all; the other exception is aarch64/virt guests, but in the latter case isn't a compelling reason for them to deviate from the widely adopted convention, especially since: * x86/q35 guests, which aarch64/virt guests are for the most part identical to, add these devices by default; * it's trivial to opt out of both default devices by setting model='none'; * higher level applications such as Nova expect at least the USB controller to be present. So add it by default for newly-defined guests. Existing guests will, of course, be left unchanged. Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1538637 Signed-off-by: Andrea Bolognani <abologna(a)redhat.com> --- src/qemu/qemu_domain.c | 34 +++++++++++++++++++++++++++++----- 1 file changed, 29 insertions(+), 5 deletions(-) diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c index 6b4bd3cca..f1c3b3d1e 100644 --- a/src/qemu/qemu_domain.c +++ b/src/qemu/qemu_domain.c @@ -2630,7 +2630,8 @@ qemuDomainDefAddImplicitInputDevice(virDomainDef *def) static int qemuDomainDefAddDefaultDevices(virDomainDefPtr def, - virQEMUCapsPtr qemuCaps) + virQEMUCapsPtr qemuCaps, + unsigned int parseFlags) { bool addDefaultUSB = true; int usbModel = -1; /* "default for machinetype" */ @@ -2680,10 +2681,33 @@ qemuDomainDefAddDefaultDevices(virDomainDefPtr def, case VIR_ARCH_ARMV7L: case VIR_ARCH_AARCH64: - addDefaultUSB = false; - addDefaultMemballoon = false; - if (qemuDomainIsVirt(def)) + if (qemuDomainIsVirt(def)) { + /* All mach-virt guests get a PCIe Root, if supported by + * the QEMU binary */ addPCIeRoot = virQEMUCapsGet(qemuCaps, QEMU_CAPS_OBJECT_GPEX); + } + + if (qemuDomainIsVirt(def) && + parseFlags & VIR_DOMAIN_DEF_PARSE_ABI_UPDATE) { + /* In addition to PCIe Root, newly-defined mach-virt guests + * also get a couple more devices so that they're more similar + * to guests on other architectures, notably x86/q35: + * + * 1) a USB3 controller, if supported by the QEMU binary; + * 2) a virtio memory balloon, as per the defaults defined + * above. + */ + if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE_QEMU_XHCI)) + usbModel = VIR_DOMAIN_CONTROLLER_MODEL_USB_QEMU_XHCI; + else + addDefaultUSB = false; + } else { + /* Other ARM guests (and existing mach-virt guests, in order + * to preserve guest ABI compatibility) don't get a PCIe Root, + * a USB controller or a memory balloon */ + addDefaultUSB = false; + addDefaultMemballoon = false; + } break; case VIR_ARCH_PPC64: @@ -3187,7 +3211,7 @@ qemuDomainDefPostParse(virDomainDefPtr def, goto cleanup; } - if (qemuDomainDefAddDefaultDevices(def, qemuCaps) < 0) + if (qemuDomainDefAddDefaultDevices(def, qemuCaps, parseFlags) < 0) goto cleanup; if (qemuCanonicalizeMachine(def, qemuCaps) < 0) -- 2.14.3

4 11

[libvirt] [PATCH v6 00/14] Move qemu command line controller checks to qemuDomainDeviceDefValidateController* checks
by John Ferlan 01 Feb '18

01 Feb '18

v5: https://www.redhat.com/archives/libvir-list/2018-January/msg00188.html Although each was ACK'd in v5, that was before alter the post parse callbacks to ensure SCSI controller model's were filled in. I post again mainly because some things do change slightly and well for sanity and being "correct". Unless there's an objection, I'll push some time on Friday. Andrea Bolognani (1): qemu: Add missing checks for pcie-root-port options John Ferlan (12): qemu: Introduce qemuDomainDeviceDefValidateControllerAttributes qemu: Move and rename qemuBuildCheckSCSIControllerModel qemu: Adjust SCSI controller switch in qemuBuildControllerDevStr qemu: Add check for iothread attribute in validate controller qemu: Introduce qemuDomainDeviceDefValidateControllerSCSI qemu: Introduce qemuDomainDeviceDefValidateControllerPCI qemu: Use virDomainPCIControllerOpts in qemuBuildControllerDevStr qemu: Move PCI command modelName check to controller def validate qemu: Move PCI command modelName TypeToString to controller def validate qemu: Move PCI more command checks to controller def validate qemu: Complete PCI command checks to controller def validate qemu: Introduce qemuDomainDeviceDefValidateControllerSATA Marc Hartmayer (1): qemu: Use switch statement for address types in qemuBuildControllerDevStr src/qemu/qemu_command.c | 505 +++++------------------------------------------- src/qemu/qemu_domain.c | 478 ++++++++++++++++++++++++++++++++++++++++++++- tests/qemumemlocktest.c | 14 ++ tests/qemuxml2xmltest.c | 5 +- 4 files changed, 545 insertions(+), 457 deletions(-) -- 2.13.6

2 16

[libvirt] [PATCH v2 00/11] Add admin protocol support for virtlogd/virtlockd
by Daniel P. Berrange 01 Feb '18

01 Feb '18

The initial admin protocol support was only integrated into libvirtd. This series extracts that code so that it is reusable with all the daemons we have (and more than we'll get). In v2: - Fixed completely broken post-exec restart support for admin servers - Misc fixes from v1 review Daniel P. Berrange (11): admin: move admins server impl/dispatch into src/admin directory libvirtd: rename virNetServerClient callback impls to match type names admin: add support for post-exec restart callbacks util: add virGetUNIXSocketPath helper rpc: clarify "void *" values passed to client callbacks rpc: pass virNetServer to post-exec restart callback in typesafe manner rpc: annotate various parameters as being required to be non-NULL rpc: add method for checking if a named server exists rpc: refactor virNetServer setup for post-exec restarts logd: add support for admin protocol in virtlogd lockd: add support for admin protocol in virtlockd .gitignore | 1 + cfg.mk | 10 +- daemon/Makefile.am | 33 +--- daemon/libvirtd.c | 10 +- daemon/libvirtd.h | 10 -- daemon/remote.c | 6 +- daemon/remote.h | 6 +- po/POTFILES.in | 6 +- src/Makefile.am | 33 +++- {daemon => src/admin}/admin_server.c | 4 +- {daemon => src/admin}/admin_server.h | 6 +- .../admin.c => src/admin/admin_server_dispatch.c | 48 ++++-- .../admin.h => src/admin/admin_server_dispatch.h | 18 +- src/libvirt-admin.c | 23 ++- src/libvirt_private.syms | 1 + src/libvirt_remote.syms | 2 +- src/locking/lock_daemon.c | 186 +++++++++++++++----- src/locking/lock_daemon_config.c | 3 + src/locking/lock_daemon_config.h | 1 + src/locking/test_virtlockd.aug.in | 4 + src/locking/virtlockd-admin.socket.in | 10 ++ src/locking/virtlockd.aug | 1 + src/locking/virtlockd.conf | 6 + src/locking/virtlockd.service.in | 1 + src/logging/log_daemon.c | 188 ++++++++++++++++----- src/logging/log_daemon_config.c | 3 + src/logging/log_daemon_config.h | 1 + src/logging/test_virtlogd.aug.in | 4 + src/logging/virtlogd-admin.socket.in | 10 ++ src/logging/virtlogd.aug | 1 + src/logging/virtlogd.service.in | 1 + src/rpc/virnetdaemon.c | 164 ++++++++++-------- src/rpc/virnetdaemon.h | 20 ++- src/rpc/virnetserver.c | 6 +- src/rpc/virnetserver.h | 7 +- src/rpc/virnetserverclient.c | 14 +- src/rpc/virnetserverclient.h | 23 ++- src/rpc/virnetserverprogram.h | 3 - src/util/virutil.c | 45 +++++ src/util/virutil.h | 1 + tests/virnetdaemontest.c | 37 +++- 41 files changed, 670 insertions(+), 287 deletions(-) rename {daemon => src/admin}/admin_server.c (99%) rename {daemon => src/admin}/admin_server.h (96%) rename daemon/admin.c => src/admin/admin_server_dispatch.c (92%) rename daemon/admin.h => src/admin/admin_server_dispatch.h (59%) create mode 100644 src/locking/virtlockd-admin.socket.in create mode 100644 src/logging/virtlogd-admin.socket.in -- 2.14.3

5 22

[libvirt] [PATCH] qemu: Expose rx/tx_queue_size in qemu.conf too
by Michal Privoznik 01 Feb '18

01 Feb '18

In 2074ef6cd4a2 and c56cdf259 (and friends) we've added two attributes to virtio NICs: rx_queue_size and tx_queue_size. However, sysadmins might want to set these on per-host basis but don't necessarily have an access to domain XML (e.g. because they are generated by some other app). So let's expose them under qemu.conf (the settings from domain XML still take precedence as they are more specific ones). Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com> --- docs/formatdomain.html.in | 12 +++++++++-- src/qemu/libvirtd_qemu.aug | 4 ++++ src/qemu/qemu.conf | 7 +++++++ src/qemu/qemu_command.c | 42 ++++++++++++++++++++++++++++++-------- src/qemu/qemu_command.h | 3 ++- src/qemu/qemu_conf.c | 4 ++++ src/qemu/qemu_conf.h | 3 +++ src/qemu/qemu_hotplug.c | 2 +- src/qemu/test_libvirtd_qemu.aug.in | 2 ++ 9 files changed, 66 insertions(+), 13 deletions(-) diff --git a/docs/formatdomain.html.in b/docs/formatdomain.html.in index d272cc1ba..c0107ab4b 100644 --- a/docs/formatdomain.html.in +++ b/docs/formatdomain.html.in @@ -5373,7 +5373,11 @@ qemu-kvm -net nic,model=? /dev/null some restrictions on actual value. For instance, latest QEMU (as of 2016-09-01) requires value to be a power of two from [256, 1024] range. - Since 2.3.0 (QEMU and KVM only) + Since 2.3.0 (QEMU and KVM only) + Then, Since 4.1.0 the default value can be + set in <code>qemu.conf</code> file and thus overrides hypervisor + default. + In general you should leave this option alone, unless you are very certain you know what you are doing. @@ -5389,7 +5393,11 @@ qemu-kvm -net nic,model=? /dev/null range. In addition to that, this may work only for a subset of interface types, e.g. aforementioned QEMU enables this option only for <code>vhostuser</code> type. - Since 3.7.0 (QEMU and KVM only) + Since 3.7.0 (QEMU and KVM only) + Then, Since 4.1.0 the default value can be + set in <code>qemu.conf</code> file and thus overrides hypervisor + default. + In general you should leave this option alone, unless you are very certain you know what you are doing. diff --git a/src/qemu/libvirtd_qemu.aug b/src/qemu/libvirtd_qemu.aug index c19bf3a43..084290296 100644 --- a/src/qemu/libvirtd_qemu.aug +++ b/src/qemu/libvirtd_qemu.aug @@ -118,6 +118,9 @@ module Libvirtd_qemu = let vxhs_entry = bool_entry "vxhs_tls" | str_entry "vxhs_tls_x509_cert_dir" + let virtio_entry = int_entry "rx_queue_size" + | int_entry "tx_queue_size" + (* Each entry in the config is one of the following ... *) let entry = default_tls_entry | vnc_entry @@ -137,6 +140,7 @@ module Libvirtd_qemu = | gluster_debug_level_entry | memory_entry | vxhs_entry + | virtio_entry let comment = [ label "#comment" . del /#[ \t]*/ "# " . store /([^ \t\n][^\n]*)?/ . del /\n/ "\n" ] let empty = [ label "#empty" . eol ] diff --git a/src/qemu/qemu.conf b/src/qemu/qemu.conf index 43dd561cc..a945ebdd5 100644 --- a/src/qemu/qemu.conf +++ b/src/qemu/qemu.conf @@ -775,3 +775,10 @@ # This directory is used for memoryBacking source if configured as file. # NOTE: big files will be stored here #memory_backing_dir = "/var/lib/libvirt/qemu/ram" + +# The following two values set the default RX/TX ring buffer size for virtio +# interfaces. These values are taken unless overridden in domain XML. Please +# note that QEMU accepts 256, 512 and 1024 only. These values correspond to +# those from domain XML. +#rx_queue_size = 1024 +#tx_queue_size = 1024 diff --git a/src/qemu/qemu_command.c b/src/qemu/qemu_command.c index b8aede32d..771c12445 100644 --- a/src/qemu/qemu_command.c +++ b/src/qemu/qemu_command.c @@ -3693,7 +3693,8 @@ qemuBuildNicStr(virDomainNetDefPtr net, char * -qemuBuildNicDevStr(virDomainDefPtr def, +qemuBuildNicDevStr(virQEMUDriverConfigPtr cfg, + virDomainDefPtr def, virDomainNetDefPtr net, int vlan, unsigned int bootindex, @@ -3813,21 +3814,40 @@ qemuBuildNicDevStr(virDomainDefPtr def, virBufferAsprintf(&buf, ",mq=on,vectors=%zu", 2 * vhostfdSize + 2); } } - if (usingVirtio && net->driver.virtio.rx_queue_size) { - if (!virQEMUCapsGet(qemuCaps, QEMU_CAPS_VIRTIO_NET_RX_QUEUE_SIZE)) { + if (usingVirtio) { + unsigned int rx_queue_size = net->driver.virtio.rx_queue_size; + + if (rx_queue_size == 0 && + virQEMUCapsGet(qemuCaps, QEMU_CAPS_VIRTIO_NET_RX_QUEUE_SIZE)) + rx_queue_size = cfg->rx_queue_size; + + + if (rx_queue_size && + virQEMUCapsGet(qemuCaps, QEMU_CAPS_VIRTIO_NET_RX_QUEUE_SIZE)) { + net->driver.virtio.rx_queue_size = rx_queue_size; + virBufferAsprintf(&buf, ",rx_queue_size=%u", rx_queue_size); + } else if (rx_queue_size) { virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s", _("virtio rx_queue_size option is not supported with this QEMU binary")); goto error; } - virBufferAsprintf(&buf, ",rx_queue_size=%u", net->driver.virtio.rx_queue_size); } - if (usingVirtio && net->driver.virtio.tx_queue_size) { - if (!virQEMUCapsGet(qemuCaps, QEMU_CAPS_VIRTIO_NET_TX_QUEUE_SIZE)) { + if (usingVirtio) { + unsigned int tx_queue_size = net->driver.virtio.tx_queue_size; + + if (tx_queue_size == 0 && + virQEMUCapsGet(qemuCaps, QEMU_CAPS_VIRTIO_NET_TX_QUEUE_SIZE)) + tx_queue_size = cfg->tx_queue_size; + + if (tx_queue_size && + virQEMUCapsGet(qemuCaps, QEMU_CAPS_VIRTIO_NET_TX_QUEUE_SIZE)) { + net->driver.virtio.tx_queue_size = tx_queue_size; + virBufferAsprintf(&buf, ",tx_queue_size=%u", tx_queue_size); + } else if (tx_queue_size) { virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s", _("virtio tx_queue_size option is not supported with this QEMU binary")); goto error; } - virBufferAsprintf(&buf, ",tx_queue_size=%u", net->driver.virtio.tx_queue_size); } if (usingVirtio && net->mtu) { @@ -8489,7 +8509,7 @@ qemuBuildVhostuserCommandLine(virQEMUDriverPtr driver, virCommandAddArg(cmd, netdev); VIR_FREE(netdev); - if (!(nic = qemuBuildNicDevStr(def, net, -1, bootindex, + if (!(nic = qemuBuildNicDevStr(cfg, def, net, -1, bootindex, queues, qemuCaps))) { virReportError(VIR_ERR_INTERNAL_ERROR, "%s", _("Error generating NIC -device string")); @@ -8526,6 +8546,7 @@ qemuBuildInterfaceCommandLine(virQEMUDriverPtr driver, int **nicindexes, bool chardevStdioLogd) { + virQEMUDriverConfigPtr cfg; int ret = -1; char *nic = NULL, *host = NULL; int *tapfd = NULL; @@ -8587,6 +8608,8 @@ qemuBuildInterfaceCommandLine(virQEMUDriverPtr driver, return -1; } + cfg = virQEMUDriverGetConfig(driver); + switch (actualType) { case VIR_DOMAIN_NET_TYPE_NETWORK: case VIR_DOMAIN_NET_TYPE_BRIDGE: @@ -8782,7 +8805,7 @@ qemuBuildInterfaceCommandLine(virQEMUDriverPtr driver, virCommandAddArgList(cmd, "-netdev", host, NULL); } if (qemuDomainSupportsNicdev(def, net)) { - if (!(nic = qemuBuildNicDevStr(def, net, vlan, bootindex, + if (!(nic = qemuBuildNicDevStr(cfg, def, net, vlan, bootindex, vhostfdSize, qemuCaps))) goto cleanup; virCommandAddArgList(cmd, "-device", nic, NULL); @@ -8826,6 +8849,7 @@ qemuBuildInterfaceCommandLine(virQEMUDriverPtr driver, VIR_FREE(host); VIR_FREE(tapfdName); VIR_FREE(vhostfdName); + virObjectUnref(cfg); return ret; } diff --git a/src/qemu/qemu_command.h b/src/qemu/qemu_command.h index bdde6f918..85f7bd1b4 100644 --- a/src/qemu/qemu_command.h +++ b/src/qemu/qemu_command.h @@ -90,7 +90,8 @@ char *qemuBuildNicStr(virDomainNetDefPtr net, int vlan); /* Current, best practice */ -char *qemuBuildNicDevStr(virDomainDefPtr def, +char *qemuBuildNicDevStr(virQEMUDriverConfigPtr cfg, + virDomainDefPtr def, virDomainNetDefPtr net, int vlan, unsigned int bootindex, diff --git a/src/qemu/qemu_conf.c b/src/qemu/qemu_conf.c index af503d31c..2fa96431f 100644 --- a/src/qemu/qemu_conf.c +++ b/src/qemu/qemu_conf.c @@ -912,6 +912,10 @@ int virQEMUDriverConfigLoadFile(virQEMUDriverConfigPtr cfg, if (virConfGetValueString(conf, "memory_backing_dir", &cfg->memoryBackingDir) < 0) goto cleanup; + if (virConfGetValueUInt(conf, "rx_queue_size", &cfg->rx_queue_size) < 0 || + virConfGetValueUInt(conf, "tx_queue_size", &cfg->tx_queue_size) < 0) + goto cleanup; + ret = 0; cleanup: diff --git a/src/qemu/qemu_conf.h b/src/qemu/qemu_conf.h index a553e30e2..3f38a76c2 100644 --- a/src/qemu/qemu_conf.h +++ b/src/qemu/qemu_conf.h @@ -206,6 +206,9 @@ struct _virQEMUDriverConfig { bool vxhsTLS; char *vxhsTLSx509certdir; + + unsigned int rx_queue_size; + unsigned int tx_queue_size; }; /* Main driver state */ diff --git a/src/qemu/qemu_hotplug.c b/src/qemu/qemu_hotplug.c index 6b245bd6a..18b460704 100644 --- a/src/qemu/qemu_hotplug.c +++ b/src/qemu/qemu_hotplug.c @@ -1116,7 +1116,7 @@ qemuDomainAttachNetDevice(virQEMUDriverPtr driver, for (i = 0; i < vhostfdSize; i++) VIR_FORCE_CLOSE(vhostfd[i]); - if (!(nicstr = qemuBuildNicDevStr(vm->def, net, vlan, 0, + if (!(nicstr = qemuBuildNicDevStr(cfg, vm->def, net, vlan, 0, queueSize, priv->qemuCaps))) goto try_remove; diff --git a/src/qemu/test_libvirtd_qemu.aug.in b/src/qemu/test_libvirtd_qemu.aug.in index 688e5b9fd..4fc4e2f4e 100644 --- a/src/qemu/test_libvirtd_qemu.aug.in +++ b/src/qemu/test_libvirtd_qemu.aug.in @@ -100,3 +100,5 @@ module Test_libvirtd_qemu = { "1" = "mount" } } { "memory_backing_dir" = "/var/lib/libvirt/qemu/ram" } +{ "rx_queue_size" = "1024" } +{ "tx_queue_size" = "1024" } -- 2.13.6

2 2

[libvirt] [PATCH v2 0/6] Last CAT fixups v2
by Martin Kletzander 01 Feb '18

01 Feb '18

Few finishing touches for the Cache Allocation Technology. This series copies the default group's mask to newly created groups, fixes a possible wrong memory access, and cleans up after domains even if daemon was restarted. Martin Kletzander (6): util: Add helpers for getting resctrl group allocs util: Use default group's mask for unspecified resctrl allocations util: Don't overwrite mask in virResctrlAllocFindUnused qemu: Restore machinename even without cgroups util: Extract path formatting into virResctrlAllocDeterminePath qemu: Restore resctrl alloc data after restart src/libvirt_private.syms | 1 + src/qemu/qemu_cgroup.c | 4 - src/qemu/qemu_process.c | 11 +++ src/util/virresctrl.c | 219 +++++++++++++++++++++++++++++------------------ src/util/virresctrl.h | 4 + 5 files changed, 152 insertions(+), 87 deletions(-) -- 2.15.1

4 15

[libvirt] [Question] nodedev-list --cap return nothing after creating a mdev
by Wuzongyong (Euler Dept) 01 Feb '18

01 Feb '18

Hi, I meet a problem that libvirt handle new mediated devices wrongly. Command `virsh nodedev-list -cap mdev` return none after I create a mdev, and the root cause I found is: // kernel code int mdev_device_create(struct kobject *kobj, struct device *dev, uuid_le uuid) { ... ret = device_register(&mdev->dev); // send udev event here --- 1 if (ret) { put_device(&mdev->dev); goto create_err; } ret = mdev_device_create_ops(kobj, mdev); if (ret) goto create_failed; ret = mdev_create_sysfs_files(&mdev->dev, type); // create mdev_type symbol link here --- 2 if (ret) { mdev_device_remove_ops(mdev, true); goto create_failed; } mdev->type_kobj = kobj; dev_dbg(&mdev->dev, "MDEV: created\n"); ... } At point 1, kernel send a udev event(add). If libvirt receive this event and handle it before kernel code reach point 2, libvirt just fail to resolve link such as '/sys/devices/pci0000:80/0000:80:02.0/0000:81:00.0/36b79575-ada8-4406-893d-8cfd8d10a984/mdev_type' because the link haven't been created yet. Then libvirt thinks this devices doesn't exsit and stop tracking info of this device. Finally, command `vish nodedev-list -cap` return nothing though the device exist at host indeed. To make it easy to reproduce this problem, I just add a `udelay(1)` before point 2 by kprobe. So I think maybe we should add 'kobject_uevent(..., KOBJ_ADD)' after point 2 instead of point 1 ? Or is there any other method to solve this problem? Thanks, Zongyong Wu

2 7