April 2016 - Devel - libvirt List Archives

[libvirt] [PATCH 00/11] vz: change vz driver to be stateful driver and other enhancements
by Maxim Nestratov 01 Apr '16

01 Apr '16

There is no benefit in providing two ways of connecting to vz driver: by connecting via daemon and directly from client. Both ways finally come to a host where vz daemon sits. Always connecting via daemon allows us to have a single list of domains and share it among all connections. Maxim Nestratov (11): virsh: report when vz driver is compiled vz: change the order of capabilities reported vz: remove drivername field from vzConn structure vz: build driver as module and don't register it on client's side vz: pass vzConnPtr to prlsdkXxx functions instead of virConnectPtr vz: make vzConn structure be a new lockable object and use it vz: implement connectGetSysinfo hypervisor callback vz: remove close callback implementations vz: remove vzDriverLock/Unlock function vz: minor cleanup vz: change vzConnectIsAlive behavior daemon/Makefile.am | 4 + daemon/libvirtd.c | 9 + src/Makefile.am | 21 ++- src/libvirt.c | 7 - src/vz/vz_driver.c | 479 ++++++++++++++++++++++++++++++----------------------- src/vz/vz_sdk.c | 36 ++-- src/vz/vz_sdk.h | 6 +- src/vz/vz_utils.c | 18 +- src/vz/vz_utils.h | 15 +- tools/virsh.c | 3 + 10 files changed, 341 insertions(+), 257 deletions(-) -- 2.4.3

2 16

[libvirt] [PATCH perl] Add APIS & constants for new perf event support
by Daniel P. Berrange 01 Apr '16

01 Apr '16

Add binding for virDomainGetPerfEvents, virDomainSetPerfEvents, VIR_DOMAIN_STATS_PERF, VIR_PERF_PARAM_CMT, and VIR_FROM_PERF Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com> --- Changes | 1 + Virt.xs | 43 +++++++++++++++++++++++++++++++++++++++++++ lib/Sys/Virt/Domain.pm | 34 ++++++++++++++++++++++++++++++++++ lib/Sys/Virt/Error.pm | 4 ++++ 4 files changed, 82 insertions(+) diff --git a/Changes b/Changes index 2f86a3d..8e08e3e 100644 --- a/Changes +++ b/Changes @@ -12,6 +12,7 @@ Revision history for perl module Sys::Virt - Add VIR_MIGRATE_PARAM_DISKS_PORT constant - Add virDomainMigrateStartPostCopy API support - Add constants related to post-copy migration + - Add APIS & constants for new perf event support 1.3.2 2016-03-01 diff --git a/Virt.xs b/Virt.xs index 3664521..50ed6cd 100644 --- a/Virt.xs +++ b/Virt.xs @@ -4213,6 +4213,46 @@ set_blkio_parameters(dom, newparams, flags=0) Safefree(params); +HV * +get_perf_events(dom, flags=0) + virDomainPtr dom; + unsigned int flags; + PREINIT: + virTypedParameter *params = NULL; + int nparams = 0; + CODE: + if (virDomainGetPerfEvents(dom, &params, &nparams, flags) < 0) { + Safefree(params); + _croak_error(); + } + + RETVAL = vir_typed_param_to_hv(params, nparams); + Safefree(params); + OUTPUT: + RETVAL + + +void +set_perf_events(dom, newparams, flags=0) + virDomainPtr dom; + HV *newparams; + unsigned int flags; + PREINIT: + virTypedParameter *params = NULL; + int nparams = 0; + PPCODE: + if (virDomainGetPerfEvents(dom, &params, &nparams, flags) < 0) { + Safefree(params); + _croak_error(); + } + + nparams = vir_typed_param_from_hv(newparams, params, nparams); + + if (virDomainSetPerfEvents(dom, params, nparams, flags) < 0) + _croak_error(); + Safefree(params); + + unsigned long get_max_memory(dom) virDomainPtr dom; @@ -7641,6 +7681,7 @@ BOOT: REGISTER_CONSTANT(VIR_DOMAIN_STATS_INTERFACE, STATS_INTERFACE); REGISTER_CONSTANT(VIR_DOMAIN_STATS_STATE, STATS_STATE); REGISTER_CONSTANT(VIR_DOMAIN_STATS_VCPU, STATS_VCPU); + REGISTER_CONSTANT(VIR_DOMAIN_STATS_PERF, STATS_PERF); REGISTER_CONSTANT(VIR_CONNECT_GET_ALL_DOMAINS_STATS_ACTIVE, GET_ALL_STATS_ACTIVE); REGISTER_CONSTANT(VIR_CONNECT_GET_ALL_DOMAINS_STATS_INACTIVE, GET_ALL_STATS_INACTIVE); @@ -7905,6 +7946,7 @@ BOOT: REGISTER_CONSTANT(VIR_DOMAIN_NUMATUNE_MEM_PREFERRED, NUMATUNE_MEM_PREFERRED); REGISTER_CONSTANT(VIR_DOMAIN_NUMATUNE_MEM_INTERLEAVE, NUMATUNE_MEM_INTERLEAVE); + REGISTER_CONSTANT_STR(VIR_PERF_PARAM_CMT, PERF_PARAM_CMT); REGISTER_CONSTANT_STR(VIR_DOMAIN_BANDWIDTH_IN_AVERAGE, BANDWIDTH_IN_AVERAGE); REGISTER_CONSTANT_STR(VIR_DOMAIN_BANDWIDTH_IN_PEAK, BANDWIDTH_IN_PEAK); @@ -8368,6 +8410,7 @@ BOOT: REGISTER_CONSTANT(VIR_FROM_ADMIN, FROM_ADMIN); REGISTER_CONSTANT(VIR_FROM_LOGGING, FROM_LOGGING); REGISTER_CONSTANT(VIR_FROM_XENXL, FROM_XENXL); + REGISTER_CONSTANT(VIR_FROM_PERF, FROM_PERF); REGISTER_CONSTANT(VIR_ERR_OK, ERR_OK); diff --git a/lib/Sys/Virt/Domain.pm b/lib/Sys/Virt/Domain.pm index 26ead9b..b095dbd 100644 --- a/lib/Sys/Virt/Domain.pm +++ b/lib/Sys/Virt/Domain.pm @@ -812,6 +812,23 @@ of the NUMA PARAMETERS constants. The C<$flags> parameter accepts one or more the CONFIG OPTION constants documented later, and defaults to 0 if omitted. +=item my $params = $dom->get_perf_events($flags=0) + +Return a hash reference containing the set of performance +events that are available for the guest. The keys in the +hash are one of the constants PERF EVENTS described later. +The C<$flags> parameter accepts one or more the CONFIG +OPTION constants documented later, and defaults to 0 if +omitted. + +=item $dom->set_perf_events($params, $flags=0) + +Update the enabled state for performance events for the +guest. The C<$params> should be a hash reference whose +keys are one of the PERF EVENTS constants. The C<$flags> +parameter accepts one or more the CONFIG OPTION constants +documented later, and defaults to 0 if omitted. + =item $dom->block_resize($disk, $newsize, $flags=0) Resize the disk C<$disk> to have new size C<$newsize> KB. If the disk @@ -2588,6 +2605,19 @@ The burstable outbound bandwidth =back +=head2 PERF EVENTS + +The following constants defined performance events +which can be monitored for a guest + +=over 4 + +=item Sys::Virt::Domain::PERF_PARAM_CMT + +The CMT event counter + +=back + =head2 VCPU FLAGS The following constants are useful when getting/setting the @@ -3716,6 +3746,10 @@ General lifecycle state Virtual CPU info +=item Sys::Virt::Domain::STATS_PERF + +Performance event counter values + =back =head2 PROCESS SIGNALS diff --git a/lib/Sys/Virt/Error.pm b/lib/Sys/Virt/Error.pm index 3f6b058..b54f876 100644 --- a/lib/Sys/Virt/Error.pm +++ b/lib/Sys/Virt/Error.pm @@ -394,6 +394,10 @@ The logging service The Xen XL driver +=item Sys::Virt::Error::FROM_PERF + +The performance events subsystem + =back =head2 ERROR CODE CONSTANTS -- 2.5.5

1 0

[libvirt] [PATCH] docs: Remove unused CSS rule
by Andrea Bolognani 01 Apr '16

01 Apr '16

Commit e14c5069c512 dropped the only user of the #sponsor CSS rule, so we can drop it from the stylesheet as well. --- docs/libvirt.css | 6 ------ 1 file changed, 6 deletions(-) diff --git a/docs/libvirt.css b/docs/libvirt.css index c6f6d0d..2310b20 100644 --- a/docs/libvirt.css +++ b/docs/libvirt.css @@ -300,12 +300,6 @@ p.image { color: #ccc; } -#sponsor { - color: #757575; - text-decoration: inherit; - font-size: 1.2em; -} - span.since { color: #3c857c; font-style: italic; -- 2.5.5

2 2

[libvirt] [PATCH] Pass the correct cpu count when calling virDomainGetCPUStats.
by Nitesh Konkar 01 Apr '16

01 Apr '16

When using the --start option, the show_count should not be set to max_id as the --start <cpu> means we dont need those many inital cpu stats. Hence, show_count should be adjusted accordingly. Signed-off-by: Nitesh Konkar <nitkon12(a)linux.vnet.ibm.com> --- Before Patch: virsh cpu-stats --domain 24 --start 158 CPU158: cpu_time 0.073570788 seconds vcpu_time 0.033277032 seconds CPU159: cpu_time 0.000000000 seconds vcpu_time 0.000000000 seconds error: Failed to retrieve CPU statistics for domain 24 error: invalid argument: start_cpu 280 larger than maximum of 159 After Patch: virsh cpu-stats --domain 24 --start 158 CPU158: cpu_time 0.073570788 seconds vcpu_time 0.033277032 seconds CPU159: cpu_time 0.000000000 seconds vcpu_time 0.000000000 seconds tools/virsh-domain.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/virsh-domain.c b/tools/virsh-domain.c index bb1ee1c..7599bdc 100644 --- a/tools/virsh-domain.c +++ b/tools/virsh-domain.c @@ -7331,7 +7331,7 @@ cmdCPUStats(vshControl *ctl, const vshCmd *cmd) if (show_count < 0 || show_count > max_id) { if (show_count > max_id) vshPrint(ctl, _("Only %d CPUs available to show\n"), max_id); - show_count = max_id; + show_count = max_id - cpu; } /* get percpu information */ -- 1.8.3.1

2 1

[libvirt] [PATCH v2] migration: convert speed from Mb/sec to bytes/sec in drive-mirror jobs
by Rudy Zhang 01 Apr '16

01 Apr '16

Commit id '7b7600b3' added qemuMigrationDriveMirror to handle NBD mirroring, but passed the migrate_speed listed in MiB/s to be used for the mirror_speed which expects a bytes/s value. This patch will convert the migrate_speed value to its mirror_speed equivalent. Signed-off-by: Rudy Zhang <rudyflyzhang(a)gmail.com> --- src/qemu/qemu_migration.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) diff --git a/src/qemu/qemu_migration.c b/src/qemu/qemu_migration.c index 8bc76bf..3d9a55f 100644 --- a/src/qemu/qemu_migration.c +++ b/src/qemu/qemu_migration.c @@ -2093,12 +2093,21 @@ qemuMigrationDriveMirror(virQEMUDriverPtr driver, char *diskAlias = NULL; char *nbd_dest = NULL; char *hoststr = NULL; + unsigned long long mirror_speed = speed; unsigned int mirror_flags = VIR_DOMAIN_BLOCK_REBASE_REUSE_EXT; int rv; virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver); VIR_DEBUG("Starting drive mirrors for domain %s", vm->def->name); + if (mirror_speed > LLONG_MAX >> 20) { + virReportError(VIR_ERR_OVERFLOW, + _("bandwidth must be less than %llu"), + LLONG_MAX >> 20); + return ret; + } + mirror_speed <<= 20; + /* steal NBD port and thus prevent its propagation back to destination */ port = mig->nbd->port; mig->nbd->port = 0; @@ -2136,7 +2145,7 @@ qemuMigrationDriveMirror(virQEMUDriverPtr driver, qemuBlockJobSyncBegin(disk); /* Force "raw" format for NBD export */ mon_ret = qemuMonitorDriveMirror(priv->mon, diskAlias, nbd_dest, - "raw", speed, 0, 0, mirror_flags); + "raw", mirror_speed, 0, 0, mirror_flags); VIR_FREE(diskAlias); VIR_FREE(nbd_dest); -- 2.6.4

2 1

[libvirt] [PATCH 00/38] Introduce APIs to change daemon's logging settings
by Erik Skultety 01 Apr '16

01 Apr '16

This series implements administration APIs to modify daemon's logging settings which include priority, filters, and outputs. It also performs all necessary changes to enable this feature. Patches also involve some slight refactors to the existing code as well as wiring up the APIs to virt-admin client. The major change in internal code is how definition of filters and outputs is achieved. Until now each filter/output has been defined and appended to the global set of filters/outputs individually whilst holding a lock that was repeatedly released after each filter/output. If modification to filters and outputs should be exported through public APIs, all these setter operations must be atomic, i.e. accomplished as a unit, not gradually. For filters and priority, these are replaced normally, i.e. reset the original and replace it with the new copy while having the lock. For outputs, there was a slight issue, since resetting the global set of outputs would result in closing all the file descriptors, which might not be exactly what we want, since the new set of outputs that is meant to replace the original set might include outputs that are also contained within the original set. The logic for such outputs is following: Each file-based output that already exists in the original set and should continue existing in the new set will copy the existing FD from the global set. For each syslog-based output, closelog and openlog must be called anyway (because syslog keeps its FD private), but will be called at the very last moment if such an output already exists to prevent from changing before the lock for unit replaced is acquired, and thus breaking the atomicity that we want. For journald-based outputs, the FD is global, we just need to prevent it from being closed if such an output shall continue to exist. All outputs that need to continue to exist are prevented from closing when being destroyed-deallocated. On the other hand, all outputs that do not exist in the new set of outputs shall be closed correctly. Erik Skultety (38): virlog: Return void instead of int in virLogReset<Foo> methods virlog: Convert virLogOutputs to a list of pointers to outputs virlog: Convert virLogFilters to a list of pointers to filters virlog: Export virLogOutputPtr through header virlog: Export virLogFilterPtr through header virlog: Introduce virLogSetFilters virlog: Introduce virLogSetOutputs daemon: Replace virLogParseOutputs with virLogSetOutputs in callers daemon: Replace virLogParseFilters with virLogSetFilters in callers virlog: Introduce virLogDefineOutputs virlog: Introduce virLogDefineFilters virlog: Rename virLogAddOutputTo to virLogNewOutput virlog: Rename virLogDefineOutput to virLogOutputNew virlog: Rename virLogDefineFilter to virLogFilterNew virlog: Introduce virLogOutputFree virlog: Introduce virLogOutputListFree virlog: Introduce virLogFilterFree virlog: Introduce virLogFilterListFree virlog: Make virLogReset methods use of virLog(Output|Filter)ListFree virlog: Split output parsing and output defining to separate operations virlog: Split filter parsing and filter defining to separate operations virlog: Split parsing and setting priority virlog: Introduce virLogOutputExists virlog: Make use of virLogOutputExists virlog: Take a special care of syslog when setting new set of log outputs virlog: Swap the new copy of outputs with the global existing one virlog: Rename virLogFiltersSerial to virLogSerial virlog: Make virLogSetDefaultPriority trigger source update as well virlog: Introduce an API mutex that serializes all setters virlog: Acquire virLogAPILock in each setter API admin: Introduce virAdmConnectGetLoggingLevel admin: Introduce virAdmConnectGetLoggingFilters admin: Introduce virAdmConnectGetLoggingOutputs admin: Introduce virAdmConnectSetLoggingLevel admin: Introduce virAdmConnectSetLoggingFilters admin: Introduce virAdmConnectSetLoggingOutputs admin: Export logging level constants via libvirt-admin.h virt-admin: Wire-up the logging APIs daemon/admin.c | 125 +++++++ daemon/libvirtd.c | 8 +- include/libvirt/libvirt-admin.h | 37 +++ src/admin/admin_protocol.x | 73 ++++- src/admin/admin_remote.c | 86 +++++ src/admin_protocol-structs | 38 +++ src/libvirt-admin.c | 219 +++++++++++++ src/libvirt_admin_private.syms | 9 + src/libvirt_admin_public.syms | 6 + src/libvirt_private.syms | 14 +- src/locking/lock_daemon.c | 8 +- src/logging/log_daemon.c | 8 +- src/util/virlog.c | 698 +++++++++++++++++++++++++++------------- src/util/virlog.h | 50 +-- tests/eventtest.c | 3 +- tests/testutils.c | 19 +- tests/virlogtest.c | 12 +- tools/virt-admin.c | 208 ++++++++++++ 18 files changed, 1361 insertions(+), 260 deletions(-) -- 2.4.3

2 43

[libvirt] [PATCH 0/5] Make all crashes non CVEs
by Michal Privoznik 01 Apr '16

01 Apr '16

Currently, whenever a crash occurs it may be very serious. It may even get its own CVE. This patch set tries to resolve the problem. Michal Privoznik (5): Introduce virConnectCrash virutil: Introduce virCrash remote: Implement virConnectCrash qemu: Implement virConnectCrash virsh: Implement virConnectCrash cfg.mk | 2 ++ daemon/remote.c | 30 +++++++++++++++++ include/libvirt/libvirt-host.h | 15 +++++++++ src/driver-hypervisor.h | 7 ++++ src/libvirt-host.c | 45 +++++++++++++++++++++++++ src/libvirt_private.syms | 1 + src/libvirt_public.syms | 1 + src/qemu/qemu_driver.c | 29 ++++++++++++++++ src/remote/remote_driver.c | 32 ++++++++++++++++++ src/remote/remote_protocol.x | 14 +++++++- src/remote_protocol-structs | 5 +++ src/util/virutil.c | 32 ++++++++++++++++++ src/util/virutil.h | 2 ++ tools/virsh-host.c | 75 ++++++++++++++++++++++++++++++++++++++++++ tools/virsh.pod | 9 +++++ 15 files changed, 298 insertions(+), 1 deletion(-) -- 2.7.3

3 7

[libvirt] [PATCH] qemu: perf: Tweak flags before using them
by Peter Krempa 01 Apr '16

01 Apr '16

@flags have a valid modification impact only after calling virDomainObjUpdateModificationImpact. virDomainObjGetOneDef calls it but doesn't update them in the caller. --- CC: mprivozn(a)redhat.com src/qemu/qemu_driver.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index 4aa1625..0434438 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -10150,6 +10150,9 @@ qemuDomainGetPerfEvents(virDomainPtr dom, if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_QUERY) < 0) goto cleanup; + if (virDomainObjUpdateModificationImpact(vm, &flags) < 0) + goto endjob; + if (!(def = virDomainObjGetOneDef(vm, flags))) goto endjob; -- 2.8.0

2 2

[libvirt] [PATCH] migration: convert speed from Mb/sec to bytes/sec in drive-mirror jobs
by Rudy Zhang 01 Apr '16

01 Apr '16

Commit id '7b7600b3' added qemuMigrationDriveMirror to handle NBD mirroring, but passed the migrate_speed listed in MiB/s to be used for the mirror_speed which expects a bytes/s value. This patch will convert the migrate_speed value to its mirror_speed equivalent. Signed-off-by: Rudy Zhang <rudyflyzhang(a)gmail.com> --- src/qemu/qemu_migration.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) diff --git a/src/qemu/qemu_migration.c b/src/qemu/qemu_migration.c index 8bc76bf..3d9a55f 100644 --- a/src/qemu/qemu_migration.c +++ b/src/qemu/qemu_migration.c @@ -2093,12 +2093,21 @@ qemuMigrationDriveMirror(virQEMUDriverPtr driver, char *diskAlias = NULL; char *nbd_dest = NULL; char *hoststr = NULL; + unsigned long long mirror_speed = speed; unsigned int mirror_flags = VIR_DOMAIN_BLOCK_REBASE_REUSE_EXT; int rv; virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver); VIR_DEBUG("Starting drive mirrors for domain %s", vm->def->name); + if (mirror_speed > LLONG_MAX >> 20) { + virReportError(VIR_ERR_OVERFLOW, + _("bandwidth must be less than %llu"), + LLONG_MAX >> 20); + return ret; + } + mirror_speed <<= 20; + /* steal NBD port and thus prevent its propagation back to destination */ port = mig->nbd->port; mig->nbd->port = 0; @@ -2136,7 +2145,7 @@ qemuMigrationDriveMirror(virQEMUDriverPtr driver, qemuBlockJobSyncBegin(disk); /* Force "raw" format for NBD export */ mon_ret = qemuMonitorDriveMirror(priv->mon, diskAlias, nbd_dest, - "raw", speed, 0, 0, mirror_flags); + "raw", mirror_speed, 0, 0, mirror_flags); VIR_FREE(diskAlias); VIR_FREE(nbd_dest); -- 2.6.4

1 0

[libvirt] [PATCH] migration: convert speed from Mb/sec to bytes/sec in drive-mirror jobs
by Rudy Zhang 01 Apr '16

01 Apr '16

Block migration speed is differect from memory migration speed, because it not convert speed from Mb/sec to bytes/sec in the drive-mirror job. Signed-off-by: Rudy Zhang <rudyflyzhang(a)gmail.com> --- src/qemu/qemu_migration.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/src/qemu/qemu_migration.c b/src/qemu/qemu_migration.c index 8bc76bf..7648d8c 100644 --- a/src/qemu/qemu_migration.c +++ b/src/qemu/qemu_migration.c @@ -2135,8 +2135,8 @@ qemuMigrationDriveMirror(virQEMUDriverPtr driver, qemuBlockJobSyncBegin(disk); /* Force "raw" format for NBD export */ - mon_ret = qemuMonitorDriveMirror(priv->mon, diskAlias, nbd_dest, - "raw", speed, 0, 0, mirror_flags); + mon_ret = qemuMonitorDriveMirror(priv->mon, diskAlias, nbd_dest,"raw", + (unsigned long long)speed << 20, 0, 0, mirror_flags); VIR_FREE(diskAlias); VIR_FREE(nbd_dest); -- 2.6.4

3 3