February 2022 - Devel - libvirt List Archives

[PATCH 0/2] tests: Don't destroy locked mutexes
by Michal Privoznik 14 Feb '22

14 Feb '22

I've rewritten our virMutexes to check for failures [1] and saw couple of problems. In most cases we are destroying a locked mutex which these patches aim to fix. Then we have virLogLock() which is locked in virFork() and then unlocked from child (a different process) which is problematic. Pthread doesn't like it when one thread locks a mutex only so that another one unlocks it. Semaphores don't care. Anyway, leave virLogMutex aside, patches here fix problems at hand. BTW: commits v8.0.0-267-g39ac285c6b and v8.0.0-236-ga7201789ab were spotted because of this branch of mine. 1: https://gitlab.com/MichalPrivoznik/libvirt/-/commits/tests_mutext/ Michal Prívozník (2): tests: Track if @vm was created by qemuMonitorCommonTestNew() tests: Destroy domain object properly tests/qemuhotplugtest.c | 13 +++++-------- tests/qemumonitortestutils.c | 7 ++++++- tests/qemusecuritytest.c | 3 ++- tests/qemuxml2argvtest.c | 22 +++++++++++++--------- 4 files changed, 26 insertions(+), 19 deletions(-) -- 2.34.1

3 5

[PATCH v3 0/7] support mode option for dirtyrate calculation
by huangy81＠chinatelecom.cn 14 Feb '22

14 Feb '22

From: Hyman Huang(黄勇) <huangy81(a)chinatelecom.cn> v3: - Rebase the master - [PATCH v2 2/6]: Fix the usage of virQEMUCapsGet - [PATCH v2 4/6]: Fix the cleanup missed in qemuDomainStartDirtyRateCalc - [PATCH v2 4/6]: Move all blocks below ACL check - [PATCH v2 4/6]: Make the qemuMonitorJSONStartDirtyRateCalc cleaner by merging the different case of qemuMonitorJSONMakeCommand - [PATCH v2 4/6]: Code clean, make the error message not be line-broken - [PATCH v2 5/6]: Abstract the enum definition into a standalone commit - [PATCH v2 5/6]: Move the validations code above calculating flags block - [PATCH v2 6/6]: Change the type of 'value' field to unsigned in struct qemuMonitorDirtyRateVcpu - [PATCH v2 6/6]: Rename the enum type qemuMonitorDirtyRateCalcMode to virDomainDirtyRateCalcMode - [PATCH v2 6/6]: Code clean, align the code in qemuDomainGetStatsDirtyRate Thanks Peter for quick and precise response, please review. Regards Yong v2: Rebase master and fix confilicts with commit "Introduce QEMU_CAPS_DEVICE_VIRTIO_MEM_PCI_PREALLOC" Thanks ! v1: This patchset introduce mode option as the supplement of qemuDomainStartDirtyRateCalc api, add calc_mode for dirtyrate statistics correspondingly. Qemu add mode parameter for calc-dirty-rate command since >= 6.2.0, either of these three mode "page-sampling, dirty-bitmap, dirty-ring" can be specified when calculating dirty page rate. Page sampling is the original mode and used as default mode. Dirty bitmap mode use kvm log sync api to fetch the dirty-bitmap and count the increased 1 bits number during measurement, thus, calculate the dirty page rate. Dirty ring mode use the dirty-ring mechanism implemented in Qemu which can count the increased dirty page on virtual cpu granularity, thus, calculate the per-vcpu dirty page rate. These three calculation mode can be used in different scenarios, and the dirty-bitmap, dirty-ring mode may be more accurate to a certain degree. So maybe it's time to support the mode option for dirtyrate calculation. This series make main modifications as the following: 1. introduce QEMU_CAPS_CALC_DIRTY_RATE capability to probe calc-dirty-rate command in case of failure since it just introduced since >= 5.2.0 2. introduce QEMU_CAPS_DIRTYRATE_MODE capability to probe mode option of calc-dirty-rate command in case of failure, same as 1. 3. implement mode option support for dirtyrate calculation. Please review, thanks ! Best Regards ! Hyman Huang(黄勇) (7): qemu_capabilities: Introduce QEMU_CAPS_CALC_DIRTY_RATE capability qemu_driver: Probe capability before calculating dirty page rate qemu_capabilities: Introduce QEMU_CAPS_DIRTYRATE_MODE capability include: Introduce enum for qemuDomainStartDirtyRateCalc qemu_driver: Add mode parameter to qemuDomainStartDirtyRateCalc virsh: Add mode option to domdirtyrate-calc virsh api qemu_driver: Add calc_mode for dirtyrate statistics docs/manpages/virsh.rst | 7 ++- include/libvirt/libvirt-domain.h | 24 ++++++++ src/libvirt-domain.c | 17 +++++- src/qemu/qemu_capabilities.c | 4 ++ src/qemu/qemu_capabilities.h | 2 + src/qemu/qemu_driver.c | 48 +++++++++++++++- src/qemu/qemu_monitor.c | 5 +- src/qemu/qemu_monitor.h | 13 ++++- src/qemu/qemu_monitor_json.c | 69 ++++++++++++++++++++++- src/qemu/qemu_monitor_json.h | 3 +- tests/qemucapabilitiesdata/caps_5.2.0.aarch64.xml | 1 + tests/qemucapabilitiesdata/caps_5.2.0.ppc64.xml | 1 + tests/qemucapabilitiesdata/caps_5.2.0.riscv64.xml | 1 + tests/qemucapabilitiesdata/caps_5.2.0.s390x.xml | 1 + tests/qemucapabilitiesdata/caps_5.2.0.x86_64.xml | 1 + tests/qemucapabilitiesdata/caps_6.0.0.aarch64.xml | 1 + tests/qemucapabilitiesdata/caps_6.0.0.s390x.xml | 1 + tests/qemucapabilitiesdata/caps_6.0.0.x86_64.xml | 1 + tests/qemucapabilitiesdata/caps_6.1.0.x86_64.xml | 1 + tests/qemucapabilitiesdata/caps_6.2.0.aarch64.xml | 1 + tests/qemucapabilitiesdata/caps_6.2.0.ppc64.xml | 2 + tests/qemucapabilitiesdata/caps_6.2.0.x86_64.xml | 2 + tests/qemucapabilitiesdata/caps_7.0.0.ppc64.xml | 2 + tests/qemucapabilitiesdata/caps_7.0.0.x86_64.xml | 2 + tools/virsh-domain.c | 28 ++++++++- 25 files changed, 226 insertions(+), 12 deletions(-) -- 1.8.3.1

5 19

[PATCH 0/2] qemu: domain: Move and rename qemuDomainStorageIdNew/Reset
by Peter Krempa 14 Feb '22

14 Feb '22

In a review adding similar APIs I was asked to use 'ID' instead of 'Id'. Rename the existing helpers. Patch 1 is a second version as 'qemuDomainStorageIdNew' can be unexported too. Peter Krempa (2): qemu: domain: Move and unexport 'qemuDomainStorageIdNew/Reset' qemu: domain: Change 'Id' to 'ID' in qemuDomainStorageIdNew/Reset src/qemu/qemu_domain.c | 62 +++++++++++++++++++++--------------------- src/qemu/qemu_domain.h | 3 -- 2 files changed, 31 insertions(+), 34 deletions(-) -- 2.34.1

2 3

[libvirt PATCH] util: Fix getting CPU frequency on Apple Silicon
by Andrea Bolognani 14 Feb '22

14 Feb '22

The hw.cpufrequency sysctl, which we use to obtain the CPU frequency on macOS, is not available when running on Apple Silicon, and as a consequence we currently report an error whenever such information is requested. The virNodeInfo.mhz field, where the CPU frequency gets stored, is documented as being zero when the information could not be obtained, and we already do that for Linux on aarch64. Extend this behavior to macOS on Apple Silicon. Signed-off-by: Andrea Bolognani <abologna(a)redhat.com> --- src/util/virhostcpu.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) diff --git a/src/util/virhostcpu.c b/src/util/virhostcpu.c index a07c00a0e9..011ef8a153 100644 --- a/src/util/virhostcpu.c +++ b/src/util/virhostcpu.c @@ -928,8 +928,14 @@ virHostCPUGetInfo(virArch hostarch G_GNUC_UNUSED, *mhz = cpu_freq; # else if (sysctlbyname("hw.cpufrequency", &cpu_freq, &cpu_freq_len, NULL, 0) < 0) { - virReportSystemError(errno, "%s", _("cannot obtain CPU freq")); - return -1; + if (errno == ENOENT) { + /* The hw.cpufrequency sysctl is not implemented on Apple Silicon. + * In that case, we report 0 instead of erroring out */ + cpu_freq = 0; + } else { + virReportSystemError(errno, "%s", _("cannot obtain CPU freq")); + return -1; + } } *mhz = cpu_freq / 1000000; -- 2.34.1

3 4

[PATCH 00/27] qemu: Fix chardev hot(un)plug issues and sanitize use of /dev/fdset/N
by Peter Krempa 14 Feb '22

14 Feb '22

This series addresses the following issues: - hot-unplug of chardevs leaks the FDset in qemu - hot-plug of chardev doesn't use virtlogd even when configured - the name of the FD used for vdpa doesn't really tie it to the device There's also improvement in the common code handling the FD passing and few other drive-by fixes. Peter Krempa (27): scripts/moc-noinline: Use full name of the required annotation in error message qemuProcessPrepareHostBackendChardevFileHelper: Always use FD passing qemuProcessPrepareHostBackendChardev: Drop unneeded arguments qemuMonitorJSONQueryFdsetsParse: Don't check value passed to g_strdup qemuMonitorRemoveFdset: Convert @fdset to unsigned int to avoid error qemu: monitor: Make 'id' in 'struct _qemuMonitorFdsetInfo' unsigned qemu: domain: Move and unexport 'qemuDomainStorageIdReset' qemu: domain: Add helper for generating 'fdset' ids for VM startup qemu: Introduce helper functions for passing FDs to qemu qemuBuildInterfaceCommandLine: Use qemuFDPass for the vdpa fd qemu: hotplug: Extract code for unplugging fdsets qemuHotplugRemoveFDSet: Prepare for proper FD unplug handling qemuBuildInterfaceCommandLine: Use new pattern for naming the VDPA fdset qemu: Rewrite chardev startup code to use qemuFDPass qemuDomainRemoveChrDevice: Detach fdset after chardev hot-unplug qemumonitorjsontest: chardev: Remove need to allow unused commands qemumonitorjsontest: Refactor chardev hotplug testing qemuMonitorJSONTestAttachChardev: Add test for TLS-secured TCP chardev qemuMonitorJSONTestAttachChardev: Add logfile to some tests qemuMonitorJSONAttachCharDevGetProps: Properly handle private data tests: Move testPrepareHostBackendChardevOne into test utils qemuMonitorJSONTestAttachChardev: Add tests for FD passing of file backend qemu: process: Add a hotplug version of qemuProcessPrepareHostBackendChardev qemu: Honour 'virtlogd' use when hotplugging chardevs virTPMCreateCancelPath: Refactor value returning qemuBuildTPMOpenBackendFDs: Construct 'cancel_path' internally qemuBuildTPMCommandLine: Use 'qemuPassFD' infrastructure po/POTFILES.in | 1 + scripts/mock-noinline.py | 2 +- src/qemu/meson.build | 1 + src/qemu/qemu_command.c | 260 +++++---------- src/qemu/qemu_command.h | 4 +- src/qemu/qemu_domain.c | 94 +++--- src/qemu/qemu_domain.h | 13 +- src/qemu/qemu_driver.c | 3 +- src/qemu/qemu_fd.c | 296 ++++++++++++++++++ src/qemu/qemu_fd.h | 56 ++++ src/qemu/qemu_hotplug.c | 100 ++++-- src/qemu/qemu_hotplug.h | 2 +- src/qemu/qemu_monitor.c | 14 +- src/qemu/qemu_monitor.h | 4 +- src/qemu/qemu_monitor_json.c | 82 +++-- src/qemu/qemu_monitor_json.h | 2 +- src/qemu/qemu_process.c | 102 ++++-- src/qemu/qemu_process.h | 6 + src/util/virtpm.c | 13 +- tests/qemuhotplugmock.c | 11 + tests/qemuhotplugtest.c | 6 +- tests/qemumonitorjsontest.c | 253 +++++++++------ .../qemuxml2argvdata/aarch64-pci-serial.args | 3 +- .../name-escape.x86_64-2.11.0.args | 3 +- .../name-escape.x86_64-latest.args | 4 +- .../net-vdpa.x86_64-latest.args | 4 +- .../qemuxml2argvdata/serial-file-chardev.args | 3 +- .../serial-file-chardev.x86_64-latest.args | 4 +- tests/qemuxml2argvdata/serial-file-log.args | 4 +- .../serial-file-log.x86_64-latest.args | 6 +- .../qemuxml2argvdata/serial-many-chardev.args | 3 +- .../serial-many-chardev.x86_64-latest.args | 4 +- .../tpm-passthrough-crb.x86_64-latest.args | 6 +- .../tpm-passthrough.x86_64-latest.args | 6 +- tests/qemuxml2argvmock.c | 1 - tests/qemuxml2argvtest.c | 87 +---- tests/testutilsqemu.c | 106 +++++++ tests/testutilsqemu.h | 3 + 38 files changed, 1011 insertions(+), 561 deletions(-) create mode 100644 src/qemu/qemu_fd.c create mode 100644 src/qemu/qemu_fd.h -- 2.34.1

2 58

[libvirt PATCH] tests: Use freecon in testSELinuxCheckLabels
by Jiri Denemark 11 Feb '22

11 Feb '22

This partially reverts commit 0fc4a43d248b86fd54ad7323beb66faec8c1043c. Signed-off-by: Jiri Denemark <jdenemar(a)redhat.com> --- tests/securityselinuxlabeltest.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/tests/securityselinuxlabeltest.c b/tests/securityselinuxlabeltest.c index dfe9d38d34..83532d8090 100644 --- a/tests/securityselinuxlabeltest.c +++ b/tests/securityselinuxlabeltest.c @@ -230,7 +230,7 @@ testSELinuxCheckLabels(testSELinuxFile *files, size_t nfiles) size_t i; for (i = 0; i < nfiles; i++) { - g_autofree char *ctx = NULL; + char *ctx = NULL; if (getfilecon(files[i].file, &ctx) < 0) { if (errno == ENODATA) { /* nothing to do */ @@ -247,8 +247,10 @@ testSELinuxCheckLabels(testSELinuxFile *files, size_t nfiles) virReportError(VIR_ERR_INTERNAL_ERROR, "File %s context '%s' did not match expected '%s'", files[i].file, ctx, files[i].context); + freecon(ctx); return -1; } + freecon(ctx); } return 0; } -- 2.35.0

3 4

[libvirt PATCH] conf: Initialize devAddr in virNodeDeviceGetPCIVPDDynamicCap
by Jiri Denemark 11 Feb '22

11 Feb '22

Otherwise devAddr.multi would be uninitialized. Signed-off-by: Jiri Denemark <jdenemar(a)redhat.com> --- src/conf/node_device_conf.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/conf/node_device_conf.c b/src/conf/node_device_conf.c index 61c8715037..eaa42c05ee 100644 --- a/src/conf/node_device_conf.c +++ b/src/conf/node_device_conf.c @@ -3018,7 +3018,7 @@ static int virNodeDeviceGetPCIVPDDynamicCap(virNodeDevCapPCIDev *devCapPCIDev) { g_autoptr(virPCIDevice) pciDev = NULL; - virPCIDeviceAddress devAddr; + virPCIDeviceAddress devAddr = {0}; g_autoptr(virPCIVPDResource) res = NULL; devAddr.domain = devCapPCIDev->domain; -- 2.35.0

2 2

[PATCH] src: Initialize stack allocated virPCIDeviceAddress variables
by Michal Privoznik 11 Feb '22

11 Feb '22

There are few places where a virPCIDeviceAddress typed variable is allocated on the stack but it's not initialized. This can lead to random values of its members which in turn can lead to a random behaviour. Generated with help of the following spatch: @@ identifier I; @@ - virPCIDeviceAddress I; + virPCIDeviceAddress I = { 0 }; And then fixing bhyveAssignDevicePCISlots() which does declare the variable and then explicitly zero it by calling memset() only to set a specific member afterwards. Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com> --- src/bhyve/bhyve_device.c | 5 +---- src/conf/domain_addr.c | 2 +- src/conf/node_device_conf.c | 4 ++-- src/hypervisor/domain_driver.c | 6 +++--- src/node_device/node_device_udev.c | 2 +- src/qemu/qemu_domain_address.c | 4 ++-- src/util/virpci.c | 2 +- 7 files changed, 11 insertions(+), 14 deletions(-) diff --git a/src/bhyve/bhyve_device.c b/src/bhyve/bhyve_device.c index 36b93c0d4c..5654028ca5 100644 --- a/src/bhyve/bhyve_device.c +++ b/src/bhyve/bhyve_device.c @@ -83,10 +83,7 @@ bhyveAssignDevicePCISlots(virDomainDef *def, virDomainPCIAddressSet *addrs) { size_t i; - virPCIDeviceAddress lpc_addr; - - memset(&lpc_addr, 0, sizeof(lpc_addr)); - lpc_addr.slot = 0x1; + virPCIDeviceAddress lpc_addr = { .slot = 0x1 }; /* If the user didn't explicitly specify slot 1 for some of the devices, reserve it for LPC, even if there's no LPC device configured. diff --git a/src/conf/domain_addr.c b/src/conf/domain_addr.c index 49745ba881..49ca775a52 100644 --- a/src/conf/domain_addr.c +++ b/src/conf/domain_addr.c @@ -1181,7 +1181,7 @@ virDomainPCIAddressReserveNextAddr(virDomainPCIAddressSet *addrs, virDomainPCIConnectFlags flags, int function) { - virPCIDeviceAddress addr; + virPCIDeviceAddress addr = { 0 }; if (virDomainPCIAddressGetNextAddr(addrs, &addr, flags, dev->isolationGroup, function) < 0) diff --git a/src/conf/node_device_conf.c b/src/conf/node_device_conf.c index 61c8715037..8b20a7bee9 100644 --- a/src/conf/node_device_conf.c +++ b/src/conf/node_device_conf.c @@ -2947,7 +2947,7 @@ virNodeDeviceGetPCIIOMMUGroupCaps(virNodeDevCapPCIDev *pci_dev) { size_t i; int tmpGroup; - virPCIDeviceAddress addr; + virPCIDeviceAddress addr = { 0 }; /* this could be a refresh, so clear out the old data */ for (i = 0; i < pci_dev->nIommuGroupDevices; i++) @@ -3018,7 +3018,7 @@ static int virNodeDeviceGetPCIVPDDynamicCap(virNodeDevCapPCIDev *devCapPCIDev) { g_autoptr(virPCIDevice) pciDev = NULL; - virPCIDeviceAddress devAddr; + virPCIDeviceAddress devAddr = { 0 }; g_autoptr(virPCIVPDResource) res = NULL; devAddr.domain = devCapPCIDev->domain; diff --git a/src/hypervisor/domain_driver.c b/src/hypervisor/domain_driver.c index 2083f06287..bb1da7ac6b 100644 --- a/src/hypervisor/domain_driver.c +++ b/src/hypervisor/domain_driver.c @@ -375,7 +375,7 @@ virDomainDriverNodeDeviceReset(virNodeDevicePtr dev, virHostdevManager *hostdevMgr) { g_autoptr(virPCIDevice) pci = NULL; - virPCIDeviceAddress devAddr; + virPCIDeviceAddress devAddr = { 0 }; g_autoptr(virNodeDeviceDef) def = NULL; g_autofree char *xml = NULL; g_autoptr(virConnect) nodeconn = NULL; @@ -421,7 +421,7 @@ virDomainDriverNodeDeviceReAttach(virNodeDevicePtr dev, virHostdevManager *hostdevMgr) { g_autoptr(virPCIDevice) pci = NULL; - virPCIDeviceAddress devAddr; + virPCIDeviceAddress devAddr = { 0 }; g_autoptr(virNodeDeviceDef) def = NULL; g_autofree char *xml = NULL; g_autoptr(virConnect) nodeconn = NULL; @@ -466,7 +466,7 @@ virDomainDriverNodeDeviceDetachFlags(virNodeDevicePtr dev, const char *driverName) { g_autoptr(virPCIDevice) pci = NULL; - virPCIDeviceAddress devAddr; + virPCIDeviceAddress devAddr = { 0 }; g_autoptr(virNodeDeviceDef) def = NULL; g_autofree char *xml = NULL; g_autoptr(virConnect) nodeconn = NULL; diff --git a/src/node_device/node_device_udev.c b/src/node_device/node_device_udev.c index 3d5e25424a..b0a5e6302c 100644 --- a/src/node_device/node_device_udev.c +++ b/src/node_device/node_device_udev.c @@ -366,7 +366,7 @@ udevProcessPCI(struct udev_device *device, virNodeDevCapPCIDev *pci_dev = &def->caps->data.pci_dev; virPCIEDeviceInfo *pci_express = NULL; virPCIDevice *pciDev = NULL; - virPCIDeviceAddress devAddr; + virPCIDeviceAddress devAddr = { 0 }; int ret = -1; char *p; bool privileged = false; diff --git a/src/qemu/qemu_domain_address.c b/src/qemu/qemu_domain_address.c index 4e7095d3a7..dd0680f57f 100644 --- a/src/qemu/qemu_domain_address.c +++ b/src/qemu/qemu_domain_address.c @@ -1743,7 +1743,7 @@ qemuDomainValidateDevicePCISlotsPIIX3(virDomainDef *def, virDomainPCIAddressSet *addrs) { size_t i; - virPCIDeviceAddress tmp_addr; + virPCIDeviceAddress tmp_addr = { 0 }; g_autofree char *addrStr = NULL; virDomainPCIConnectFlags flags = (VIR_PCI_CONNECT_AUTOASSIGN | VIR_PCI_CONNECT_TYPE_PCI_DEVICE); @@ -1853,7 +1853,7 @@ qemuDomainValidateDevicePCISlotsQ35(virDomainDef *def, virDomainPCIAddressSet *addrs) { size_t i; - virPCIDeviceAddress tmp_addr; + virPCIDeviceAddress tmp_addr = { 0 }; g_autofree char *addrStr = NULL; virDomainPCIConnectFlags flags = VIR_PCI_CONNECT_TYPE_PCIE_DEVICE; diff --git a/src/util/virpci.c b/src/util/virpci.c index adc255f438..d141fde814 100644 --- a/src/util/virpci.c +++ b/src/util/virpci.c @@ -1865,7 +1865,7 @@ virPCIDeviceAddressIOMMUGroupIterate(virPCIDeviceAddress *orig, } while ((direrr = virDirRead(groupDir, &ent, groupPath)) > 0) { - virPCIDeviceAddress newDev; + virPCIDeviceAddress newDev = { 0 }; if (virPCIDeviceAddressParse(ent->d_name, &newDev) < 0) { virReportError(VIR_ERR_INTERNAL_ERROR, -- 2.34.1

2 1

[libvirt PATCH v2 0/2] qemu_migration_cookie: Properly fetch cert DN
by Jiri Denemark 11 Feb '22

11 Feb '22

Jiri Denemark (2): qemu_migration_cookie: Rename ret in qemuDomainExtractTLSSubject qemu_migration_cookie: Properly fetch cert DN src/qemu/qemu_migration_cookie.c | 34 ++++++++++++++++++++------------ 1 file changed, 21 insertions(+), 13 deletions(-) -- 2.35.0

2 3

[libvirt PATCH] conf: Avoid NULL-dereference in virDomainObjGetMessages
by Jiri Denemark 11 Feb '22

11 Feb '22

All callers currently guarantee flags passed to virDomainObjGetMessages are either zero or contain at least one of the supported flags. But it doesn't mean we should not check for the possibility an unknown flag was the only one passed to virDomainObjGetMessages. Signed-off-by: Jiri Denemark <jdenemar(a)redhat.com> --- src/conf/domain_conf.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c index ab8f2a52cc..093b719b2c 100644 --- a/src/conf/domain_conf.c +++ b/src/conf/domain_conf.c @@ -31472,7 +31472,8 @@ virDomainObjGetMessages(virDomainObj *vm, } } - (*msgs)[nmsgs] = NULL; + if (*msgs) + (*msgs)[nmsgs] = NULL; rv = nmsgs; -- 2.35.0

2 1