- Devel - libvirt List Archives

[libvirt] [PATCH] tests: fix #ifdef indentation from previous commit
by Daniel P. Berrangé 20 Aug '19

20 Aug '19

Signed-off-by: Daniel P. Berrangé <berrange(a)redhat.com> --- Pushed under the brown paper bag rule :-( tests/virpcimock.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/tests/virpcimock.c b/tests/virpcimock.c index e9a3791aea..c0dcd377d5 100644 --- a/tests/virpcimock.c +++ b/tests/virpcimock.c @@ -32,9 +32,9 @@ static int (*real_access)(const char *path, int mode); static int (*real_open)(const char *path, int flags, ...); -#ifdef __GLIBC__ +# ifdef __GLIBC__ static int (*real___open_2)(const char *path, int flags); -#endif /* ! __GLIBC__ */ +# endif /* ! __GLIBC__ */ static int (*real_close)(int fd); static DIR * (*real_opendir)(const char *name); static char *(*real_virFileCanonicalizePath)(const char *path); @@ -957,9 +957,9 @@ init_syms(void) VIR_MOCK_REAL_INIT(access); VIR_MOCK_REAL_INIT(open); -#ifdef __GLIBC__ +# ifdef __GLIBC__ VIR_MOCK_REAL_INIT(__open_2); -#endif /* ! __GLIBC__ */ +# endif /* ! __GLIBC__ */ VIR_MOCK_REAL_INIT(close); VIR_MOCK_REAL_INIT(opendir); VIR_MOCK_REAL_INIT(virFileCanonicalizePath); @@ -1088,7 +1088,7 @@ open(const char *path, int flags, ...) } -#ifdef __GLIBC__ +# ifdef __GLIBC__ /* in some cases this function may not be present in headers, so we need * a declaration to silence the complier */ int @@ -1118,7 +1118,7 @@ __open_2(const char *path, int flags) return ret; } -#endif /* ! __GLIBC__ */ +# endif /* ! __GLIBC__ */ DIR * opendir(const char *path) -- 2.21.0

1 0

[libvirt] [PATCH] tests: don't try to mock __open_2 on non-GLibc builds
by Daniel P. Berrangé 20 Aug '19

20 Aug '19

Mocking of the __open_2 function was added in commit 459f071cacf30af9df93b7d090b1bda71b0ef20f Author: Michal Privoznik <mprivozn(a)redhat.com> Date: Thu Aug 15 16:37:17 2019 +0200 virpcimock: Mock __open_2() This function only exists in glibc, however, and the mocking code runs on systems not using glibc, such as FreeBSD. Even Linux hosts might be using a different libc impl, though we don't actively try to support that. Signed-off-by: Daniel P. Berrangé <berrange(a)redhat.com> --- Pushed as a CI build fix for FreeBSD tests/virpcimock.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/tests/virpcimock.c b/tests/virpcimock.c index ca6cc060fd..e9a3791aea 100644 --- a/tests/virpcimock.c +++ b/tests/virpcimock.c @@ -32,7 +32,9 @@ static int (*real_access)(const char *path, int mode); static int (*real_open)(const char *path, int flags, ...); +#ifdef __GLIBC__ static int (*real___open_2)(const char *path, int flags); +#endif /* ! __GLIBC__ */ static int (*real_close)(int fd); static DIR * (*real_opendir)(const char *name); static char *(*real_virFileCanonicalizePath)(const char *path); @@ -955,7 +957,9 @@ init_syms(void) VIR_MOCK_REAL_INIT(access); VIR_MOCK_REAL_INIT(open); +#ifdef __GLIBC__ VIR_MOCK_REAL_INIT(__open_2); +#endif /* ! __GLIBC__ */ VIR_MOCK_REAL_INIT(close); VIR_MOCK_REAL_INIT(opendir); VIR_MOCK_REAL_INIT(virFileCanonicalizePath); @@ -1084,6 +1088,7 @@ open(const char *path, int flags, ...) } +#ifdef __GLIBC__ /* in some cases this function may not be present in headers, so we need * a declaration to silence the complier */ int @@ -1113,7 +1118,7 @@ __open_2(const char *path, int flags) return ret; } - +#endif /* ! __GLIBC__ */ DIR * opendir(const char *path) -- 2.21.0

1 0

[libvirt] [PATCHv2 00/11] util/resctrl cleanups and refactors
by Wang Huaqiang 20 Aug '19

20 Aug '19

Patches submitted for purpose of refactoring existing 'resctrl' related source code, including some code cleanups as well as some fixes. This is also a preparation for memory bandwidth monitor codes. Plan to support Resctrl Control Monitors, which is a feature introduced by kernel 'resctrl' sub-model. Submit some cleanup and refactoring patches for upcoming memory bandwidth resource monitoring (MBM) monitors. Related MBM RFC is https://www.redhat.com/archives/libvir-list/2019-April/msg01409.html. This RFC is not actively discussed since libvirt already implemented similar resctrl cache monitoring (CMT), and lots details have been discussed and implemented during the work of CMT. The cleanups and refactoring includes: v2 changes: 1. Addressed comments of v1. 2. Introduce a new algorithm for verifying new monitor vcpus and existing monitors and allocations. 3. Fixes for creating default-allocation-monitor in 'resctrl' file system.(patch 0001). v1 changes: 1. Removing some reluctant lines and white spaces that is existing in current code and not meet the libvirt coding style. 2. Replace 'virResctrlAllocIsEmpty' with @n==0 for indicating no resctrl allocation in configuration file. 3. Private API changes, removed 'virResctrlMonitorGetCacheOccupancy' and exported a new API named 'virResctrlMonitorGetStats' with similar functionality, but with capability to be used for retrieving MBM statistical information. 4. Refactoring 'virResctrlMonitorFreeStats' for more reusing in code. 5. Extend data structure 'virResctrlMonitorStats' with the capability to carry multiple statistical information from monitor. Wang Huaqiang (11): util,conf: Handle default monitor group of an allocation properly conf: code cleanup, remove empty line and one space conf: code cleanup for return error code directly conf: some code cleanup conf: refactor 'virDomainResctrlVcpuMatch' and some code cleanup conf: Append 'resctrl' object according to number of monitor group directly util: Refactor and rename 'virResctrlMonitorFreeStats' util: Refactor 'virResctrlMonitorStats' util: Extend virresctl API to retrieve multiple monitor statistics util: Remove unused virResctrlMonitorGetCacheOccupancy conf: Refactor and rename the function to validate a new resctrl monitor src/conf/domain_conf.c | 145 ++++++++++++++++++++++++----------------------- src/libvirt_private.syms | 5 +- src/qemu/qemu_driver.c | 41 ++++++++++---- src/qemu/qemu_process.c | 3 +- src/util/virresctrl.c | 75 ++++++++++-------------- src/util/virresctrl.h | 23 +++++--- 6 files changed, 156 insertions(+), 136 deletions(-) -- 2.7.4

5 19

[libvirt] [jenkins-ci PATCH] guests: remove obsolete install of intltool
by Daniel P. Berrangé 20 Aug '19

20 Aug '19

libvirt-glib, libvirt-sandbox and virt-viewer have all ditched usage of intltool, so this no longer needs to be installed. Signed-off-by: Daniel P. Berrangé <berrange(a)redhat.com> --- guests/vars/projects/libvirt-glib.yml | 1 - guests/vars/projects/libvirt-sandbox.yml | 1 - guests/vars/projects/virt-viewer.yml | 1 - 3 files changed, 3 deletions(-) diff --git a/guests/vars/projects/libvirt-glib.yml b/guests/vars/projects/libvirt-glib.yml index 13a5128..baa5b6a 100644 --- a/guests/vars/projects/libvirt-glib.yml +++ b/guests/vars/projects/libvirt-glib.yml @@ -3,6 +3,5 @@ packages: - glib2 - gobject-introspection - gtk-doc - - intltool - libxml2 - vala diff --git a/guests/vars/projects/libvirt-sandbox.yml b/guests/vars/projects/libvirt-sandbox.yml index 46b903f..f444333 100644 --- a/guests/vars/projects/libvirt-sandbox.yml +++ b/guests/vars/projects/libvirt-sandbox.yml @@ -4,7 +4,6 @@ packages: - glibc-static - gobject-introspection - gtk-doc - - intltool - libcap-ng - libselinux - libxml2 diff --git a/guests/vars/projects/virt-viewer.yml b/guests/vars/projects/virt-viewer.yml index cd32176..c03d50b 100644 --- a/guests/vars/projects/virt-viewer.yml +++ b/guests/vars/projects/virt-viewer.yml @@ -3,7 +3,6 @@ packages: - glib2 - gtk-vnc2 - gtk3 - - intltool - libgovirt - libxml2 - spice-gtk3 -- 2.21.0

3 2

Re: [libvirt] [Libvirt-ci] Build failed in Jenkins: virt-manager-check » libvirt-fedora-30 #1020
by Cole Robinson 19 Aug '19

19 Aug '19

On 8/19/19 3:44 PM, ci(a)centos.org wrote: > See <https://ci.centos.org/job/virt-manager-check/systems=libvirt-fedora-30/1020…> > > ------------------------------------------ > Started by upstream project "virt-manager-check" build number 1020 > originally caused by: > Started by upstream project "virt-manager-build" build number 1024 > originally caused by: > Started by an SCM change > Running as SYSTEM > [EnvInject] - Loading node environment variables. > Building remotely on libvirt-fedora-30 (libvirt) in workspace <https://ci.centos.org/job/virt-manager-check/systems=libvirt-fedora-30/ws/> > [virt-manager] $ /bin/sh -xe /tmp/jenkins5817546305801793439.sh > + /usr/bin/python3 ./setup.py test > running test New XML building in virt-install is triggering a crash in libvirt: https://www.redhat.com/archives/libvir-list/2019-August/msg00757.html - Cole

1 0

[libvirt] [PATCH] ci: Comment tweaks in Makefile.ci
by Eric Blake 19 Aug '19

19 Aug '19

Fix some typos and grammar (calling something safer and error-prone is odd, and 'ther eneeds' is an obvious typo), and reflow some long lines. Signed-off-by: Eric Blake <eblake(a)redhat.com> --- Makefile.ci | 38 +++++++++++++++++++++----------------- 1 file changed, 21 insertions(+), 17 deletions(-) diff --git a/Makefile.ci b/Makefile.ci index 8857c953b2..df2195ea72 100644 --- a/Makefile.ci +++ b/Makefile.ci @@ -86,7 +86,7 @@ endif # is liable to mess with SELinux labelling which will # then prevent the host accessing them. And podman cannot # relabel the files due to it running rootless. So -# copying them first is safer and error-prone. +# copying them first is safer and less error-prone. CI_PWDB_MOUNTS = \ --volume $(CI_SCRATCHDIR)/group:/etc/group:ro,z \ --volume $(CI_SCRATCHDIR)/passwd:/etc/passwd:ro,z \ @@ -98,22 +98,26 @@ CI_PWDB_MOUNTS = \ CI_ULIMIT_FILES = 1024 ifeq ($(CI_ENGINE),podman) - # Podman cannot reuse host namespace when running non-root containers. Until - # support for --keep-uid is added we can just create another mapping that will - # do that for us. Beware, that in {uid,git}map=container_id:host_id:range, - # the host_id does actually refer to the uid in the first mapping where 0 - # (root) is mapped to the current user and rest is offset. - - # In order to set up this mapping, we need to keep all the user IDs to prevent - # possible errors as some images might expect UIDs up to 90000 (looking at you - # fedora), so we don't want the overflowuid to be used for them. For mapping - # all the other users properly ther eneeds to be some math done. Don't worry, - # it's just addition and subtraction. - - # 65536 ought to be enough (tm), but for really rare cases the maximums might - # need to be higher, but that only happens when your /etc/sub{u,g}id allow - # users to have more IDs. Unless --keep-uid is supported, let's do this in a - # way that should work for everyone. + # Podman cannot reuse host namespace when running non-root + # containers. Until support for --keep-uid is added we can + # just create another mapping that will do that for us. + # Beware, that in {uid,git}map=container_id:host_id:range, the + # host_id does actually refer to the uid in the first mapping + # where 0 (root) is mapped to the current user and rest is + # offset. + # + # In order to set up this mapping, we need to keep all the + # user IDs to prevent possible errors as some images might + # expect UIDs up to 90000 (looking at you fedora), so we don't + # want the overflowuid to be used for them. For mapping all + # the other users properly, some math needs to be done. + # Don't worry, it's just addition and subtraction. + # + # 65536 ought to be enough (tm), but for really rare cases the + # maximums might need to be higher, but that only happens when + # your /etc/sub{u,g}id allow users to have more IDs. Unless + # --keep-uid is supported, let's do this in a way that should + # work for everyone. CI_MAX_UID = $(shell sed -n "s/^$USER:[^:]\+://p" /etc/subuid) CI_MAX_GID = $(shell sed -n "s/^$USER:[^:]\+://p" /etc/subgid) ifeq ($(CI_MAX_UID),) -- 2.20.1

1 1

[libvirt] [PATCH 1/1] qemuxml2xmltest.c: mock /etc/qemu/firmware path
by Daniel Henrique Barboza 19 Aug '19

19 Aug '19

Since commit 'b66ca0220a' the 'tpm-emulator' tests in qemuxml2xmltest started to fail unless 'make check' is either executed as root or with an user with access to /etc/qemu/firmware file. This is the error that happens in a regular 'make check' run with an unprivileged user: ------ 546) QEMU XML-2-XML-inactive tpm-emulator ... 1315 In '/home/danielhb/kvm-project/libvirt/tests/qemuxml2xmloutdata/tpm-emulator.xml': 1316 Offset 909 1317 Expect [1.2] 1318 Actual [default] 1319 ... Expected result code=0 but received code=4libvirt: error : cannot open directory '/home/danielhb/kvm-pr oject/libvirt/usr/etc/qemu/firmware': Permission denied 1320 FAILED 1321 547) QEMU XML-2-XML-active tpm-emulator ... 1322 In '/home/danielhb/kvm-project/libvirt/tests/qemuxml2xmloutdata/tpm-emulator.xml': 1323 Offset 909 1324 Expect [1.2] 1325 Actual [default] 1326 ... Expected result code=0 but received code=4libvirt: error : cannot open directory '/home/danielhb/kvm-pr oject/libvirt/usr/etc/qemu/firmware': Permission denied 1327 FAILED ------ The reason is that commit b66ca0220a changed the domain validation to fetch the capabilities in qemuDomainDeviceDefValidate time, directly impacting existing tpm-emulator tests in qemuxml2xmltest.c that that since then will attempt to read the /etc/qemu/firmware file that wasn't being mocked. This patch fixes it by adding the mock paths for /etc/qemu/firmware in qemuxml2xmltest.c as well, using the virFileWrapper API like it is done in commit 5b9819eedc71. Fixes: b66ca0220a ("qemu: domain: Call virDomainCapsDeviceDefValidate") Signed-off-by: Daniel Henrique Barboza <danielhb413(a)gmail.com> --- tests/Makefile.am | 3 ++- tests/qemuxml2xmltest.c | 13 +++++++++++++ 2 files changed, 15 insertions(+), 1 deletion(-) diff --git a/tests/Makefile.am b/tests/Makefile.am index 1c92e3ca6f..31a22ebefd 100644 --- a/tests/Makefile.am +++ b/tests/Makefile.am @@ -573,7 +573,8 @@ qemuxml2argvmock_la_LIBADD = $(MOCKLIBS_LIBS) qemuxml2xmltest_SOURCES = \ qemuxml2xmltest.c testutilsqemu.c testutilsqemu.h \ - testutils.c testutils.h + testutils.c testutils.h \ + virfilewrapper.c virfilewrapper.h qemuxml2xmltest_LDADD = $(qemu_LDADDS) qemumonitorjsontest_SOURCES = \ diff --git a/tests/qemuxml2xmltest.c b/tests/qemuxml2xmltest.c index 525eb9a740..51be6e0263 100644 --- a/tests/qemuxml2xmltest.c +++ b/tests/qemuxml2xmltest.c @@ -14,6 +14,9 @@ # include "qemu/qemu_domain.h" # include "testutilsqemu.h" # include "virstring.h" +# include "virfilewrapper.h" +# include "testutils.h" +# include "configmake.h" # define VIR_FROM_THIS VIR_FROM_NONE @@ -167,6 +170,15 @@ mymain(void) setenv("LIBVIRT_FAKE_ROOT_DIR", fakerootdir, 1); + /* Required for tpm-emulator tests + */ + virFileWrapperAddPrefix(SYSCONFDIR "/qemu/firmware", + abs_srcdir "/qemufirmwaredata/etc/qemu/firmware"); + virFileWrapperAddPrefix(PREFIX "/share/qemu/firmware", + abs_srcdir "/qemufirmwaredata/usr/share/qemu/firmware"); + virFileWrapperAddPrefix("/home/user/.config/qemu/firmware", + abs_srcdir "/qemufirmwaredata/home/user/.config/qemu/firmware"); + if (qemuTestDriverInit(&driver) < 0) return EXIT_FAILURE; @@ -1325,6 +1337,7 @@ mymain(void) virHashFree(capslatest); qemuTestDriverFree(&driver); VIR_FREE(fakerootdir); + virFileWrapperClearPrefixes(); return ret == 0 ? EXIT_SUCCESS : EXIT_FAILURE; } -- 2.21.0

3 3

[libvirt] [PATCH] src: Don't check lxc_monitor_protocol-struct when LXC is disabled
by Michal Privoznik 19 Aug '19

19 Aug '19

If LXC is disabled at build time then there is no libvirt_driver_lxc_impl_la-*.lo to run the 'check-protocol' against. Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com> --- src/Makefile.am | 2 ++ 1 file changed, 2 insertions(+) diff --git a/src/Makefile.am b/src/Makefile.am index 3ae4b87abb..817a7ecf34 100644 --- a/src/Makefile.am +++ b/src/Makefile.am @@ -318,9 +318,11 @@ $(srcdir)/remote_protocol-struct \ $(srcdir)/virnetprotocol-struct $(srcdir)/virkeepaliveprotocol-struct: \ $(srcdir)/%-struct: rpc/libvirt_net_rpc_la-%.lo $(PDWTAGS) +if WITH_LXC $(srcdir)/lxc_monitor_protocol-struct: \ $(srcdir)/%-struct: lxc/libvirt_driver_lxc_impl_la-%.lo $(PDWTAGS) +endif WITH_LXC $(srcdir)/lock_protocol-struct: \ $(srcdir)/%-struct: locking/lockd_la-%.lo $(PDWTAGS) -- 2.21.0

2 1

[libvirt] [PATCH] qemu: alias: Generate 'qomName' of disk with useraliases
by Peter Krempa 19 Aug '19

19 Aug '19

Commit fb64e176f4f forgot to delete the check that short-circuits the disk alias creation if the alias is already present. The side efect of this is that the creation qomName which is necessary to be able to refer to disk frontends when -blockdev is used was skipped when user aliases are used. Fix it by deleting the check. Also prevent any potential memory leaks from calling this function repeatedly by creating the qomName only when it's not present. https://bugzilla.redhat.com/show_bug.cgi?id=1741838 Signed-off-by: Peter Krempa <pkrempa(a)redhat.com> --- src/qemu/qemu_alias.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) diff --git a/src/qemu/qemu_alias.c b/src/qemu/qemu_alias.c index 585cc972ba..216a18d0d9 100644 --- a/src/qemu/qemu_alias.c +++ b/src/qemu/qemu_alias.c @@ -182,9 +182,6 @@ qemuAssignDeviceDiskAlias(virDomainDefPtr def, const char *prefix = virDomainDiskBusTypeToString(disk->bus); int controllerModel = -1; - if (disk->info.alias) - return 0; - if (!disk->info.alias) { if (disk->info.type == VIR_DOMAIN_DEVICE_ADDRESS_TYPE_DRIVE) { if (disk->bus == VIR_DOMAIN_DISK_BUS_SCSI) { @@ -220,7 +217,8 @@ qemuAssignDeviceDiskAlias(virDomainDefPtr def, * on the alias in qemu. While certain disk types use just the alias, some * need the full path into /machine/peripheral as a historical artifact. */ - if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_BLOCKDEV)) { + if (!diskPriv->qomName && + virQEMUCapsGet(qemuCaps, QEMU_CAPS_BLOCKDEV)) { switch ((virDomainDiskBus) disk->bus) { case VIR_DOMAIN_DISK_BUS_FDC: case VIR_DOMAIN_DISK_BUS_IDE: -- 2.21.0

2 1

[libvirt] [PATCH] util: mdev: support persistent devices with mdevctl
by Jonathon Jongsma 19 Aug '19

19 Aug '19

When a host is rebooted, mediated devices disappear from sysfs. mdevctl (https://github.com/mdevctl/mdevctl) is a utility for managing and persisting these devices. It maintains a registry of mediated devices and can start and stop them by UUID. when libvirt attempts to create a new mediated device object, we currently fail if the path does not exist in sysfs. This patch tries a little bit harder by using mdevctl to attempt to activate the device. The approach is fairly naive -- it just calls 'mdevctl start -u $UUID' without checking whether this UUID is registered with mdevctl or not. See https://bugzilla.redhat.com/show_bug.cgi?id=1699274 Signed-off-by: Jonathon Jongsma <jjongsma(a)redhat.com> --- NOTES: - an argument could be made that we should simply do nothing here. mdevctl does have support for automatically activating the mediated device when the parent device becomes available (via udev). So if the administrator set up the mdev to start automatically, this patch should not even be necessary. That said, I think this patch could still be useful. - As I said above, the approach is pretty naive. I could have checked whether the device was defined in mdevctl's registry and given a different error message ("try registering this device with mdevctl") in that scenario. But I chose to keep it simple. src/util/virmdev.c | 22 +++++++++++++++++++--- 1 file changed, 19 insertions(+), 3 deletions(-) diff --git a/src/util/virmdev.c b/src/util/virmdev.c index 3d5488cdae..7a2f0adedc 100644 --- a/src/util/virmdev.c +++ b/src/util/virmdev.c @@ -25,6 +25,7 @@ #include "virfile.h" #include "virstring.h" #include "viralloc.h" +#include "vircommand.h" #define VIR_FROM_THIS VIR_FROM_NONE @@ -148,9 +149,24 @@ virMediatedDeviceNew(const char *uuidstr, virMediatedDeviceModelType model) return NULL; if (!virFileExists(sysfspath)) { - virReportError(VIR_ERR_DEVICE_MISSING, - _("mediated device '%s' not found"), uuidstr); - return NULL; + bool activated = false; + /* if the mdev device path doesn't exist, we may still be able to start + * the device using mdevctl + */ + char *mdevctl = virFindFileInPath("mdevctl"); + if (mdevctl) { + const char *runargs[] = { mdevctl, "start", "-u", uuidstr, NULL }; + if (virRun(runargs, NULL) == 0) { + /* check to see if it the device exists now */ + activated = virFileExists(sysfspath); + } + } + + if (!activated) { + virReportError(VIR_ERR_DEVICE_MISSING, + _("mediated device '%s' not found"), uuidstr); + return NULL; + } } if (VIR_ALLOC(dev) < 0) -- 2.21.0

4 4