July 2013 - Devel - libvirt List Archives

[libvirt] [PATCH] caps: use -device for primary video when qemu >=1.6
by Guannan Ren 29 Jul '13

29 Jul '13

https://bugzilla.redhat.com/show_bug.cgi?id=981094 The commit 0ad9025ef introduce qemu flag QEMU_CAPS_DEVICE_VIDEO_PRIMARY for using -device VGA, -device cirrus-vga, -device vmware-svga and -device qxl-vga. In use, for -device qxl-vga, mouse doesn't display in guest window like the desciption in above bug. This patch try to use -device for primary video when qemu >=1.6 which is safe. --- src/qemu/qemu_capabilities.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/src/qemu/qemu_capabilities.c b/src/qemu/qemu_capabilities.c index 5dc3c9e..08406b8 100644 --- a/src/qemu/qemu_capabilities.c +++ b/src/qemu/qemu_capabilities.c @@ -1189,8 +1189,6 @@ virQEMUCapsComputeCmdFlags(const char *help, virQEMUCapsSet(qemuCaps, QEMU_CAPS_VNC_SHARE_POLICY); } - if (version >= 1002000) - virQEMUCapsSet(qemuCaps, QEMU_CAPS_DEVICE_VIDEO_PRIMARY); return 0; } @@ -2424,7 +2422,6 @@ virQEMUCapsInitQMPBasic(virQEMUCapsPtr qemuCaps) virQEMUCapsSet(qemuCaps, QEMU_CAPS_MACHINE_OPT); virQEMUCapsSet(qemuCaps, QEMU_CAPS_DUMP_GUEST_CORE); virQEMUCapsSet(qemuCaps, QEMU_CAPS_VNC_SHARE_POLICY); - virQEMUCapsSet(qemuCaps, QEMU_CAPS_DEVICE_VIDEO_PRIMARY); } /* Capabilities that are architecture depending @@ -2597,6 +2594,9 @@ virQEMUCapsInitQMP(virQEMUCapsPtr qemuCaps, if (qemuCaps->version >= 1003001) virQEMUCapsSet(qemuCaps, QEMU_CAPS_VNC_WEBSOCKET); + if (qemuCaps->version >= 1006000) + virQEMUCapsSet(qemuCaps, QEMU_CAPS_DEVICE_VIDEO_PRIMARY); + if (virQEMUCapsProbeQMPCommands(qemuCaps, mon) < 0) goto cleanup; if (virQEMUCapsProbeQMPEvents(qemuCaps, mon) < 0) -- 1.8.3.1

3 5

Re: [libvirt] Entering freeze for libvirt-1.1.1
by Daniel Veillard 29 Jul '13

29 Jul '13

On Fri, Jul 26, 2013 at 06:01:23AM -0400, Alex Jia wrote: > Hello Daniel, > I gave a basic try for libvirt-1.1.1-rc1 on RHEL6. > > [root@202 libvirt-1.1.1]# make > > Notes, sometimes, I can meet the following errors, it's not reproducible each time, > maybe, it's a network issue. > > <slice> > > I/O error : Attempt to load network entity http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd > formatdomain.html.in:2: warning: failed to load external entity "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd" > C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd" > ^ > formatdomain.html.in:4582: parser error : Entity 'mdash' not defined > <li> 'i6300esb' — the recommended device, you need to install xhtml1-dtds for local validation of generated HTML [...] > TEST: virnetsockettest > ........!!!.!!! 15 FAIL weird ... > Notes, works well with PATH environment setting, but it still is failed for > running 'make -C tests valgrind' due to previous 'make check' fails, whether > we my bypass/fix this for 'make check'. BTW, I also checked libvirt upstream > with "make -C tests valgrind", it's fine and without memory leaks. > > > [root@202 libvirt-1.1.1]# make syntax-check > GEN HACKING > GEN bracket-spacing-check > ./build-aux/vc-list-files: Failed to determine type of version control used in /home/ajia/Workspace/rc1/libvirt-1.1.1 > Can't open perl script "./build-aux/bracket-spacing.pl": No such file or directory > maint.mk: incorrect whitespace, see HACKING for rules > make: *** [bracket-spacing-check] Error 1 > > [root@202 libvirt-1.1.1]# ls ./build-aux/bracket-spacing.pl > ls: cannot access ./build-aux/bracket-spacing.pl: No such file or directory > > Notes, it's okay on libvirt upstream and can find ./build-aux/bracket-spacing.pl file. > > > To run coverity on libvirt-1.1.1-rc1 and attach report as an attachment. thanks ! Daniel -- Daniel Veillard | Open Source and Standards, Red Hat veillard(a)redhat.com | libxml Gnome XML XSLT toolkit http://xmlsoft.org/ http://veillard.com/ | virtualization library http://libvirt.org/

3 2

[libvirt] [PATCH] build: avoid uninitialized use warning
by Eric Blake 27 Jul '13

27 Jul '13

Otherwise, with new enough gcc compiling at -O2, the build fails with: ../../src/conf/domain_conf.c: In function ‘virDomainDeviceDefPostParse’: ../../src/conf/domain_conf.c:2821:29: error: ‘cnt’ may be used uninitialized in this function [-Werror=maybe-uninitialized] for (i = 0; i < *cnt; i++) { ^ ../../src/conf/domain_conf.c:2795:20: note: ‘cnt’ was declared here size_t i, *cnt; ^ ../../src/conf/domain_conf.c:2794:30: error: ‘arrPtr’ may be used uninitialized in this function [-Werror=maybe-uninitialized] virDomainChrDefPtr **arrPtr; ^ * src/conf/domain_conf.c (virDomainChrGetDomainPtrs): Always assign into output parameters. Signed-off-by: Eric Blake <eblake(a)redhat.com> --- Pushing under the build-breaker rule. I actually found this while running ./autobuild.sh; but there, we run libtool in both static and shared modes, and libtool, in its "infinite wisdom", assumes that if the static build succeeds, then the shared build can have compiler output redirected to /dev/null, which meant I was stuck debugging this output: CC libvirt_driver_test_la-test_driver.lo CC libvirt_driver_remote_la-remote_driver.lo CC libvirt_driver_remote_la-remote_protocol.lo make[3]: *** [libvirt_conf_la-domain_conf.lo] Error 1 make[3]: *** Waiting for unfinished jobs.... thanks for nothing, libtool. src/conf/domain_conf.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c index 0e74039..a86be8c 100644 --- a/src/conf/domain_conf.c +++ b/src/conf/domain_conf.c @@ -10221,6 +10221,8 @@ virDomainChrGetDomainPtrs(virDomainDefPtr vmdef, break; case VIR_DOMAIN_CHR_DEVICE_TYPE_LAST: + *arrPtr = NULL; + *cntPtr = NULL; break; } } -- 1.8.3.1

1 0

[libvirt] [PATCH] Fix probing of legacy Xen driver to not leave URI set
by Daniel P. Berrange 26 Jul '13

26 Jul '13

From: "Daniel P. Berrange" <berrange(a)redhat.com> When the legacy Xen driver probes with a NULL URI, and finds itself running on Xen, it will set conn->uri. A little bit later though it checks to see if libxl support exists, and if so declines the driver. This leaves the conn->uri set to 'xen:///', so if libxl also declines it, it prevents probing of the QEMU driver. Once a driver has set the conn->uri, it must *never* decline an open request. So we must move the libxl check earlier Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com> --- src/xen/xen_driver.c | 20 +++++++++++++------- 1 file changed, 13 insertions(+), 7 deletions(-) diff --git a/src/xen/xen_driver.c b/src/xen/xen_driver.c index 39334b7..4ae38d3 100644 --- a/src/xen/xen_driver.c +++ b/src/xen/xen_driver.c @@ -365,6 +365,13 @@ xenUnifiedConnectOpen(virConnectPtr conn, virConnectAuthPtr auth, unsigned int f if (!xenUnifiedProbe()) return VIR_DRV_OPEN_DECLINED; +#ifdef WITH_LIBXL + /* Decline xen:// URI if xend is not running and libxenlight + * driver is potentially available. */ + if (!xenUnifiedXendProbe()) + return VIR_DRV_OPEN_DECLINED; +#endif + if (!(conn->uri = virURIParse("xen:///"))) return VIR_DRV_OPEN_ERROR; } else { @@ -374,6 +381,12 @@ xenUnifiedConnectOpen(virConnectPtr conn, virConnectAuthPtr auth, unsigned int f STRCASENEQ(conn->uri->scheme, "http")) return VIR_DRV_OPEN_DECLINED; +#ifdef WITH_LIBXL + /* Decline xen:// URI if xend is not running and libxenlight + * driver is potentially available. */ + if (!xenUnifiedXendProbe()) + return VIR_DRV_OPEN_DECLINED; +#endif /* Return an error if the path isn't '' or '/' */ if (conn->uri->path && @@ -395,13 +408,6 @@ xenUnifiedConnectOpen(virConnectPtr conn, virConnectAuthPtr auth, unsigned int f } } -#ifdef WITH_LIBXL - /* Decline xen:// URI if xend is not running and libxenlight - * driver is potentially available. */ - if (!xenUnifiedXendProbe()) - return VIR_DRV_OPEN_DECLINED; -#endif - /* We now know the URI is definitely for this driver, so beyond * here, don't return DECLINED, always use ERROR */ -- 1.8.3.1

2 1

[libvirt] [PATCH] Set default partition in libvirtd instead of libvirt_lxc
by Daniel P. Berrange 26 Jul '13

26 Jul '13

From: "Daniel P. Berrange" <berrange(a)redhat.com> By setting the default partition in libvirt_lxc it is not visible when querying the live XML. Move setting of the default partition into libvirtd virLXCProcessStart Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com> --- src/lxc/lxc_cgroup.c | 14 -------------- src/lxc/lxc_process.c | 14 ++++++++++++++ 2 files changed, 14 insertions(+), 14 deletions(-) diff --git a/src/lxc/lxc_cgroup.c b/src/lxc/lxc_cgroup.c index af91b04..0b0ca02 100644 --- a/src/lxc/lxc_cgroup.c +++ b/src/lxc/lxc_cgroup.c @@ -433,20 +433,6 @@ virCgroupPtr virLXCCgroupCreate(virDomainDefPtr def) { virCgroupPtr cgroup = NULL; - if (!def->resource) { - virDomainResourceDefPtr res; - - if (VIR_ALLOC(res) < 0) - goto cleanup; - - if (VIR_STRDUP(res->partition, "/machine") < 0) { - VIR_FREE(res); - goto cleanup; - } - - def->resource = res; - } - if (def->resource->partition[0] != '/') { virReportError(VIR_ERR_CONFIG_UNSUPPORTED, _("Resource partition '%s' must start with '/'"), diff --git a/src/lxc/lxc_process.c b/src/lxc/lxc_process.c index 1a5686f..247e516 100644 --- a/src/lxc/lxc_process.c +++ b/src/lxc/lxc_process.c @@ -1007,6 +1007,20 @@ int virLXCProcessStart(virConnectPtr conn, return -1; } + if (!vm->def->resource) { + virDomainResourceDefPtr res; + + if (VIR_ALLOC(res) < 0) + goto cleanup; + + if (VIR_STRDUP(res->partition, "/machine") < 0) { + VIR_FREE(res); + goto cleanup; + } + + vm->def->resource = res; + } + if (virAsprintf(&logfile, "%s/%s.log", cfg->logDir, vm->def->name) < 0) return -1; -- 1.8.1.4

2 1

[libvirt] [PATCH v2 0/2] Split driver StateAutoStart from StateInitialization
by John Ferlan 26 Jul '13

26 Jul '13

This is an update to change the 'int' AutoStart function to 'void' as was discussed in the initial review: https://www.redhat.com/archives/libvir-list/2013-July/msg01674.html John Ferlan (2): Separate out StateAutoStart from StateInitialize virStateDriver - Separate AutoStart from Initialize src/driver.h | 4 ++++ src/libvirt.c | 14 +++++++++++++- src/libxl/libxl_driver.c | 16 +++++++++++++--- src/lxc/lxc_driver.c | 16 ++++++++++++++-- src/network/bridge_driver.c | 18 +++++++++++++++++- src/qemu/qemu_driver.c | 17 +++++++++++++++-- src/storage/storage_driver.c | 18 +++++++++++++++++- src/uml/uml_driver.c | 17 +++++++++++++++-- 8 files changed, 108 insertions(+), 12 deletions(-) -- 1.8.1.4

1 3

[libvirt] [PATCH 0/7] Probe QEMU binary for host CPU and use it for computations
by Jiri Denemark 26 Jul '13

26 Jul '13

Since QEMU and kvm may filter some host CPU features or add efficiently emulated features, asking QEMU binary for host CPU data provides better results when we later use the data for building guest CPUs. Jiri Denemark (7): cpu: Add support for loading and storing CPU data cpu: Export few x86-specific APIs x86: Ignore CPUID functions greater than 10 qemu: Add monitor APIs to fetch CPUID data from QEMU qemu: Make QMP probing process reusable qemu: Probe QEMU binary for host CPU qemu: Use host CPU from QEMU for computations src/cpu/cpu.c | 41 ++++ src/cpu/cpu.h | 13 ++ src/cpu/cpu_x86.c | 161 +++++++++++--- src/cpu/cpu_x86.h | 10 + src/cpu/cpu_x86_data.h | 1 + src/libvirt_private.syms | 9 + src/qemu/qemu_capabilities.c | 234 ++++++++++++++------- src/qemu/qemu_capabilities.h | 2 + src/qemu/qemu_command.c | 32 ++- src/qemu/qemu_domain.c | 21 +- src/qemu/qemu_monitor.c | 21 ++ src/qemu/qemu_monitor.h | 3 + src/qemu/qemu_monitor_json.c | 162 ++++++++++++++ src/qemu/qemu_monitor_json.h | 6 + tests/Makefile.am | 1 + .../qemumonitorjson-getcpu-empty.data | 2 + .../qemumonitorjson-getcpu-empty.json | 46 ++++ .../qemumonitorjson-getcpu-filtered.data | 4 + .../qemumonitorjson-getcpu-filtered.json | 46 ++++ .../qemumonitorjson-getcpu-full.data | 4 + .../qemumonitorjson-getcpu-full.json | 46 ++++ .../qemumonitorjson-getcpu-host.data | 5 + .../qemumonitorjson-getcpu-host.json | 45 ++++ tests/qemumonitorjsontest.c | 74 +++++++ 24 files changed, 881 insertions(+), 108 deletions(-) create mode 100644 tests/qemumonitorjsondata/qemumonitorjson-getcpu-empty.data create mode 100644 tests/qemumonitorjsondata/qemumonitorjson-getcpu-empty.json create mode 100644 tests/qemumonitorjsondata/qemumonitorjson-getcpu-filtered.data create mode 100644 tests/qemumonitorjsondata/qemumonitorjson-getcpu-filtered.json create mode 100644 tests/qemumonitorjsondata/qemumonitorjson-getcpu-full.data create mode 100644 tests/qemumonitorjsondata/qemumonitorjson-getcpu-full.json create mode 100644 tests/qemumonitorjsondata/qemumonitorjson-getcpu-host.data create mode 100644 tests/qemumonitorjsondata/qemumonitorjson-getcpu-host.json -- 1.8.3.2

5 32

[libvirt] [PATCH 0/4] fix two issues about diskchain
by Guannan Ren 26 Jul '13

26 Jul '13

1, Report an backing missing error on guest boot if its disks have broken backing file chain 2, List volume even if its diskchain is broken Guannan Ren(4) [PATCH 1/4] qemu: refactor qemuDomainCheckDiskPresence for only disk [PATCH 2/4] qemu: report error if disk backing files doesn't exist [PATCH 3/4] qemu: check presence of each disk in chain [PATCH 4/4] storage: list volumes even if its diskchain is broken src/qemu/qemu_domain.c | 108 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-------------------------------------------- src/qemu/qemu_process.c | 7 ------- src/storage/storage_backend_fs.c | 4 +++- src/util/virstoragefile.c | 23 +++++++++++++++-------- tests/virstoragetest.c | 16 ++++++++-------- 5 files changed, 90 insertions(+), 68 deletions(-)

2 8

[libvirt] [PATCH] build: avoid -lgcrypt with newer gnutls
by Eric Blake 26 Jul '13

26 Jul '13

https://bugzilla.redhat.com/show_bug.cgi?id=951637 Newer gnutls uses nettle, rather than gcrypt, which is a lot nicer regarding initialization. Yet we were unconditionally initializing gcrypt even when gnutls wouldn't be using it, and having two crypto libraries linked into libvirt.so is pointless. The ldd probe in configure borrows from our libnl-1 vs. libnl-3 code. * configure.ac (WITH_GNUTLS): Probe whether to add -lgcrypt, and define a witness WITH_GNUTLS_GCRYPT. * src/libvirt.c (virTLSMutexInit, virTLSMutexDestroy) (virTLSMutexLock, virTLSMutexUnlock, virTLSThreadImpl) (virGlobalInit): Honor the witness. * libvirt.spec.in (BuildRequires): Make gcrypt usage conditional, no longer needed in Fedora 19. Signed-off-by: Eric Blake <eblake(a)redhat.com> --- Tested with 'ldd src/.libs/libvirt.so | grep -E "(gcry|net|tls)"': - on RHEL 6.4 and Fedora 18, pre- and post-patch remain unchanged (use of just libgnutls/libgcrypt) - on Fedora 19, pre-patch linked against libgnutls, libgcrypt, and libnettle, post-patch linked against just libgnutls and libnettle This should probably go in for 1.1.1, but it's not a build-breaker so it needs review. configure.ac | 27 +++++++++++++++++++++------ libvirt.spec.in | 2 ++ src/libvirt.c | 10 ++++++---- 3 files changed, 29 insertions(+), 10 deletions(-) diff --git a/configure.ac b/configure.ac index cc9942a..de209e2 100644 --- a/configure.ac +++ b/configure.ac @@ -1098,13 +1098,28 @@ if test "x$with_gnutls" != "xno"; then AC_MSG_ERROR([You must install the GnuTLS library in order to compile and run libvirt]) fi else - dnl Not all versions of gnutls include -lgcrypt, and so we add - dnl it explicitly for the calls to gcry_control/check_version - GNUTLS_LIBS="$GNUTLS_LIBS -lgcrypt" - dnl We're not using gcrypt deprecated features so define - dnl GCRYPT_NO_DEPRECATED to avoid deprecated warnings - GNUTLS_CFLAGS="$GNUTLS_CFLAGS -DGCRYPT_NO_DEPRECATED" + dnl If gnutls linked against -lgcrypt, then we must initialize gcrypt + dnl prior to using gnutls. Newer versions of gnutls use -lnettle, in + dnl which case we don't want to drag in gcrypt ourselves. + gnutls_ldd= + for dir in /usr/lib64 /usr/lib /usr/lib/*-linux-gnu*; do + if test -f $dir/libgnutls.so; then + gnutls_ldd=`(ldd $dir/libgnutls.so) 2>&1` + break + fi + done + case $gnutls_ldd in + '' | *libgcrypt.so*) + GNUTLS_LIBS="$GNUTLS_LIBS -lgcrypt" + dnl We're not using gcrypt deprecated features so define + dnl GCRYPT_NO_DEPRECATED to avoid deprecated warnings + GNUTLS_CFLAGS="$GNUTLS_CFLAGS -DGCRYPT_NO_DEPRECATED" + AC_DEFINE_UNQUOTED([WITH_GNUTLS_GCRYPT], 1, + [whether GNUTLS uses gcrypt]) + ;; + *) ;; # Assume no gcrypt usage + esac dnl gnutls 3.x moved some declarations to a new header AC_CHECK_HEADERS([gnutls/crypto.h], [], [], [[ diff --git a/libvirt.spec.in b/libvirt.spec.in index e0e0004..4320281 100644 --- a/libvirt.spec.in +++ b/libvirt.spec.in @@ -422,7 +422,9 @@ BuildRequires: readline-devel BuildRequires: ncurses-devel BuildRequires: gettext BuildRequires: libtasn1-devel +%if (0%{?rhel} && 0%{?rhel} < 7) || (0%{?fedora} && 0%{?fedora} < 19) BuildRequires: libgcrypt-devel +%endif BuildRequires: gnutls-devel BuildRequires: libattr-devel %if %{with_libvirtd} diff --git a/src/libvirt.c b/src/libvirt.c index 444c1c3..9775b97 100644 --- a/src/libvirt.c +++ b/src/libvirt.c @@ -55,7 +55,9 @@ #include "intprops.h" #include "virconf.h" #if WITH_GNUTLS -# include <gcrypt.h> +# if WITH_GNUTLS_GCRYPT +# include <gcrypt.h> +# endif # include "rpc/virnettlscontext.h" #endif #include "vircommand.h" @@ -270,7 +272,7 @@ winsock_init(void) #endif -#ifdef WITH_GNUTLS +#ifdef WITH_GNUTLS_GCRYPT static int virTLSMutexInit(void **priv) { virMutexPtr lock = NULL; @@ -323,7 +325,7 @@ static struct gcry_thread_cbs virTLSThreadImpl = { virTLSMutexUnlock, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL }; -#endif +#endif /* WITH_GNUTLS_GCRYPT */ /* Helper macros to implement VIR_DOMAIN_DEBUG using just C99. This * assumes you pass fewer than 15 arguments to VIR_DOMAIN_DEBUG, but @@ -407,7 +409,7 @@ virGlobalInit(void) virErrorInitialize() < 0) goto error; -#ifdef WITH_GNUTLS +#ifdef WITH_GNUTLS_GCRYPT /* * This sequence of API calls it copied exactly from * gnutls 2.12.23 source lib/gcrypt/init.c, with -- 1.8.3.1

2 3

[libvirt] [PATCH 0/2] Split driver StateAutoStart from StateInitialization
by John Ferlan 26 Jul '13

26 Jul '13

The post push review/comments for the chap authentication determined that trying to connect to qemu driver from within the storage auto start would not be successful, see the following and followups https://www.redhat.com/archives/libvir-list/2013-July/msg01409.html These patches will split the virStateInitialize() into two phases - the first being the bulk of the existing code and the second being just running the auto start functionality for each of the drivers that needs/support it. I realize it's probably too late for 1.1.1, but figured it'd be good to be ready when the barn door opens again. John Ferlan (2): Separate out StateAutoStart from StateInitialize virStateDriver - Separate AutoStart from Initialize src/driver.h | 4 ++++ src/libvirt.c | 23 ++++++++++++++++++++++- src/libxl/libxl_driver.c | 18 +++++++++++++++--- src/lxc/lxc_driver.c | 18 ++++++++++++++++-- src/network/bridge_driver.c | 20 +++++++++++++++++++- src/qemu/qemu_driver.c | 18 ++++++++++++++++-- src/storage/storage_driver.c | 19 ++++++++++++++++++- src/uml/uml_driver.c | 18 ++++++++++++++++-- 8 files changed, 126 insertions(+), 12 deletions(-) -- 1.8.1.4

4 9