May 2009 - Devel - libvirt List Archives

[libvirt] [PATCH v2 0/4] Improve QEMU process startup error
by Cole Robinson 11 May '09

11 May '09

Error reporting for failed QEMU processes isn't too great at the moment. A couple cases I have hit: - Trying to start a VM with a cdrom pointing to a nonexistent file. The svirt hook function choked on this and indicated as much in the domain log, but I still had to wait for the pidfile timeout (10 seconds), and the only error reported was 'Domain did not show up'. - QEMU dies quickly at startup. Happened with the pulse audio issue, or if we build cli args incorrectly. The reporting here varies: if QEMU dies before writing out the pidfile, the above case applies. If the pidfile is written, we timeout in ReadLogOutput, and it's hit or miss whether we even report the logfile output to the user. The following patches improve the above cases by ensuring we never have to wait for a timeout, and that the logfile output is always reported in the error message. Changes since v1: Drop patch 1 + 2 (applied separately) Patch 7 is pretty isolated, so was resent to list separately. Don't waitpid in virExecDaemonize if virExec fails. Remove needless pidfile loop in qemudStartVMDaemon. Thanks, Cole Cole Robinson (4): Add helper function virExecDaemonize Report qemu log data if we fail to daemonize the process. Add pidfile argument to __virExec Check that QEMU is still alive while reading startup output. src/libvirt_private.syms | 2 +- src/proxy_internal.c | 16 +--- src/qemu_conf.c | 7 -- src/qemu_driver.c | 89 +++++++++++-------- src/remote_internal.c | 15 +--- src/uml_driver.c | 11 +-- src/util.c | 90 +++++++++++++++++--- src/util.h | 15 +++- .../qemuxml2argvdata/qemuxml2argv-boot-cdrom.args | 2 +- .../qemuxml2argvdata/qemuxml2argv-boot-floppy.args | 2 +- .../qemuxml2argv-boot-network.args | 2 +- .../qemuxml2argvdata/qemuxml2argv-bootloader.args | 2 +- .../qemuxml2argv-clock-localtime.args | 2 +- tests/qemuxml2argvdata/qemuxml2argv-clock-utc.args | 2 +- .../qemuxml2argv-console-compat.args | 2 +- .../qemuxml2argv-disk-cdrom-empty.args | 2 +- .../qemuxml2argvdata/qemuxml2argv-disk-cdrom.args | 2 +- .../qemuxml2argv-disk-drive-boot-cdrom.args | 2 +- .../qemuxml2argv-disk-drive-boot-disk.args | 2 +- .../qemuxml2argv-disk-drive-cache-v1-none.args | 2 +- .../qemuxml2argv-disk-drive-cache-v1-wb.args | 2 +- .../qemuxml2argv-disk-drive-cache-v1-wt.args | 2 +- .../qemuxml2argv-disk-drive-cache-v2-none.args | 2 +- .../qemuxml2argv-disk-drive-cache-v2-wb.args | 2 +- .../qemuxml2argv-disk-drive-cache-v2-wt.args | 2 +- .../qemuxml2argv-disk-drive-fmt-qcow.args | 2 +- .../qemuxml2argv-disk-drive-shared.args | 2 +- .../qemuxml2argvdata/qemuxml2argv-disk-floppy.args | 2 +- tests/qemuxml2argvdata/qemuxml2argv-disk-many.args | 2 +- tests/qemuxml2argvdata/qemuxml2argv-disk-usb.args | 2 +- .../qemuxml2argvdata/qemuxml2argv-disk-virtio.args | 2 +- .../qemuxml2argvdata/qemuxml2argv-disk-xenvbd.args | 2 +- .../qemuxml2argv-graphics-sdl-fullscreen.args | 2 +- .../qemuxml2argv-graphics-sdl.args | 2 +- .../qemuxml2argv-graphics-vnc-sasl.args | 2 +- .../qemuxml2argv-graphics-vnc-tls.args | 2 +- .../qemuxml2argv-graphics-vnc.args | 2 +- .../qemuxml2argv-hostdev-pci-address.args | 2 +- .../qemuxml2argv-hostdev-usb-address.args | 2 +- .../qemuxml2argv-hostdev-usb-product.args | 2 +- .../qemuxml2argv-input-usbmouse.args | 2 +- .../qemuxml2argv-input-usbtablet.args | 2 +- tests/qemuxml2argvdata/qemuxml2argv-input-xen.args | 2 +- tests/qemuxml2argvdata/qemuxml2argv-migrate.args | 2 +- tests/qemuxml2argvdata/qemuxml2argv-minimal.args | 2 +- tests/qemuxml2argvdata/qemuxml2argv-misc-acpi.args | 2 +- .../qemuxml2argv-misc-no-reboot.args | 2 +- tests/qemuxml2argvdata/qemuxml2argv-misc-uuid.args | 2 +- .../qemuxml2argv-net-eth-ifname.args | 2 +- tests/qemuxml2argvdata/qemuxml2argv-net-eth.args | 2 +- tests/qemuxml2argvdata/qemuxml2argv-net-user.args | 2 +- .../qemuxml2argvdata/qemuxml2argv-net-virtio.args | 2 +- .../qemuxml2argv-parallel-tcp.args | 2 +- .../qemuxml2argvdata/qemuxml2argv-restore-v1.args | 2 +- .../qemuxml2argvdata/qemuxml2argv-restore-v2.args | 2 +- .../qemuxml2argvdata/qemuxml2argv-serial-dev.args | 2 +- .../qemuxml2argvdata/qemuxml2argv-serial-file.args | 2 +- .../qemuxml2argvdata/qemuxml2argv-serial-many.args | 2 +- .../qemuxml2argvdata/qemuxml2argv-serial-pty.args | 2 +- .../qemuxml2argv-serial-tcp-telnet.args | 2 +- .../qemuxml2argvdata/qemuxml2argv-serial-tcp.args | 2 +- .../qemuxml2argvdata/qemuxml2argv-serial-udp.args | 2 +- .../qemuxml2argvdata/qemuxml2argv-serial-unix.args | 2 +- tests/qemuxml2argvdata/qemuxml2argv-serial-vc.args | 2 +- tests/qemuxml2argvdata/qemuxml2argv-sound.args | 2 +- 65 files changed, 211 insertions(+), 148 deletions(-)

2 9

[libvirt] [PATCH 2/3] support for "Host only" and "Internal" networks in VirtualBox driver
by Pritesh Kothari 11 May '09

11 May '09

Hi All, As discussed on the list resending the networking patch's. the patch's are as below: [PATCH 1/3]: contains support for "Internal" network in libvirt [PATCH 2/3]: contains support for "Host only" and "Internal" networks in VirtualBox driver [PATCH 3/3]: contains networking API for hostonly networks in VirtualBox driver in libvirt (it contains all the fix's proposed on list along with two extra *DefinedNetworks functions) The patches work as per cvs checkin today. (git SHA1 ID: 9136ae2056b45ea83854d3fe31d860f645b8c883) Regards, Pritesh

3 2

[libvirt] [PATCH 1/3] support for "Internal" network in libvirt
by Pritesh Kothari 11 May '09

11 May '09

Hi All, As discussed on the list resending the networking patch's. the patch's are as below: [PATCH 1/3]: contains support for "Internal" network in libvirt [PATCH 2/3]: contains support for "Host only" and "Internal" networks in VirtualBox driver [PATCH 3/3]: contains networking API for hostonly networks in VirtualBox driver in libvirt (it contains all the fix's proposed on list along with two extra *DefinedNetworks functions) The patches work as per cvs checkin today. (git SHA1 ID: 9136ae2056b45ea83854d3fe31d860f645b8c883) Regards, Pritesh

3 2

[libvirt] Status and plans for next release
by Daniel Veillard 11 May '09

11 May '09

Seems I didn't send any reminder about the next release plans on Friday, so here is a recap: - feature freeze on the 22nd so we have 2 more week to work on reviewing and adding OpenNebula/Power drivers and try to get the NPIV, netcf and secure migration patches in. It's likely not everything will make the release cut but we can try ! - target for next release is Friday 29th So far we have mostly a lot of bug fixes and VirtualBox driver updates commited since 0.6.3, Daniel -- Daniel Veillard | libxml Gnome XML XSLT toolkit http://xmlsoft.org/ daniel(a)veillard.com | Rpmfind RPM search engine http://rpmfind.net/ http://veillard.com/ | virtualization library http://libvirt.org/

1 0

[libvirt] Java binding blockStats bug (libvirt-java-0.2.1)
by Zvi Dubitzky 11 May '09

11 May '09

running blcokStats() with Java binding (libvirt-java-0.2.1 ) I get an exception Inspecting the src/jni/org_libvirt_Domain.c is see the following lines of code : JNIEXPORT jobject JNICALL Java_org_libvirt_Domain__1blockStats (JNIEnv *env, jobject obj, jlong VDP, jstring j_path){ struct _virDomainBlockStats stats; jobject j_stats; jclass stats_cls=(*env)->FindClass(env, "org/libvirt/DomainInterfaceStats"); const char *path = (*env)->GetStringUTFChars(env, j_path, NULL); if(virDomainBlockStats((virDomainPtr)VDP, path, &stats, sizeof(struct _virDomainBlockStats))<0){ (*env)->ReleaseStringUTFChars(env, j_path, path); return NULL; } . . . The bold line should be replaced by : jclass stats_cls=(*env)->FindClass(env, "org/libvirt/DomainBlockStats") as we deal here with blockStats and not network interface Stats ! thanks Zvi Dubitzky Virtualization and System Architecture Email:dubi@il.ibm.com IBM Haifa Research Laboratory Phone: +972-4-8296182 Haifa, 31905, ISRAEL

2 1

[libvirt] Fw: Java api for domain migration
by Zvi Dubitzky 10 May '09

10 May '09

Sorry , clarification: Bottom line is that the Java binding of libvirt should accept 'dname' parameter = null as the C libvirt case does without crashing The below stament that " If I try to specify a meaningful 'dname' then if it is the same name as the original host domain 'name' I will get a message saying : invalid domain pointer in no domain with matching id -1." is not correct. The migrate works under Java in this case as in the C case i.e the domain is migrated to the destination host but remains paused there for a long time until resumed there. If the 'dname' is null then the migration is faster but that works ok only with the libvirt C library. It crashes with the Java binding . Zvi Dubitzky Virtualization and System Architecture Email:dubi@il.ibm.com IBM Haifa Research Laboratory Phone: +972-4-8296182 Haifa, 31905, ISRAEL ----- Forwarded by Zvi Dubitzky/Haifa/IBM on 10/05/2009 13:29 ----- Zvi Dubitzky/Haifa/IBM 10/05/2009 13:14 To libvir-list(a)redhat.com cc Subject Fw: Java api for domain migration Hi I think I detected the origin of the libvirt Java binding crash in the case of domain migration though I do not know how to fix it yet . The fix should be in the src/jni/org_libvirt_Domain.c / migrate function. First: It turns out the that the C libvirt interface for migrate ( libvirt 0.6.1. + kvm-84) has a limitation with specifying the 'dname' parameter . It should be NULL to work ok , meaning the domain is migrated from the old host to the new host under the same domain name . If I try to specify a meaningful 'dname' then if it is the same name as the original host domain 'name' I will get a message saying : invalid domain pointer in no domain with matching id -1. If I am using 'dname' different than the original host domain 'name' it will complain that no such domain is found on the target host. So using NULL for 'dname' is the only one that makes the migrate work Using 'dname' = NULL with the Java binding causes it though to crash as below (see .log file). Fixing the jni code as below (see added 'if') does not help: src/jni/org_libvirt_Domain.c -------------------------------- JNIEXPORT jlong JNICALL Java_org_libvirt_Domain__1migrate (JNIEnv *env, jobject obj, jlong VDP, jobject dconn, jlong flags, jstring j_dname, jstring j_uri, jlong bandwidth){ virConnectPtr destVCP; const char *dname=NULL; const char *uri=NULL; //if String="", we pass NULL to the library if((*env)->GetStringLength(env, j_dname)>0) dname=(*env)->GetStringUTFChars(env, j_dname, NULL); //if String="", we pass NULL to the library if((*env)->GetStringLength(env, j_uri)>0) uri=(*env)->GetStringUTFChars(env, j_uri, NULL); //Extract the destination Conn Ptr destVCP=(virConnectPtr)(*env)->GetLongField(env, dconn, (*env)->GetFieldID(env, (*env)->GetObjectClass(env, dconn), "VCP", "J")); jlong retval = (jlong)virDomainMigrate((virDomainPtr)VDP, destVCP, flags, dname, uri, bandwidth) ; if (dname) (*env)->ReleaseStringUTFChars(env, j_dname, dname); if (uri) (*env)->ReleaseStringUTFChars(env, j_uri, uri); return retval; } Please your advice on how to solve the migrate for the Java binding thanks Zvi Dubitzky Virtualization and System Architecture Email:dubi@il.ibm.com IBM Haifa Research Laboratory Phone: +972-4-8296182 Haifa, 31905, ISRAEL ----- Forwarded by Zvi Dubitzky/Haifa/IBM on 10/05/2009 12:59 ----- Zvi Dubitzky/Haifa/IBM 07/05/2009 13:49 To libvir-list(a)redhat.com cc Subject Java api for libvirt migration I am trying to do a migrate of VM with libvirt 0.6.1 and kvm-84 . I am successful in doing it many times via the C interface . I can do most of the basic libvirt activity via the libvirt JAVA library (libvirt-java-0.2.1). When it comes to a migrate the JVM fails wit a log file attached below . Can someone help. My memory consumption in my JAVA test code is small. The migrate parameters I am sending with the Java are identical to those I use in the C case and is: destDomain = localDomain.migrate (dcon, // destination host Connect object VIR_MIGRATE_LIVE, null, //const char *dname destination domain name target_uri, //"tcp:9.148.28.13:4444" 0 // unsigned long bandwidth) - do not specify ); thanks Zvi Dubitzky Virtualization and System Architecture Email:dubi@il.ibm.com IBM Haifa Research Laboratory Phone: +972-4-8296182 Haifa, 31905, ISRAEL

1 0

[libvirt] Fw: Java api for domain migration
by Zvi Dubitzky 10 May '09

10 May '09

Hi I think I detected the origin of the libvirt Java binding crash in the case of domain migration though I do not know how to fix it yet . The fix should be in the src/jni/org_libvirt_Domain.c / migrate function. First: It turns out the that the C libvirt interface for migrate ( libvirt 0.6.1. + kvm-84) has a limitation with specifying the 'dname' parameter . It should be NULL to work ok , meaning the domain is migrated from the old host to the new host under the same domain name . If I try to specify a meaningful 'dname' then if it is the same name as the original host domain 'name' I will get a message saying : invalid domain pointer in no domain with matching id -1. If I am using 'dname' different than the original host domain 'name' it will complain that no such domain is found on the target host. So using NULL for 'dname' is the only one that makes the migrate work Using 'dname' = NULL with the Java binding causes it though to crash as below (see .log file). Fixing the jni code as below (see added 'if') does not help: src/jni/org_libvirt_Domain.c -------------------------------- JNIEXPORT jlong JNICALL Java_org_libvirt_Domain__1migrate (JNIEnv *env, jobject obj, jlong VDP, jobject dconn, jlong flags, jstring j_dname, jstring j_uri, jlong bandwidth){ virConnectPtr destVCP; const char *dname=NULL; const char *uri=NULL; //if String="", we pass NULL to the library if((*env)->GetStringLength(env, j_dname)>0) dname=(*env)->GetStringUTFChars(env, j_dname, NULL); //if String="", we pass NULL to the library if((*env)->GetStringLength(env, j_uri)>0) uri=(*env)->GetStringUTFChars(env, j_uri, NULL); //Extract the destination Conn Ptr destVCP=(virConnectPtr)(*env)->GetLongField(env, dconn, (*env)->GetFieldID(env, (*env)->GetObjectClass(env, dconn), "VCP", "J")); jlong retval = (jlong)virDomainMigrate((virDomainPtr)VDP, destVCP, flags, dname, uri, bandwidth) ; if (dname) (*env)->ReleaseStringUTFChars(env, j_dname, dname); if (uri) (*env)->ReleaseStringUTFChars(env, j_uri, uri); return retval; } Please your advice on how to solve the migrate for the Java binding thanks Zvi Dubitzky Virtualization and System Architecture Email:dubi@il.ibm.com IBM Haifa Research Laboratory Phone: +972-4-8296182 Haifa, 31905, ISRAEL ----- Forwarded by Zvi Dubitzky/Haifa/IBM on 10/05/2009 12:59 ----- Zvi Dubitzky/Haifa/IBM 07/05/2009 13:49 To libvir-list(a)redhat.com cc Subject Java api for libvirt migration I am trying to do a migrate of VM with libvirt 0.6.1 and kvm-84 . I am successful in doing it many times via the C interface . I can do most of the basic libvirt activity via the libvirt JAVA library (libvirt-java-0.2.1). When it comes to a migrate the JVM fails wit a log file attached below . Can someone help. My memory consumption in my JAVA test code is small. The migrate parameters I am sending with the Java are identical to those I use in the C case and is: destDomain = localDomain.migrate (dcon, // destination host Connect object VIR_MIGRATE_LIVE, null, //const char *dname destination domain name target_uri, //"tcp:9.148.28.13:4444" 0 // unsigned long bandwidth) - do not specify ); thanks Zvi Dubitzky Virtualization and System Architecture Email:dubi@il.ibm.com IBM Haifa Research Laboratory Phone: +972-4-8296182 Haifa, 31905, ISRAEL

1 0

[libvirt] migration fails
by jens 09 May '09

09 May '09

Hello, I tried to migrate a kvm domain to a second identical machine with the command: virsh migrate --live dom1 qemu+ssh://second.machine.org/system which leads me to an unresponsive virsh on the second machine - 'virsh list' just takes forever and qemu wasn't started (the domain was not migrated), finally I have to kill libvirtd on the second machine. Because I'm new to libvirt I searched the web to find a solution for this behavior but without success. Please can someone on the list point me to the right direction how I can use libvirt to migrate kvm domains? I use kernel 2.6.29.1 and kvm-85 with libvirt 0.6.2 on an debian unstable system. Attached are debug logs generated by the source- and destination-machine. Thank you very much Jens

5 7

[libvirt] [PATCH] change permissions of directories in cgroups
by Ryota Ozaki 08 May '09

08 May '09

Hi, This patch creates a directory in cgroups with an ordinary permission 0755 (rwxr-xr-x) instead of 0655 (rw-r-xr-x). I guess 0655 is not expected and just a mistake, or is there a special reason? Thanks, ozaki-r Signed-off-by: Ryota Ozaki <ozaki.ryota(a)gmail.com> >From 4829855bc0baa3c75806f106bc0e54eb1da75eea Mon Sep 17 00:00:00 2001 From: Ryota Ozaki <ozaki.ryota(a)gmail.com> Date: Fri, 8 May 2009 03:23:50 +0900 Subject: [PATCH] change permissions of directories in cgroups This patch creates a directory in cgroups with an ordinary permission 0755 (rwxr-xr-x) instead of 0655 (rw-r-xr-x). --- src/cgroup.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/src/cgroup.c b/src/cgroup.c index d1d44a2..50517e2 100644 --- a/src/cgroup.c +++ b/src/cgroup.c @@ -436,7 +436,7 @@ static int virCgroupMakeGroup(const char *name) virCgroupFree(&root); if (access(path, F_OK) != 0) { - if (mkdir(path, 0655) < 0) { + if (mkdir(path, 0755) < 0) { rc = -errno; VIR_FREE(path); break; -- 1.6.0.6

3 3

[libvirt] [PATCH] lxc: fix mknod file type
by Ryota Ozaki 08 May '09

08 May '09

Hi, This patch fixes the 2nd argument of mknod syscall. The argument should include a file type, i.e., S_IFCHR in this case, otherwise created files will be regular files. Thanks, ozaki-r Signed-off-by: Ryota Ozaki <ozaki.ryota(a)gmail.com> >From 60009b725da855f131e8272f0397455ae27bafc3 Mon Sep 17 00:00:00 2001 From: Ryota Ozaki <ozaki.ryota(a)gmail.com> Date: Wed, 6 May 2009 03:47:04 +0900 Subject: [PATCH] lxc: fix mknod file type This patch fixes the 2nd argument of mknod syscall. The argument should include a file type, i.e., S_IFCHR in this case, otherwise created files will be regular files. --- src/lxc_container.c | 4 ++-- 1 files changed, 2 insertions(+), 2 deletions(-) diff --git a/src/lxc_container.c b/src/lxc_container.c index 3946b84..c30daec 100644 --- a/src/lxc_container.c +++ b/src/lxc_container.c @@ -440,7 +440,7 @@ static int lxcContainerPopulateDevices(void) /* Populate /dev/ with a few important bits */ for (i = 0 ; i < ARRAY_CARDINALITY(devs) ; i++) { dev_t dev = makedev(devs[i].maj, devs[i].min); - if (mknod(devs[i].path, 0, dev) < 0 || + if (mknod(devs[i].path, S_IFCHR, dev) < 0 || chmod(devs[i].path, devs[i].mode)) { virReportSystemError(NULL, errno, _("failed to make device %s"), @@ -457,7 +457,7 @@ static int lxcContainerPopulateDevices(void) } } else { dev_t dev = makedev(LXC_DEV_MAJ_TTY, LXC_DEV_MIN_PTMX); - if (mknod("/dev/ptmx", 0, dev) < 0 || + if (mknod("/dev/ptmx", S_IFCHR, dev) < 0 || chmod("/dev/ptmx", 0666)) { virReportSystemError(NULL, errno, "%s", _("failed to make device /dev/ptmx")); -- 1.6.0.6

2 4