September 2009 - Devel - libvirt List Archives

[libvirt] [PATCH] Fix a memory leak in virsh.
by Laine Stump 04 Sep '09

04 Sep '09

I found this while looking for examples of using virNodeDeviceGetXMLDesc(). AFAIK, *all* of the *GetXMLDesc() functions return a newly allocated chunk of memory that is owned by the caller, who must free it when they're done... --- src/virsh.c | 10 +++++++++- 1 files changed, 9 insertions(+), 1 deletions(-) diff --git a/src/virsh.c b/src/virsh.c index 15e0cef..910d860 100644 --- a/src/virsh.c +++ b/src/virsh.c @@ -5521,6 +5521,7 @@ cmdNodeDeviceDumpXML (vshControl *ctl, const vshCmd *cmd) { const char *name; virNodeDevicePtr device; + char *xml; if (!vshConnectionUsability(ctl, ctl->conn, TRUE)) return FALSE; @@ -5531,7 +5532,14 @@ cmdNodeDeviceDumpXML (vshControl *ctl, const vshCmd *cmd) return FALSE; } - vshPrint(ctl, "%s\n", virNodeDeviceGetXMLDesc(device, 0)); + xml = virNodeDeviceGetXMLDesc(device, 0); + if (!xml) { + virNodeDeviceFree(device); + return FALSE; + } + + vshPrint(ctl, "%s\n", xml); + free(xml); virNodeDeviceFree(device); return TRUE; } -- 1.6.2.5

3 2

[libvirt] [PATCH] Fix sexpr2string() to handle empty list
by Jim Fehlig 04 Sep '09

04 Sep '09

Finally have some time to submit this small patch for a bug I'm seeing with Xen 3.3.1. If s-expression returned by xend contains an empty list, sexpr2string() failed to serialize it. E.g. sexpr containing (cpus (()())) would cause sexpr2string() to fail. I spoke with danpb on IRC about this bug and he suggested adding a XML/sexpr pair that exhibits the problem to tests/xml2sexprtest.c. I can easily produce the sexpr but not the case with XML. The empty cpus lists indicate no explicit affinity and AFAIK the only way to express this in XML is absence of cpuset attribute in vcpu element. But IMO we should not produce the (cpus (()()...)) sexpr in this case. It should only be produced when user has explicitly specified affinity. Does this sound reasonable? Regards, Jim

2 1

[libvirt] [PATCH] Fix ID field in virDomainPtr after starting Xen guest
by Daniel P. Berrange 04 Sep '09

04 Sep '09

* src/xend_internal.c: Lookup guest after starting persistent config in order to refresh the domain ID field --- src/xend_internal.c | 14 +++++++++++++- 1 files changed, 13 insertions(+), 1 deletions(-) diff --git a/src/xend_internal.c b/src/xend_internal.c index 7bcee7d..459b6bc 100644 --- a/src/xend_internal.c +++ b/src/xend_internal.c @@ -4537,6 +4537,8 @@ virDomainPtr xenDaemonDomainDefineXML(virConnectPtr conn, const char *xmlDesc) { int xenDaemonDomainCreate(virDomainPtr domain) { xenUnifiedPrivatePtr priv; + int ret; + virDomainPtr tmp; if ((domain == NULL) || (domain->conn == NULL) || (domain->name == NULL)) { virXendError((domain ? domain->conn : NULL), VIR_ERR_INVALID_ARG, @@ -4549,7 +4551,17 @@ int xenDaemonDomainCreate(virDomainPtr domain) if (priv->xendConfigVersion < 3) return(-1); - return xend_op(domain->conn, domain->name, "op", "start", NULL); + ret = xend_op(domain->conn, domain->name, "op", "start", NULL); + + if (ret != -1) { + /* Need to force a refresh of this object's ID */ + tmp = virDomainLookupByName(domain->conn, domain->name); + if (tmp) { + domain->id = tmp->id; + virDomainFree(tmp); + } + } + return ret; } int xenDaemonDomainUndefine(virDomainPtr domain) -- 1.6.4

2 1

[libvirt] Save & Restore Synchronization
by Erkan Unal 04 Sep '09

04 Sep '09

Hi, I am using libvirt for qemu/kvm. I have following questions in terms of save and restore (Version 0.6.2): 1) As I checked the libvirt code, restore command is asynchronous. There is a macro called VIR_EXEC_NONBLOCK but there is no macro called VIR_EXEC_BLOCK. Is it possible to execute the restore in blocking mode so that I can measure the whole restore time? 2) I tried to measure the time spent to save the VM. I am executing save command and getting a timing. However when I execute the following operations in order I got an error right after script issues the restore operation (No delay between them): "save->restore->save" ... the error is: 20:19:02.478: error : internal error Unable to open monitor path /dev/pts/3 libvir: QEMU error : internal error Unable to open monitor path /dev/pts/3 20:19:02.478: error : internal error unable to start guest: char device redirected to /dev/pts/3 inet_listen: bind(ipv4,127.0.0.1,5912): Address already in use inet_listen: FAILED libvir: QEMU error : internal error unable to start guest: char device redirected to /dev/pts/3 inet_listen: bind(ipv4,127.0.0.1,5912): Address already in use inet_listen: FAILED 20:19:02.482: error : operation failed: failed to start VM libvir: QEMU error : operation failed: failed to start VM error: Failed to restore domain from /path/to/the/save/file error: operation failed: failed to start VM (You can find the detailed error report in the attachement when LIBVIRT_DEBUG flag is 1) I think that some of the cleanup is non-blocking which are cleanup of pty device and vncserver. However, I am not sure if destroying the qemu process is asynchronous?? P.S.: If I put one second delay between first save and restore, there is no error. Save and restore work fine. Thanks, Erkan Unal 20:26:07.506: debug : virInitialize:287 : register drivers 20:26:07.506: debug : virRegisterDriver:660 : registering Test as driver 0 20:26:07.506: debug : virRegisterNetworkDriver:560 : registering Test as network driver 0 20:26:07.506: debug : virRegisterStorageDriver:591 : registering Test as storage driver 0 20:26:07.506: debug : virRegisterDeviceMonitor:622 : registering Test as device driver 0 20:26:07.506: debug : virRegisterDriver:660 : registering Xen as driver 1 20:26:07.506: debug : virRegisterDriver:660 : registering OPENVZ as driver 2 20:26:07.506: debug : virRegisterDriver:660 : registering remote as driver 3 20:26:07.506: debug : virRegisterNetworkDriver:560 : registering remote as network driver 1 20:26:07.506: debug : virRegisterStorageDriver:591 : registering remote as storage driver 1 20:26:07.506: debug : virRegisterDeviceMonitor:622 : registering remote as device driver 1 20:26:07.506: debug : virConnectOpenAuth:1089 : name=qemu:///session, auth=0x7fa21cbae660, flags=0 20:26:07.506: debug : do_open:909 : name "qemu:///session" to URI components: scheme qemu opaque (null) authority (null) server (null) user (null) port 0 path /session 20:26:07.506: debug : do_open:919 : trying driver 0 (Test) ... 20:26:07.506: debug : do_open:925 : driver 0 Test returned DECLINED 20:26:07.506: debug : do_open:919 : trying driver 1 (Xen) ... 20:26:07.506: debug : do_open:925 : driver 1 Xen returned DECLINED 20:26:07.506: debug : do_open:919 : trying driver 2 (OPENVZ) ... 20:26:07.506: debug : do_open:925 : driver 2 OPENVZ returned DECLINED 20:26:07.506: debug : do_open:919 : trying driver 3 (remote) ... 20:26:07.506: debug : remoteOpen:974 : Auto-remote UNIX socket 20:26:07.506: debug : remoteOpen:992 : Auto-spawn user daemon instance 20:26:07.506: debug : doRemoteOpen:511 : proceeding with name = qemu:///session 20:26:07.507: debug : call:6473 : Doing call 66 (nil) 20:26:07.507: debug : call:6543 : We have the buck 66 0x7fa21bdb5010 0x7fa21bdb5010 20:26:07.507: debug : processCallRecvLen:6131 : Got length, now need 36 total (32 more) 20:26:07.507: debug : processCalls:6399 : Giving up the buck 66 0x7fa21bdb5010 (nil) 20:26:07.507: debug : call:6574 : All done with our call 66 (nil) 0x7fa21bdb5010 20:26:07.507: debug : call:6473 : Doing call 1 (nil) 20:26:07.507: debug : call:6543 : We have the buck 1 0xf86d80 0xf86d80 20:26:07.509: debug : processCallRecvLen:6131 : Got length, now need 28 total (24 more) 20:26:07.509: debug : processCalls:6399 : Giving up the buck 1 0xf86d80 (nil) 20:26:07.509: debug : call:6574 : All done with our call 1 (nil) 0xf86d80 20:26:07.509: debug : doRemoteOpen:822 : Adding Handler for remote events 20:26:07.509: debug : doRemoteOpen:829 : virEventAddHandle failed: No addHandleImpl defined. continuing without events. 20:26:07.509: debug : do_open:925 : driver 3 remote returned SUCCESS 20:26:07.509: debug : do_open:945 : network driver 0 Test returned DECLINED 20:26:07.509: debug : do_open:945 : network driver 1 remote returned SUCCESS 20:26:07.509: debug : do_open:967 : storage driver 0 Test returned DECLINED 20:26:07.509: debug : do_open:967 : storage driver 1 remote returned SUCCESS 20:26:07.510: debug : do_open:988 : node driver 0 Test returned DECLINED 20:26:07.510: debug : do_open:988 : node driver 1 remote returned SUCCESS 20:26:07.510: debug : virDomainRestore:2012 : conn=0xf82430, from=/path/to/save/file 20:26:07.510: debug : call:6473 : Doing call 54 (nil) 20:26:07.510: debug : call:6543 : We have the buck 54 0xf86d80 0xf86d80 20:26:07.625: error : internal error Unable to open monitor path /dev/pts/3 libvir: QEMU error : internal error Unable to open monitor path /dev/pts/3 20:26:07.625: error : internal error unable to start guest: char device redirected to /dev/pts/3 inet_listen: bind(ipv4,127.0.0.1,5912): Address already in use inet_listen: FAILED libvir: QEMU error : internal error unable to start guest: char device redirected to /dev/pts/3 inet_listen: bind(ipv4,127.0.0.1,5912): Address already in use inet_listen: FAILED 20:26:07.629: error : operation failed: failed to start VM libvir: QEMU error : operation failed: failed to start VM 20:26:07.629: debug : processCallRecvLen:6131 : Got length, now need 160 total (156 more) 20:26:07.629: debug : processCalls:6399 : Giving up the buck 54 0xf86d80 (nil) 20:26:07.629: debug : call:6574 : All done with our call 54 (nil) 0xf86d80 error: Failed to restore domain from /path/to/save/file error: operation failed: failed to start VM 20:26:07.629: debug : virConnectClose:1107 : conn=0xf82430 20:26:07.629: debug : call:6473 : Doing call 2 (nil) 20:26:07.629: debug : call:6543 : We have the buck 2 0xf86d80 0xf86d80 20:26:07.629: debug : processCallRecvLen:6131 : Got length, now need 28 total (24 more) 20:26:07.629: debug : processCalls:6399 : Giving up the buck 2 0xf86d80 (nil) 20:26:07.629: debug : call:6574 : All done with our call 2 (nil) 0xf86d80 20:26:07.629: debug : virUnrefConnect:210 : unref connection 0xf82430 1 20:26:07.629: debug : virReleaseConnect:171 : release connection 0xf82430

2 1

[libvirt] libvirt socket closed unexpectedly / libvirt broken pipe
by Dave 04 Sep '09

04 Sep '09

Hello, I'm trying to get eucalyptus v1.5.2 working with debian lenny, the relevant versions of libvirt software installed are from the apt-get repos, which are: # dpkg -l|grep libvirt ii libvirt-bin 0.4.6-10 the programs for the libvirt library ii libvirt0 0.4.6-10 library for interfacing with different virtualization systems The the conf file looks like so: # cat libvirtd.conf|grep -v '#' unix_sock_group = "libvirt" unix_sock_rw_perms = "0770" auth_unix_ro = "none" auth_unix_rw = "none" ----- my eucalyptus nc.log has the following error, which i believe is the issue, basically any instance i start, terminates about a minute later by itself, any guidance or help is much appreciated [Sat Aug 15 03:20:41 2009][004602][EUCAERROR ] libvirt: socket closed unexpectedly (code=39) [Sat Aug 15 03:20:43 2009][004602][EUCAERROR ] libvirt: Broken pipe (code=38) thoughts? ideas?

2 1

[libvirt] [PATCH] Allow libvirtd to RPC to external libvirtd.
by Chris Lalancette 04 Sep '09

04 Sep '09

Allow the daemon itself to make RPCs to an external libvirtd, but only if the URI is fully specified. While this isn't used at the moment, it will be for the tunnelled migration support in the future. Signed-off-by: Chris Lalancette <clalance(a)redhat.com> --- src/remote_internal.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/src/remote_internal.c b/src/remote_internal.c index bc69c70..a416cf3 100644 --- a/src/remote_internal.c +++ b/src/remote_internal.c @@ -1001,7 +1001,7 @@ remoteOpen (virConnectPtr conn, int ret, rflags = 0; const char *autostart = getenv("LIBVIRT_AUTOSTART"); - if (inside_daemon) + if (inside_daemon && (!conn->uri || (conn->uri && !conn->uri->server))) return VIR_DRV_OPEN_DECLINED; if (!(priv = remoteAllocPrivateData(conn))) -- 1.6.0.6

2 3

[libvirt] On pidfile security and using -daemonize to start qemu/kvm
by Charles Duffy 03 Sep '09

03 Sep '09

I had a brief conversation with Daniel Berrange on IRC this morning regarding race conditions in libvirt's invocation of qemu domains. The preferred approach from qemu's maintainership is using -daemonize at invocation, and then waiting for the process to exit before attempting to connect to the monitor; after this occurs, qemu guarantees monitor availability, so no retry/timeout behavior is necessary. Dan pointed out an issue with this approach: To determine the PID of the qemu instance started, the -pidfile argument to qemu needs to be used with -daemonize. However, -pidfile writes the requested pidfile with qemu's permissions; if the guest were subverted, it could rewrite its pidfile and lead libvirt to kill other, arbitrary processes. Neither changing permissions with -runas rather than prior to invocation nor chown'ing the pidfile from libvirt after creation helps, as qemu keeps a file handle on it open throughout operation (to maintain a POSIX lock on the file). However, a workaround exists, and it meshes well with a security fix which needs to be implemented anyhow: At present, on initial startup or libvirtd restart, we read the PIDs associated with vCPUs from the monitor in virDetectVcpuPIDs(); doing this on libvirtd restart poses a security risk for the reason given above, so we need to have libvirtd record the PIDs on initial startup (before any untrusted code within the guest is ever run) regardless. If we also read the libvirtd pidfile at this same early time (before the guest has been started and thus had an opportunity to run untrusted code) and include its contents when writing out the vCPU PIDs, this would eliminate the issue with pidfile ownership blocking use of -daemonize. Thoughts?

2 1

[libvirt] [PATCH] xen_internal.c: remove two unused local variables
by Jim Meyering 03 Sep '09

03 Sep '09

Obviously unnecessary. Maybe someone had plans for these long ago? >From 94730ec1d99076cd18b04b950adad7c7a576071f Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Thu, 3 Sep 2009 17:11:14 +0200 Subject: [PATCH] xen_internal.c: remove two unused local variables * src/xen_internal.c (xenHypervisorSetSchedulerParameters): Remove set-but-never-used local variables, cap_set and weight_set. --- src/xen_internal.c | 5 ----- 1 files changed, 0 insertions(+), 5 deletions(-) diff --git a/src/xen_internal.c b/src/xen_internal.c index ae78f84..30fd8bd 100644 --- a/src/xen_internal.c +++ b/src/xen_internal.c @@ -1299,9 +1299,6 @@ xenHypervisorSetSchedulerParameters(virDomainPtr domain, TODO return(-1); case XEN_SCHEDULER_CREDIT: { - int weight_set = 0; - int cap_set = 0; - memset(&op_dom, 0, sizeof(op_dom)); op_dom.cmd = XEN_V2_OP_SCHEDULER; op_dom.domain = (domid_t) domain->id; @@ -1326,7 +1323,6 @@ xenHypervisorSetSchedulerParameters(virDomainPtr domain, return(-1); } op_dom.u.getschedinfo.u.credit.weight = val; - weight_set = 1; } else if (STREQ (params[i].field, str_cap) && params[i].type == VIR_DOMAIN_SCHED_FIELD_UINT) { val = params[i].value.ui; @@ -1336,7 +1332,6 @@ xenHypervisorSetSchedulerParameters(virDomainPtr domain, return(-1); } op_dom.u.getschedinfo.u.credit.cap = val; - cap_set = 1; } else { virXenErrorFunc (domain->conn, VIR_ERR_INVALID_ARG, __FUNCTION__, "Credit scheduler accepts 'cap' and 'weight' integer parameters", -- 1.6.4.2.395.ge3d52

2 1

[libvirt] [PATCH] python: let libvirt_virConnectDomainEventCallback indicate success
by Jim Meyering 03 Sep '09

03 Sep '09

First dead-store-related bug fix: >From 6024fb05dfa2dd408ad79abbac88791f5048af11 Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Thu, 3 Sep 2009 14:09:49 +0200 Subject: [PATCH] python: let libvirt_virConnectDomainEventCallback indicate success * python/libvir.c (libvirt_virConnectDomainEventCallback): Return 0 when successful, rather than always returning -1. clang flagged this function for its dead-store of "ret=0". Once "ret" was set to 0, it was never used, and the function would always return -1. --- python/libvir.c | 4 ++-- 1 files changed, 2 insertions(+), 2 deletions(-) diff --git a/python/libvir.c b/python/libvir.c index e2c196a..0c00b80 100644 --- a/python/libvir.c +++ b/python/libvir.c @@ -4,7 +4,7 @@ * entry points where an automatically generated stub is * unpractical * - * Copyright (C) 2005, 2007, 2008 Red Hat, Inc. + * Copyright (C) 2005, 2007-2009 Red Hat, Inc. * * Daniel Veillard <veillard(a)redhat.com> */ @@ -1812,7 +1812,7 @@ libvirt_virConnectDomainEventCallback(virConnectPtr conn ATTRIBUTE_UNUSED, cleanup: LIBVIRT_RELEASE_THREAD_STATE; - return -1; + return ret; } static PyObject * -- 1.6.4.2.395.ge3d52

2 1

[libvirt] [PATCH 0/3] QEMU directory/monitor socket fixes/cleanup
by Daniel P. Berrange 03 Sep '09

03 Sep '09

This patch series does three things - Moves QEMU monitor sockets in /var/lib/libvirt/qemu, since this is created file and thus should be separate from libvirtd created data to avoid risk from QEMU compromise. - Sets QEMU directory ownership to match configured qemu user/group - Fixes a minor memory leak / socket unlink

2 4