- Devel - libvirt List Archives

[libvirt] Some close/fclose/closedir calls are missing
by Matthias Bolte 07 Sep '09

07 Sep '09

While debugging something that looked like an file descriptor leak, we (Maximilian Wilhelm and me) found some places where close/fclose/closedir calls are missing, mostly in error handling if-branches. See attached patch. Max and Matthias

3 2

[libvirt] [PATCH 0/6] sVirt AppArmor security driver
by Jamie Strandboge 07 Sep '09

07 Sep '09

This patch series implements the AppArmor security driver for sVirt. This implementation was developed for the Ubuntu AppArmorLibvirtProfile specification[1], but is general enough for any AppArmor deployment (such as Ubuntu, *SUSE and Mandriva). This patch has seen quite a bit of real world testing in Ubuntu 9.10 (our development release) in our 0.7.0-1ubuntu3 package. I did make a few small changes after going through HACKING, but mostly I got the tests going and added documentation. DESIGN ------ When a virtual machine is started, determine if a profile is currently defined for the machine, and use it if available. If not, generate a new profile for the machine based on a template, which is by default a very restrictive profile allowing access to disk files, and anything else needed to run, such as the pid, monitor and log files. Virtual machines should have a unique profile specific to that machine. To ensure uniqueness, the profile name will be derived from the UUID of the virtual machine. These profiles should be configurable, either by adjusting the profile template for new machines, creating/modifying the VM profile directly or through the use of AppArmor abstractions. This will allow for administrators to fine-tune confinement for individual machines if desired. If enabled at compile time, the sVirt security model will be activated if AppArmor is available on the host OS and a profile for the libvirtd daemon is loaded when libvirtd is started. libvirtd should not be allowed to create arbitrary profiles or modify profiles directly, so as to not allow libvirtd to potentially (ie via a security bug in libvirtd itself) bootstrap out of AppArmor confinement. Because root privileges are needed to manipulate AppArmor profiles, qemu:///session will not be supported at this time, but the implementation must allow for a confined libvirtd with qemu:///session guests running unconfined. This can be revisited when AppArmor supports per-user profiles. Please see the specification[1] for more details. PATCHES ------- The patches are all against trunk as of yesterday. Testing was done on trunk and there seem to be no regressions over the the 0.7.0-1ubuntu3 package in Ubuntu. [PATCH 1*] patch_1a_reenable-nonfile-labels.patch: When James Morris originally submitted his sVirt patches (as seen in libvirt 0.6.1), he did not require on disk labelling for virSecurityDomainRestoreImageLabel. A later commit[2] changed this behavior to assume on disk labelling, which halts implementations for path-based MAC systems such as AppArmor and TOMOYO where vm->def->seclabel is required to obtain the label. This patch simply adds the 'virDomainObjPtr vm' argument back to *RestoreImageLabel. patch_1b_optional.patch: Due to the above change, 'make syntax-check' fails because SELinuxRestoreSecurityImageLabel() does not use the 'virDomainObjPtr vm'. patch_1b_optional.patch is a simple patch to fix this by checking if vm->def->seclabel == NULL and returns with error if it does. I realize this may not be desired in the long term, but it should be harmless enough to include. [PATCH 2] patch_2_security_c.patch: Updates src/security.c for AppArmor [PATCH 3] patch_3_security_apparmor.patch: Adds security_apparmor.c, security_apparmor.h, virt-aa-helper.c and updates po/POTFILES.in. virt-aa-helper.c is a new binary which is used exclusively by the AppArmor security driver to manipulate AppArmor. These files compile without warning and pass syntax-check. [PATCH 4] patch_4_tests.patch: Adds tests for virt-aa-helper and the security driver. secaatest.c is identical to seclabeltest.c except it initializes the 'apparmor' driver instead of 'selinux'. These tests are integrated into 'make check' and pass. [PATCH 5] patch_5_docs.patch: Updates docs/drvqemu.html.in for AppArmor and adds profile examples to examples/apparmor. [PATCH 6] patch_6_autoconf.patch: Updates Makefile.am and configure.in for AppArmor. It is based on and should operate the same as the SELinux configuration. Caveats and known issues: 1. it does not take advantage of the recent host device labelling functionality yet 2. it does not properly handle hot-plugging of devices yet 3. qemu:///session runs unconfined (see above) Thanks! Jamie (jdstrand on Freenode and OFTC) [1] https://wiki.ubuntu.com/SecurityTeam/Specifications/AppArmorLibvirtProfile [2] http://libvirt.org/git/?p=libvirt.git;a=commit;h=c86afc85ee0d1ec6d76c2d254b… -- Jamie Strandboge | http://www.canonical.com

3 10

[libvirt] [PATCH 0/6] Automatic sVirt management of host device labelling
by Daniel P. Berrange 07 Sep '09

07 Sep '09

This patch series extends the libvirt security driver API, and sVirt implementation to cover management of host device labelling. Previously users would have to set a global boolean tunable virt_use_pci/usb to allow all domains access to all host devices. With this series applied libvirt will automatically relabel only the individual PCI/USB devices which are assigned to a guest. ie it should make host device assignment 'just work' when sVirt is enforcing, and improve security It also attempts to address a problem with restoration of disk labels. The current code uses matchpathcon() to find the defalt label for a path. This works fine for locations which have a defined label in the policy (eg like /var/lib/libvirt/images), but if storing disk images in non-defualt locations, eg a external USB drive mounted under a place like /media/myusbdisk/virtual-images/, matchpathcon() returns NULL. In this scenario the disk would remain labelled with the MCS level specific to the just stopped VM. Since MCS labels are allocated on demand on each boot, this could allow a future VMs to access disks that it ought not to be able to. Dan Walsh suggested that we default to using the label defined for matchpathcon("/var/libvirt/images/00-DEFAULT") in this case, but this doesn't work for restoring USB/PCI device labels[1]. In all the case I've had this problem so far, the files' original label matched that of the directory it was contained in, so this patch just uses the containing directory's label when restoring labels. Dan didn't like this idea when I first mentioned it in IRC though, so perhaps I need to implement different logic still... ? Regards, Daniel [1] PCI device access from VMs requires labelling /sys/bus/pci/devices/$DOMAIN:$BUS:$SLOT:FUNCTION/{config, resource*, rom} while USB device access requires labelling /dev/bus/usb/$BUS/$DEVICE

5 18

[libvirt] [PATCH] Fix several memory leaks
by Ryota Ozaki 05 Sep '09

05 Sep '09

Hi, This patch fixes several memory leaks being scattered over libvirt. Thanks, ozaki-r >From 6fc282eae5192cccda208d8d4fd14c5c0676992b Mon Sep 17 00:00:00 2001 From: Ryota Ozaki <ozaki.ryota(a)gmail.com> Date: Tue, 1 Sep 2009 01:34:40 +0900 Subject: [PATCH] Fix several memory leaks --- qemud/qemud.c | 2 +- src/domain_conf.c | 1 + src/network_conf.c | 5 +++-- src/qemu_conf.c | 13 ++++++++----- src/storage_backend_fs.c | 5 +++++ 5 files changed, 18 insertions(+), 8 deletions(-) diff --git a/qemud/qemud.c b/qemud/qemud.c index df275e6..17ba44a 100644 --- a/qemud/qemud.c +++ b/qemud/qemud.c @@ -1733,7 +1733,7 @@ readmore: /* Possibly need to create another receive buffer */ if ((client->nrequests < max_client_requests && - VIR_ALLOC(client->rx) < 0)) { + !client->rx && VIR_ALLOC(client->rx) < 0)) { qemudDispatchClientFailure(client); } else { if (client->rx) diff --git a/src/domain_conf.c b/src/domain_conf.c index 1d2cc7c..4b64219 100644 --- a/src/domain_conf.c +++ b/src/domain_conf.c @@ -4361,6 +4361,7 @@ int virDomainSaveXML(virConnectPtr conn, cleanup: if (fd != -1) close(fd); + VIR_FREE(configFile); return ret; } diff --git a/src/network_conf.c b/src/network_conf.c index bb649a4..58a4f32 100644 --- a/src/network_conf.c +++ b/src/network_conf.c @@ -820,6 +820,7 @@ int virNetworkDeleteConfig(virConnectPtr conn, { char *configFile = NULL; char *autostartLink = NULL; + int ret = -1; if ((configFile = virNetworkConfigFile(conn, configDir, net->def->name)) == NULL) goto error; @@ -836,12 +837,12 @@ int virNetworkDeleteConfig(virConnectPtr conn, goto error; } - return 0; + ret = 0; error: VIR_FREE(configFile); VIR_FREE(autostartLink); - return -1; + return ret; } char *virNetworkConfigFile(virConnectPtr conn, diff --git a/src/qemu_conf.c b/src/qemu_conf.c index 22f5edd..32d6a48 100644 --- a/src/qemu_conf.c +++ b/src/qemu_conf.c @@ -1034,7 +1034,7 @@ qemudNetworkIfaceConnect(virConnectPtr conn, virDomainNetDefPtr net, int qemuCmdFlags) { - char *brname; + char *brname = NULL; int err; int tapfd = -1; int vnet_hdr = 0; @@ -1053,7 +1053,7 @@ qemudNetworkIfaceConnect(virConnectPtr conn, if (brname == NULL) return -1; } else if (net->type == VIR_DOMAIN_NET_TYPE_BRIDGE) { - brname = net->data.bridge.brname; + brname = strdup(net->data.bridge.brname); } else { qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR, _("Network type %d is not supported"), net->type); @@ -1063,7 +1063,7 @@ qemudNetworkIfaceConnect(virConnectPtr conn, if (!driver->brctl && (err = brInit(&driver->brctl))) { virReportSystemError(conn, err, "%s", _("cannot initialize bridge support")); - return -1; + goto cleanup; } if (!net->ifname || @@ -1072,7 +1072,7 @@ qemudNetworkIfaceConnect(virConnectPtr conn, VIR_FREE(net->ifname); if (!(net->ifname = strdup("vnet%d"))) { virReportOOMError(conn); - return -1; + goto cleanup; } /* avoid exposing vnet%d in dumpxml or error outputs */ template_ifname = 1; @@ -1100,9 +1100,12 @@ qemudNetworkIfaceConnect(virConnectPtr conn, } if (template_ifname) VIR_FREE(net->ifname); - return -1; + tapfd = -1; } +cleanup: + VIR_FREE(brname); + return tapfd; } diff --git a/src/storage_backend_fs.c b/src/storage_backend_fs.c index ca6d329..222e591 100644 --- a/src/storage_backend_fs.c +++ b/src/storage_backend_fs.c @@ -856,6 +856,8 @@ virStorageBackendFileSystemRefresh(virConnectPtr conn, vol->type = VIR_STORAGE_VOL_FILE; vol->target.format = VIR_STORAGE_VOL_FILE_RAW; /* Real value is filled in during probe */ + if (vol->target.path) + VIR_FREE(vol->target.path); if (virAsprintf(&vol->target.path, "%s/%s", pool->def->target.path, vol->name) == -1) @@ -1022,6 +1024,9 @@ virStorageBackendFileSystemVolCreate(virConnectPtr conn, vol->type = VIR_STORAGE_VOL_FILE; + if (vol->target.path) + VIR_FREE(vol->target.path); + if (virAsprintf(&vol->target.path, "%s/%s", pool->def->target.path, vol->name) == -1) { -- 1.6.0.6

4 7

[libvirt] [PATCH] storage_driver.c: remove two useless calls to virStorageBackendForType
by Jim Meyering 04 Sep '09

04 Sep '09

>From dead-store warnings: >From c925b02828083a016f7b9a7ae93eb3a7bf5f4609 Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Fri, 4 Sep 2009 17:18:29 +0200 Subject: [PATCH] storage_driver.c: remove two useless calls to virStorageBackendForType * src/storage_driver.c (storagePoolGetInfo, storagePoolDefine): The result was never used, and the function has no side effects. --- src/storage_driver.c | 8 -------- 1 files changed, 0 insertions(+), 8 deletions(-) diff --git a/src/storage_driver.c b/src/storage_driver.c index e9ecb20..e8d43b7 100644 --- a/src/storage_driver.c +++ b/src/storage_driver.c @@ -519,15 +519,11 @@ storagePoolDefine(virConnectPtr conn, virStoragePoolDefPtr def; virStoragePoolObjPtr pool = NULL; virStoragePoolPtr ret = NULL; - virStorageBackendPtr backend; storageDriverLock(driver); if (!(def = virStoragePoolDefParseString(conn, xml))) goto cleanup; - if ((backend = virStorageBackendForType(def->type)) == NULL) - goto cleanup; - if (!(pool = virStoragePoolObjAssignDef(conn, &driver->pools, def))) goto cleanup; @@ -847,7 +843,6 @@ storagePoolGetInfo(virStoragePoolPtr obj, virStoragePoolInfoPtr info) { virStorageDriverStatePtr driver = obj->conn->storagePrivateData; virStoragePoolObjPtr pool; - virStorageBackendPtr backend; int ret = -1; storageDriverLock(driver); @@ -860,9 +855,6 @@ storagePoolGetInfo(virStoragePoolPtr obj, goto cleanup; } - if ((backend = virStorageBackendForType(pool->def->type)) == NULL) - goto cleanup; - memset(info, 0, sizeof(virStoragePoolInfo)); if (pool->active) info->state = VIR_STORAGE_POOL_RUNNING; -- 1.6.4.2.409.g85dc3

2 3

[libvirt] [PATCH 6/7] VMware ESX: Add esxDomainUndefine() based on esxVI_UnregisterVM()
by Matthias Bolte 04 Sep '09

04 Sep '09

* src/esx/esx_driver.c: add esxDomainUndefine() based on esxVI_UnregisterVM() * src/esx/esx_vi_methods.[ch]: add esxVI_UnregisterVM()

2 3

[libvirt] [PATCH 5/7] VMware ESX: Add esxGetCapabilities() with basic default capabilities
by Matthias Bolte 04 Sep '09

04 Sep '09

* src/esx/esx_driver.c: add esxCapsInit() with default caps and add esxGetCapabilities()

2 3

[libvirt] [PATCH 1/7] VMware ESX: Fix VMX path parsing and URL encoding
by Matthias Bolte 04 Sep '09

04 Sep '09

* src/esx/esx_driver.c: handle spaces in VMX file path and use a virBuffer to encode spaces correctly in the resulting URL * src/esx/esx_vi.c: include the URL in the error message in case of a download error

2 3

[libvirt] [PATCH] domain_conf.c: remove dead store
by Jim Meyering 04 Sep '09

04 Sep '09

Another dead store: >From 06c5ed162cbdcfec74461e9d09c82244242cc6a2 Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Fri, 4 Sep 2009 17:27:34 +0200 Subject: [PATCH] domain_conf.c: remove dead store * src/domain_conf.c (virDomainSaveXML): ...and decl of "err". --- src/domain_conf.c | 3 +-- 1 files changed, 1 insertions(+), 2 deletions(-) diff --git a/src/domain_conf.c b/src/domain_conf.c index 8dde5dd..4e8ad37 100644 --- a/src/domain_conf.c +++ b/src/domain_conf.c @@ -4456,12 +4456,11 @@ int virDomainSaveXML(virConnectPtr conn, char *configFile = NULL; int fd = -1, ret = -1; size_t towrite; - int err; if ((configFile = virDomainConfigFile(conn, configDir, def->name)) == NULL) goto cleanup; - if ((err = virFileMakePath(configDir))) { + if (virFileMakePath(configDir)) { virReportSystemError(conn, errno, _("cannot create config directory '%s'"), configDir); -- 1.6.4.2.409.g85dc3

2 1

[libvirt] [PATCH] xm_internal.c: remove two ret=... dead stores
by Jim Meyering 04 Sep '09

04 Sep '09

Two more dead stores: >From 92bf654b02e5c1eb3334fc13ef52f5b0679512ea Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Fri, 4 Sep 2009 17:22:19 +0200 Subject: [PATCH] xm_internal.c: remove two ret=... dead stores * src/xm_internal.c (xenXMDomainCreate): Remove dead stores. --- src/xm_internal.c | 4 ++-- 1 files changed, 2 insertions(+), 2 deletions(-) diff --git a/src/xm_internal.c b/src/xm_internal.c index e675672..e948fa0 100644 --- a/src/xm_internal.c +++ b/src/xm_internal.c @@ -1851,10 +1851,10 @@ int xenXMDomainCreate(virDomainPtr domain) { goto error; domain->id = ret; - if ((ret = xend_wait_for_devices(domain->conn, domain->name)) < 0) + if (xend_wait_for_devices(domain->conn, domain->name) < 0) goto error; - if ((ret = xenDaemonDomainResume(domain)) < 0) + if (xenDaemonDomainResume(domain) < 0) goto error; xenUnifiedUnlock(priv); -- 1.6.4.2.409.g85dc3

2 1

[libvirt] [PATCH] lxc_container.c: avoid a leak on error paths
by Jim Meyering 04 Sep '09

04 Sep '09

This started with a dead-store report: File: lxc_container.c Location: line 417, column 10 Description: Although the value stored to 'rc' is used in the enclosing expression, the value is never actually read from 'rc' But there was a leak, too, upon any but the first failure. This fixes both. >From 141ba8bab2de00c911017bff6fca240c72a80633 Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Fri, 4 Sep 2009 16:12:35 +0200 Subject: [PATCH] lxc_container.c: avoid a leak on error paths * src/lxc_container.c (lxcContainerMountBasicFS): Don't leak upon failure. Add "cleanup:" label and change each post-allocation failure to use "goto cleanup" rather than returning immediately. --- src/lxc_container.c | 20 ++++++++++++-------- 1 files changed, 12 insertions(+), 8 deletions(-) diff --git a/src/lxc_container.c b/src/lxc_container.c index 2073864..f4381e7 100644 --- a/src/lxc_container.c +++ b/src/lxc_container.c @@ -1,6 +1,6 @@ /* * Copyright IBM Corp. 2008 - * Copyright Red Hat 2008 + * Copyright Red Hat 2008-2009 * * lxc_container.c: file description * @@ -384,12 +384,12 @@ static int lxcContainerMountBasicFS(virDomainFSDefPtr root) { "none", "/selinux", "selinuxfs" }, #endif }; - int i, rc; + int i, rc = -1; char *devpts; if (virAsprintf(&devpts, "/.oldroot%s/dev/pts", root->src) < 0) { virReportOOMError(NULL); - return -1; + return rc; } for (i = 0 ; i < ARRAY_CARDINALITY(mnts) ; i++) { @@ -397,31 +397,35 @@ static int lxcContainerMountBasicFS(virDomainFSDefPtr root) virReportSystemError(NULL, errno, _("failed to mkdir %s"), mnts[i].src); - return -1; + goto cleanup; } if (mount(mnts[i].src, mnts[i].dst, mnts[i].type, 0, NULL) < 0) { virReportSystemError(NULL, errno, _("failed to mount %s on %s"), mnts[i].type, mnts[i].type); - return -1; + goto cleanup; } } if ((rc = virFileMakePath("/dev/pts") < 0)) { virReportSystemError(NULL, rc, "%s", _("cannot create /dev/pts")); - return -1; + goto cleanup; } VIR_DEBUG("Trying to move %s to %s", devpts, "/dev/pts"); if ((rc = mount(devpts, "/dev/pts", NULL, MS_MOVE, NULL)) < 0) { virReportSystemError(NULL, errno, "%s", _("failed to mount /dev/pts in container")); - return -1; + goto cleanup; } + + rc = 0; + + cleanup: VIR_FREE(devpts); - return 0; + return rc; } static int lxcContainerPopulateDevices(void) -- 1.6.4.2.409.g85dc3

2 2

[libvirt] Dead code vs. XXX comment: remove or not?
by Jim Meyering 04 Sep '09

04 Sep '09

clang reported that this assignment to type is a dead store, since type is never used after this point. This is xm_internal.c, line 1074: /* XXX Forcing to pretend its a bridge */ if (type == -1) { type = 1; } Normally I'd just remove the whole if block, but the XXX comment makes me think someone has plans for this code. I'll wait for a second opinion.

2 3

[libvirt] address 3 more dead stores
by Jim Meyering 04 Sep '09

04 Sep '09

These changes fix two dead stores and add a comment suggesting why *not* to remove the third one. >From 41c778fdbaef9abf52bc4c3e278d3389d8198c34 Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Fri, 4 Sep 2009 10:55:55 +0200 Subject: [PATCH 1/3] interface_conf.c: remove a dead-store and declaration * src/interface_conf.c (virInterfaceDefParseDhcp): Remove unused "old". --- src/interface_conf.c | 2 -- 1 files changed, 0 insertions(+), 2 deletions(-) diff --git a/src/interface_conf.c b/src/interface_conf.c index a74d17e..96e0710 100644 --- a/src/interface_conf.c +++ b/src/interface_conf.c @@ -228,11 +228,9 @@ static int virInterfaceDefParseDhcp(virConnectPtr conn, virInterfaceDefPtr def, xmlNodePtr dhcp, xmlXPathContextPtr ctxt) { char *tmp; - xmlNodePtr old; int ret = 0; def->proto.dhcp = 1; - old = ctxt->node; ctxt->node = dhcp; /* Not much to do in the current version */ tmp = virXPathString(conn, "string(./@peerdns)", ctxt); -- 1.6.4.2.409.g85dc3 >From 10037ca58c8d93e427c480082704e7dd52424525 Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Fri, 4 Sep 2009 11:01:00 +0200 Subject: [PATCH 2/3] hash.c: remove a dead store * src/hash.c (virHashFree): Remove useless assignment to inside_table. --- src/hash.c | 1 - 1 files changed, 0 insertions(+), 1 deletions(-) diff --git a/src/hash.c b/src/hash.c index 9308c0c..45c5747 100644 --- a/src/hash.c +++ b/src/hash.c @@ -231,7 +231,6 @@ virHashFree(virHashTablePtr table, virHashDeallocator f) inside_table = 0; iter = next; } - inside_table = 0; } VIR_FREE(table->table); } -- 1.6.4.2.409.g85dc3 >From 3b16e6d50d364a5307284a78a7b23d8631146663 Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Fri, 4 Sep 2009 11:20:23 +0200 Subject: [PATCH 3/3] qemu_conf.c: add a comment suggesting why we leave a dead-store * src/qemu_conf.c (qemuBuildHostNetStr): Do not remove the type_sep=',' dead store, since not having it would be a problem if we ever add a new attribute=VAL option. --- src/qemu_conf.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/src/qemu_conf.c b/src/qemu_conf.c index 2c4a37d..6ffca2f 100644 --- a/src/qemu_conf.c +++ b/src/qemu_conf.c @@ -1269,7 +1269,7 @@ qemuBuildHostNetStr(virConnectPtr conn, if (net->hostnet_name) { virBufferVSprintf(&buf, "%cname=%s", type_sep, net->hostnet_name); - type_sep = ','; + type_sep = ','; /* dead-store, but leave it, in case... */ } if (virBufferError(&buf)) { virReportOOMError(conn); -- 1.6.4.2.409.g85dc3

2 1

[libvirt] [PATCH 7/7] VMware ESX: Check the result of virUUIDParse() in esxUtil_GetConfigUUID()
by Matthias Bolte 04 Sep '09

04 Sep '09

* src/esx/esx_util.c: let esxUtil_GetConfigUUID() report an error if virUUIDParse() fails

2 1

[libvirt] [PATCH 4/7] VMware ESX: Simplify SOAP request and response handling
by Matthias Bolte 04 Sep '09

04 Sep '09

* src/esx/esx_vi.[ch]: convert esxVI_RemoteRequest_Execute() to a simpler esxVI_Context_Execute() version, remove esxVI_RemoteRequest and convert esxVI_RemoteResponse to esxVI_Response * src/esx/esx_vi_methods.c: update and simplify callers to use esxVI_Context_Execute() instead of esxVI_RemoteRequest_Execute()

2 1

[libvirt] [PATCH 3/7] VMware ESX: Use virXPathNode*() to simplify XPath handling
by Matthias Bolte 04 Sep '09

04 Sep '09

* src/esx/esx_vi.[ch]: use virXPathNode*() in esxVI_RemoteRequest_Execute() and remove esxVI_RemoteResponse_DeserializeXPathObject*() * src/esx/esx_vi_methods.c: update callers to use the new syntax of esxVI_RemoteRequest_Execute()

2 1

[libvirt] [PATCH 2/7] VMware ESX: Make esxVI_GetVirtualMachineIdentity() more robust
by Matthias Bolte 04 Sep '09

04 Sep '09

Check the config status before requesting config subelements * src/esx/esx_driver.c: add configStatus to the requested properties to check it in esxVI_GetVirtualMachineIdentity() * src/esx/esx_vi.[ch]: add esxVI_GetManagedEntityStatus() and use it in esxVI_GetVirtualMachineIdentity() * src/esx/esx_vi_types.[ch]: add VI type esxVI_ManagedEntityStatus

2 1

[libvirt] [PATCH] Fix more OOM handling bugs
by Daniel P. Berrange 04 Sep '09

04 Sep '09

* src/qemu_conf.c: Fix leak of values upon OOM * src/xend_internal.c: Fix missing check for OOM failure * tests/qemuargv2xmltest.c, tests/qemuxml2argvtest.c: Free stateDir upon exit to avoid leak --- src/qemu_conf.c | 21 ++++++++++++++++++--- src/xend_internal.c | 6 ++++++ tests/qemuargv2xmltest.c | 1 + tests/qemuxml2argvtest.c | 1 + 4 files changed, 26 insertions(+), 3 deletions(-) diff --git a/src/qemu_conf.c b/src/qemu_conf.c index 22f5edd..ad41104 100644 --- a/src/qemu_conf.c +++ b/src/qemu_conf.c @@ -1722,6 +1722,8 @@ int qemudBuildCommandLine(virConnectPtr conn, continue; } + ADD_ARG_SPACE; + if (idx < 0) { qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR, _("unsupported disk type '%s'"), disk->dst); @@ -1773,7 +1775,10 @@ int qemudBuildCommandLine(virConnectPtr conn, optstr = virBufferContentAndReset(&opt); - ADD_ARG_LIT("-drive"); + if ((qargv[qargc++] = strdup("-drive")) == NULL) { + VIR_FREE(optstr); + goto no_memory; + } ADD_ARG(optstr); } } else { @@ -1829,6 +1834,7 @@ int qemudBuildCommandLine(virConnectPtr conn, net->vlan = i; + ADD_ARG_SPACE; if ((qemuCmdFlags & QEMUD_CMD_FLAG_NET_NAME) && qemuAssignNetNames(def, net) < 0) goto no_memory; @@ -1836,9 +1842,14 @@ int qemudBuildCommandLine(virConnectPtr conn, if (qemuBuildNicStr(conn, net, NULL, ',', net->vlan, &nic) < 0) goto error; - ADD_ARG_LIT("-net"); + if ((qargv[qargc++] = strdup("-net")) == NULL) { + VIR_FREE(nic); + goto no_memory; + } ADD_ARG(nic); + + ADD_ARG_SPACE; if (net->type == VIR_DOMAIN_NET_TYPE_NETWORK || net->type == VIR_DOMAIN_NET_TYPE_BRIDGE) { int tapfd = qemudNetworkIfaceConnect(conn, driver, net, qemuCmdFlags); @@ -1862,7 +1873,10 @@ int qemudBuildCommandLine(virConnectPtr conn, goto error; } - ADD_ARG_LIT("-net"); + if ((qargv[qargc++] = strdup("-net")) == NULL) { + VIR_FREE(host); + goto no_memory; + } ADD_ARG(host); VIR_FREE(tapfd_name); @@ -2230,6 +2244,7 @@ static int qemuStringToArgvEnv(const char *args, goto no_memory; for (i = 0 ; i < envend ; i++) { progenv[i] = arglist[i]; + arglist[i] = NULL; } progenv[i] = NULL; } diff --git a/src/xend_internal.c b/src/xend_internal.c index 99847b0..20ddb89 100644 --- a/src/xend_internal.c +++ b/src/xend_internal.c @@ -5212,6 +5212,9 @@ xenDaemonFormatSxprChr(virConnectPtr conn, break; } + if (virBufferError(buf)) + return -1; + return 0; } @@ -5535,6 +5538,9 @@ xenDaemonFormatSxprSound(virConnectPtr conn, virBufferVSprintf(buf, "%s%s", i ? "," : "", str); } + if (virBufferError(buf)) + return -1; + return 0; } diff --git a/tests/qemuargv2xmltest.c b/tests/qemuargv2xmltest.c index 7861520..4a92280 100644 --- a/tests/qemuargv2xmltest.c +++ b/tests/qemuargv2xmltest.c @@ -223,6 +223,7 @@ mymain(int argc, char **argv) DO_TEST_FULL("restore-v2", QEMUD_CMD_FLAG_MIGRATE_QEMU_EXEC, "exec:cat"); DO_TEST_FULL("migrate", QEMUD_CMD_FLAG_MIGRATE_QEMU_TCP, "tcp:10.0.0.1:5000"); + free(driver.stateDir); virCapabilitiesFree(driver.caps); return(ret==0 ? EXIT_SUCCESS : EXIT_FAILURE); diff --git a/tests/qemuxml2argvtest.c b/tests/qemuxml2argvtest.c index 6f25e7d..2f91288 100644 --- a/tests/qemuxml2argvtest.c +++ b/tests/qemuxml2argvtest.c @@ -270,6 +270,7 @@ mymain(int argc, char **argv) DO_TEST_FULL("restore-v2", QEMUD_CMD_FLAG_MIGRATE_QEMU_EXEC, "exec:cat"); DO_TEST_FULL("migrate", QEMUD_CMD_FLAG_MIGRATE_QEMU_TCP, "tcp:10.0.0.1:5000"); + free(driver.stateDir); virCapabilitiesFree(driver.caps); return(ret==0 ? EXIT_SUCCESS : EXIT_FAILURE); -- 1.6.2.5

2 1

[libvirt] [PATCH] Fix logging buffer overrun read
by Daniel P. Berrange 04 Sep '09

04 Sep '09

* src/logging.c: Fix buffer offset in logging read --- src/logging.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/src/logging.c b/src/logging.c index e46e2be..07c2b0e 100644 --- a/src/logging.c +++ b/src/logging.c @@ -240,7 +240,7 @@ static void virLogStr(const char *str, int len) { tmp = LOG_BUFFER_SIZE - virLogEnd; memcpy(&virLogBuffer[virLogEnd], str, tmp); virLogBuffer[LOG_BUFFER_SIZE] = 0; - memcpy(&virLogBuffer[0], &str[len], len - tmp); + memcpy(&virLogBuffer[0], &str[tmp], len - tmp); virLogEnd = len - tmp; } else { memcpy(&virLogBuffer[virLogEnd], str, len); -- 1.6.2.5

3 2

[libvirt] [PATCH] uml_conf.c: don't return an uninitialized pointer
by Jim Meyering 04 Sep '09

04 Sep '09

Another "real" bug: >From 6697607bf0b023ffb692576b31d652d10719b08a Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Thu, 3 Sep 2009 12:05:52 +0200 Subject: [PATCH] uml_conf.c: don't return an uninitialized pointer * src/uml_conf.c (umlBuildCommandLineChr): Initialize "ret" also in the final switch cases. --- src/uml_conf.c | 1 + 1 files changed, 1 insertions(+), 0 deletions(-) diff --git a/src/uml_conf.c b/src/uml_conf.c index 838fedd..2e9c25c 100644 --- a/src/uml_conf.c +++ b/src/uml_conf.c @@ -331,6 +331,7 @@ umlBuildCommandLineChr(virConnectPtr conn, default: umlReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR, _("unsupported chr device type %d"), def->type); + ret = NULL; break; } -- 1.6.4.2.395.ge3d52

3 9

[libvirt] three more clang-inspired dead-store-fixing patches
by Jim Meyering 04 Sep '09

04 Sep '09

>From a2d03c987bb724283207dbeef873178c08a6c4c5 Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Thu, 3 Sep 2009 18:14:48 +0200 Subject: [PATCH 1/3] storage_backend_logical.c: appease clang: remove useless increment * src/storage_backend_logical.c (virStorageBackendLogicalBuildPool): Don't increment "n" when we won't use the result. --- src/storage_backend_logical.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/src/storage_backend_logical.c b/src/storage_backend_logical.c index bc40dd7..43117e8 100644 --- a/src/storage_backend_logical.c +++ b/src/storage_backend_logical.c @@ -437,7 +437,7 @@ virStorageBackendLogicalBuildPool(virConnectPtr conn, goto cleanup; } - vgargv[n++] = NULL; + vgargv[n] = NULL; /* Now create the volume group itself */ if (virRun(conn, vgargv, NULL) < 0) -- 1.6.4.2.395.ge3d52 >From b72ed1ba5d5e6444f000fd9be706a51a90dd2292 Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Thu, 3 Sep 2009 18:23:51 +0200 Subject: [PATCH 2/3] openvz_conf.c: Remove dead store to copy_fd * src/openvz_conf.c (openvz_copyfile): Remove unused assignment. --- src/openvz_conf.c | 1 - 1 files changed, 0 insertions(+), 1 deletions(-) diff --git a/src/openvz_conf.c b/src/openvz_conf.c index 41c6684..2f0471c 100644 --- a/src/openvz_conf.c +++ b/src/openvz_conf.c @@ -692,7 +692,6 @@ openvz_copyfile(char* from_path, char* to_path) fd = -1; if (close(copy_fd) < 0) goto error; - copy_fd = -1; return 0; -- 1.6.4.2.395.ge3d52 >From a8d2eca13635578084e3c2cc2093562d9a8fcfed Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Thu, 3 Sep 2009 18:25:03 +0200 Subject: [PATCH 3/3] eventtest.c: detect write failure and avoid dead stores * tests/eventtest.c (mymain): Exit nonzero upon write failure. This also avoids several dead stores of the form ret = safewrite... --- tests/eventtest.c | 15 +++++++++------ 1 files changed, 9 insertions(+), 6 deletions(-) diff --git a/tests/eventtest.c b/tests/eventtest.c index d25381d..ef37ff1 100644 --- a/tests/eventtest.c +++ b/tests/eventtest.c @@ -248,7 +248,6 @@ resetAll(void) static int mymain(int argc, char **argv) { - int ret = 0; char *progname; int i; pthread_t eventThread; @@ -304,7 +303,8 @@ mymain(int argc, char **argv) /* First time, is easy - just try triggering one of our * registered handles */ startJob("Simple write", &test); - ret = safewrite(handles[1].pipeFD[1], &one, 1); + if (safewrite(handles[1].pipeFD[1], &one, 1) != 1) + return EXIT_FAILURE; if (finishJob(1, -1) != EXIT_SUCCESS) return EXIT_FAILURE; @@ -314,7 +314,8 @@ mymain(int argc, char **argv) * try triggering another handle */ virEventRemoveHandleImpl(handles[0].watch); startJob("Deleted before poll", &test); - ret = safewrite(handles[1].pipeFD[1], &one, 1); + if (safewrite(handles[1].pipeFD[1], &one, 1) != 1) + return EXIT_FAILURE; if (finishJob(1, -1) != EXIT_SUCCESS) return EXIT_FAILURE; @@ -346,8 +347,9 @@ mymain(int argc, char **argv) * about */ startJob("Deleted during dispatch", &test); handles[2].delete = handles[3].watch; - ret = safewrite(handles[2].pipeFD[1], &one, 1); - ret = safewrite(handles[3].pipeFD[1], &one, 1); + if (safewrite(handles[2].pipeFD[1], &one, 1) != 1 + || safewrite(handles[3].pipeFD[1], &one, 1) != 1) + return EXIT_FAILURE; if (finishJob(2, -1) != EXIT_SUCCESS) return EXIT_FAILURE; @@ -356,7 +358,8 @@ mymain(int argc, char **argv) /* Extreme fun, lets delete ourselves during dispatch */ startJob("Deleted during dispatch", &test); handles[2].delete = handles[2].watch; - ret = safewrite(handles[2].pipeFD[1], &one, 1); + if (safewrite(handles[2].pipeFD[1], &one, 1) != 1) + return EXIT_FAILURE; if (finishJob(2, -1) != EXIT_SUCCESS) return EXIT_FAILURE; -- 1.6.4.2.395.ge3d52

2 2

[libvirt] [PATCH] Added Storage Volume support to VIrtualBox
by Pritesh Kothari 04 Sep '09

04 Sep '09

Hi All, I have added support for Storage Volumes in VirtualBox, the patch for the same is attached here. (needs the "little bit cleanup" patch previously posted) Regards, Pritesh

2 1

[libvirt] [PATCH] openvz_conf.c: don't use undefined local, "net"
by Jim Meyering 04 Sep '09

04 Sep '09

This looks like a real bug. >From 170af3320e68a0ac2cfe854fba28abe0e4040d2c Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Thu, 3 Sep 2009 11:24:44 +0200 Subject: [PATCH] openvz_conf.c: don't use undefined local, "net" * src/openvz_conf.c (openvzReadNetworkConf): Upon openvzRead... failure, simply return -1, rather than "goto error;" where an uninitialized "net" could be dereferenced. --- src/openvz_conf.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/src/openvz_conf.c b/src/openvz_conf.c index a172fe3..b1cb31a 100644 --- a/src/openvz_conf.c +++ b/src/openvz_conf.c @@ -197,7 +197,7 @@ openvzReadNetworkConf(virConnectPtr conn, openvzError(conn, VIR_ERR_INTERNAL_ERROR, _("Could not read 'IP_ADDRESS' from config for container %d"), veid); - goto error; + return -1; } else if (ret > 0) { token = strtok_r(temp, " ", &saveptr); while (token != NULL) { -- 1.6.4.2.395.ge3d52

3 4

[libvirt] [PATCH] storage_backend.c: assure clang that inputvol can't be NULL
by Jim Meyering 04 Sep '09

04 Sep '09

clang was complaining that a NULL inputvol would be dereferenced in that "could not open..." diagnostic. Since the two sole callers of this function are careful to call it only when inputvol is non-NULL, this is a good case for giving the parameter the nonnull attribute: >From 314278acb021a1f2e63494fab352bd6e9a4a38bb Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Thu, 3 Sep 2009 10:22:57 +0200 Subject: [PATCH] storage_backend.c: assure clang that inputvol can't be NULL * src/storage_backend.c (virStorageBackendCopyToFD): Declare inputvol parameter to be "nonnull". Remove test for non-NULL inputvol. Both callers ensure it's non-NULL. --- src/storage_backend.c | 16 +++++++--------- 1 files changed, 7 insertions(+), 9 deletions(-) diff --git a/src/storage_backend.c b/src/storage_backend.c index c818142..8d1187e 100644 --- a/src/storage_backend.c +++ b/src/storage_backend.c @@ -1,7 +1,7 @@ /* * storage_backend.c: internal storage driver backend contract * - * Copyright (C) 2007-2008 Red Hat, Inc. + * Copyright (C) 2007-2009 Red Hat, Inc. * Copyright (C) 2007-2008 Daniel P. Berrange * * This library is free software; you can redistribute it and/or @@ -106,7 +106,7 @@ enum { static int virStorageBackendCopyToFD(virConnectPtr conn, virStorageVolDefPtr vol, - virStorageVolDefPtr inputvol, + virStorageVolDefPtr inputvol ATTRIBUTE_NONNULL, int fd, unsigned long long *total, int is_dest_file) @@ -119,13 +119,11 @@ virStorageBackendCopyToFD(virConnectPtr conn, char zerobuf[512]; char *buf = NULL; - if (inputvol) { - if ((inputfd = open(inputvol->target.path, O_RDONLY)) < 0) { - virReportSystemError(conn, errno, - _("could not open input path '%s'"), - inputvol->target.path); - goto cleanup; - } + if ((inputfd = open(inputvol->target.path, O_RDONLY)) < 0) { + virReportSystemError(conn, errno, + _("could not open input path '%s'"), + inputvol->target.path); + goto cleanup; } bzero(&zerobuf, sizeof(zerobuf)); -- 1.6.4.2.395.ge3d52

3 4

[libvirt] [PATCH] little bit cleanup
by Pritesh Kothari 04 Sep '09

04 Sep '09

Hi All, Just cleaned up some code to make the it more readable. The patch is attached here with. Directly applies to the HEAD as of today. Regards, Pritesh

3 2

[libvirt] [PATCH] Minor comment changes.
by Laine Stump 04 Sep '09

04 Sep '09

Fix some minor grammer (and one other) nits in comments that end up in generated API reference documentation. No functional/binary differences. --- src/libvirt.c | 24 ++++++++++++------------ 1 files changed, 12 insertions(+), 12 deletions(-) diff --git a/src/libvirt.c b/src/libvirt.c index ca8e003..ce3d6ba 100644 --- a/src/libvirt.c +++ b/src/libvirt.c @@ -592,9 +592,9 @@ virRegisterNetworkDriver(virNetworkDriverPtr driver) /** * virRegisterInterfaceDriver: - * @driver: pointer to a interface driver block + * @driver: pointer to an interface driver block * - * Register a interface virtualization driver + * Register an interface virtualization driver * * Returns the driver priority or -1 in case of error. */ @@ -5365,7 +5365,7 @@ error: /** * virNetworkGetXMLDesc: * @network: a network object - * @flags: and OR'ed set of extraction flags, not used yet + * @flags: an OR'ed set of extraction flags, not used yet * * Provide an XML description of the network. The description may be reused * later to relaunch the network with virNetworkCreateXML(). @@ -5545,7 +5545,7 @@ error: /** * virInterfaceGetConnect: - * @iface: pointer to a interface + * @iface: pointer to an interface * * Provides the connection pointer associated with an interface. The * reference counter on the connection is not increased by this @@ -5819,7 +5819,7 @@ error: /** * virInterfaceGetName: - * @iface: a interface object + * @iface: an interface object * * Get the public name for that interface * @@ -5842,9 +5842,9 @@ virInterfaceGetName(virInterfacePtr iface) /** * virInterfaceGetMACString: - * @iface: a interface object + * @iface: an interface object * - * Get the MAC for a interface as string. For more information about + * Get the MAC for an interface as string. For more information about * MAC see RFC4122. * * Returns a pointer to the MAC address (in null-terminated ASCII @@ -5867,11 +5867,11 @@ virInterfaceGetMACString(virInterfacePtr iface) /** * virInterfaceGetXMLDesc: - * @iface: a interface object - * @flags: and OR'ed set of extraction flags, not used yet + * @iface: an interface object + * @flags: an OR'ed set of extraction flags, not used yet * * Provide an XML description of the interface. The description may be reused - * later to recreate the interface with virInterfaceCreateXML(). + * later to redefine the interface with virInterfaceDefineXML(). * * Returns a 0 terminated UTF-8 encoded XML instance, or NULL in case of error. * the caller must free() the returned value. @@ -6100,7 +6100,7 @@ error: * This method is typically useful for applications where multiple * threads are using a connection, and it is required that the * connection remain open until all threads have finished using - * it. ie, each new thread using a interface would increment + * it. ie, each new thread using an interface would increment * the reference count. * * Returns 0 in case of success, -1 in case of failure. @@ -6121,7 +6121,7 @@ virInterfaceRef(virInterfacePtr iface) /** * virInterfaceFree: - * @iface: a interface object + * @iface: an interface object * * Free the interface object. The interface itself is unaltered. * The data structure is freed and should not be used thereafter. -- 1.6.2.5

3 2

[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

[libvirt] [PATCH 2/2]: VirtualBox: Updated vboxNetworkUndefine() and vboxNetworkDestroy()
by Pritesh Kothari 03 Sep '09

03 Sep '09

Hi All, I have made some changes to the functions vboxNetworkCreateXML(), vboxNetworkDefineXML(), vboxNetworkUndefine() and vboxNetworkDestroy() to handle multiple host only interfaces as multiple host only interfaces are supported by VirtualBox 3.0 and greater. The patch's are as below: PATCH 1/2: Merged vboxNetworkCreateXML() and vboxNetworkDefineXML() and added code to handle multiple hostonly interfaces. PATCH 2/2: Merged vboxNetworkUndefine() and vboxNetworkDestroy() and added code to handle multiple hostonly interfaces. Regards, Pritesh

3 2

[libvirt] [PATCH] esx_vi: remove unreachable code
by Jim Meyering 03 Sep '09

03 Sep '09

This removes dead code and suppresses a warning from clang. However, see below: >From 2ed548cbe948aa7ef37ccd121bd084bd05019ac8 Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Thu, 3 Sep 2009 16:35:47 +0200 Subject: [PATCH] esx_vi: remove unreachable code * src/esx/esx_vi.c (esxVI_Enumeration_Deserialize): Remove unreachable code. --- src/esx/esx_vi.c | 2 -- 1 files changed, 0 insertions(+), 2 deletions(-) diff --git a/src/esx/esx_vi.c b/src/esx/esx_vi.c index 0da908b..c689a7a 100644 --- a/src/esx/esx_vi.c +++ b/src/esx/esx_vi.c @@ -857,8 +857,6 @@ esxVI_Enumeration_Deserialize(virConnectPtr conn, failure: goto cleanup; - - result = -1; } -------------------------------------------- This function does not allow the caller to distinguish via its return value whether it has failed. It always returns 0. I see that this function's return value is used in numerous places (each esxVI_.*Deserialize function), so this should be fixed. int esxVI_Enumeration_Deserialize(virConnectPtr conn, const esxVI_Enumeration *enumeration, xmlNodePtr node, int *value) { int i; int result = 0; char *name = NULL; if (value == NULL) { ESX_VI_ERROR(conn, VIR_ERR_INTERNAL_ERROR, "Invalid argument"); goto failure; } *value = 0; /* undefined */ if (esxVI_String_DeserializeValue(conn, node, &name) < 0) { goto failure; } for (i = 0; enumeration->values[i].name != NULL; ++i) { if (STREQ(name, enumeration->values[i].name)) { *value = enumeration->values[i].value; goto cleanup; } } ESX_VI_ERROR(conn, VIR_ERR_INTERNAL_ERROR, "Unknown value '%s' for %s", name, enumeration->type); cleanup: VIR_FREE(name); return result; failure: goto cleanup; }

2 2

[libvirt] [PATCH 0/1] Support huge pages in guests
by Daniel P. Berrange 03 Sep '09

03 Sep '09

This patch is an update of John Cooper's original work from last month http://www.redhat.com/archives/libvir-list/2009-July/msg00753.html Changes since that patch - We create a directory $MOUNTPOINT/libvirt/qemu inside the hugetlbfs mount point. This is neccessary because QEMU now runs as 'qemu:qemu' in Fedora and thus cannot create files in the root. libvirtd will automatically chown() this subdir to allow QEMU guests access - Don't automatically probe for hugetlbfs mount if runing as an unprivileged libvirtd, since user won't have access to the mount unless administrator has manually created them a subdir. - Change the XML to look like <memoryBacking> <hugepages/> </memoryBacking> This is because I fear we may have more config options for memory backing store in the future, such as whether KVM can use KSM for sharing on the guest. - Add a test case for it in the XML -> ARGV convetor - Document the new XML options - Add the new XML options to the RNG schema - Fix setup & auto-detection of hugetlbfs mount if no qemu.conf file exists at all - Add config parameter to the augeas schema - Move the code which finds a mount point into util.c since its not specific to QEMU

4 5

[libvirt] PATCH 0/0: Support setting a disk drive serial (number) v2
by Daniel P. Berrange 03 Sep '09

03 Sep '09

This is a second version of the patch posted here http://www.redhat.com/archives/libvir-list/2009-August/msg00341.html In this version, I make the QEMU driver refuse to run with any serial that doesn't contain 0-9, a-Z, -, _, since there's no way to escape special chars in QEMU command line args. Hopefully we can lift this restriction in a future version of QEMU I also add a check that serial=s is actually supported by QEMU

3 3

[libvirt] [PATCH] mdns.c: remove dead initialization
by Jim Meyering 03 Sep '09

03 Sep '09

The only following uses of group look like this: (always preceded by initialization): AVAHI_DEBUG("Client collision/connecting %p", mdns->client); group = mdns->group; while (group) { if (group->handle) avahi_entry_group_reset(group->handle); group = group->next; } So the value used in the initialization at the top is never used. >From 515e4a4542ce396b8775c618cbe17a30b78299b4 Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Thu, 3 Sep 2009 13:45:22 +0200 Subject: [PATCH] mdns.c: remove dead initialization * qemud/mdns.c (libvirtd_mdns_client_callback): Remove dead initialization. --- qemud/mdns.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/qemud/mdns.c b/qemud/mdns.c index ae6ff08..73b7d11 100644 --- a/qemud/mdns.c +++ b/qemud/mdns.c @@ -178,7 +178,7 @@ static void libvirtd_mdns_create_services(struct libvirtd_mdns_group *group) { static void libvirtd_mdns_client_callback(AvahiClient *c, AvahiClientState state, void *userdata) { struct libvirtd_mdns *mdns = (struct libvirtd_mdns *)userdata; - struct libvirtd_mdns_group *group = mdns->group; + struct libvirtd_mdns_group *group; if (!mdns->client) mdns->client = c; -- 1.6.4.2.395.ge3d52

2 1

[libvirt] [PATCH] node_device_conf.c: remove dead initialization
by Jim Meyering 03 Sep '09

03 Sep '09

The following use of "caps" is the assignment in a for-loop, so the initialization is unnecessary. >From 7032a0b370db1a9489cf19fce0cabbb6959320ab Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Thu, 3 Sep 2009 13:38:57 +0200 Subject: [PATCH] node_device_conf.c: remove dead initialization * src/node_device_conf.c (virNodeDeviceDefFormat): Remove dead initialization. --- src/node_device_conf.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/src/node_device_conf.c b/src/node_device_conf.c index 56a9bb5..5b25bb7 100644 --- a/src/node_device_conf.c +++ b/src/node_device_conf.c @@ -213,7 +213,7 @@ char *virNodeDeviceDefFormat(virConnectPtr conn, const virNodeDeviceDefPtr def) { virBuffer buf = VIR_BUFFER_INITIALIZER; - virNodeDevCapsDefPtr caps = def->caps; + virNodeDevCapsDefPtr caps; char *tmp; virBufferAddLit(&buf, "<device>\n"); -- 1.6.4.2.395.ge3d52

2 1

[libvirt] [PATCH] Fix misc OOM bugs
by Daniel P. Berrange 03 Sep '09

03 Sep '09

* tests/testutils.c: Run test function twice, once to prime it for static allocations, once to count the non-static allocations. * tests/testutilsqemu.c: Initialize variable correctl * src/capabilities.c: Don't free machines variable upon failure since caller must do that * src/xm_internal.c: Add missing check for OOM in building VIF config param --- src/capabilities.c | 17 +++++++++-------- src/xm_internal.c | 3 +++ tests/testutils.c | 12 ++++++++---- tests/testutilsqemu.c | 2 +- 4 files changed, 21 insertions(+), 13 deletions(-) diff --git a/src/capabilities.c b/src/capabilities.c index 193a9fe..289180d 100644 --- a/src/capabilities.c +++ b/src/capabilities.c @@ -354,10 +354,6 @@ virCapabilitiesAddGuest(virCapsPtr caps, if (loader && (guest->arch.defaultInfo.loader = strdup(loader)) == NULL) goto no_memory; - if (nmachines) { - guest->arch.defaultInfo.nmachines = nmachines; - guest->arch.defaultInfo.machines = machines; - } if (VIR_REALLOC_N(caps->guests, caps->nguests + 1) < 0) @@ -365,6 +361,11 @@ virCapabilitiesAddGuest(virCapsPtr caps, caps->guests[caps->nguests] = guest; caps->nguests++; + if (nmachines) { + guest->arch.defaultInfo.nmachines = nmachines; + guest->arch.defaultInfo.machines = machines; + } + return guest; no_memory: @@ -407,10 +408,6 @@ virCapabilitiesAddGuestDomain(virCapsGuestPtr guest, if (loader && (dom->info.loader = strdup(loader)) == NULL) goto no_memory; - if (nmachines) { - dom->info.nmachines = nmachines; - dom->info.machines = machines; - } if (VIR_REALLOC_N(guest->arch.domains, guest->arch.ndomains + 1) < 0) @@ -418,6 +415,10 @@ virCapabilitiesAddGuestDomain(virCapsGuestPtr guest, guest->arch.domains[guest->arch.ndomains] = dom; guest->arch.ndomains++; + if (nmachines) { + dom->info.nmachines = nmachines; + dom->info.machines = machines; + } return dom; diff --git a/src/xm_internal.c b/src/xm_internal.c index 71b852e..66f1e1c 100644 --- a/src/xm_internal.c +++ b/src/xm_internal.c @@ -2048,6 +2048,9 @@ static int xenXMDomainConfigFormatNet(virConnectPtr conn, virBufferVSprintf(&buf, ",vifname=%s", net->ifname); + if (virBufferError(&buf)) + goto cleanup; + if (VIR_ALLOC(val) < 0) { virReportOOMError(conn); goto cleanup; diff --git a/tests/testutils.c b/tests/testutils.c index 7a1dbdc..5072fec 100644 --- a/tests/testutils.c +++ b/tests/testutils.c @@ -367,10 +367,7 @@ int virtTestMain(int argc, } } - if (testOOM) - virAllocTestInit(); - - /* Run once to count allocs, and ensure it passes :-) */ + /* Run once to prime any static allocations & ensure it passes */ ret = (func)(argc, argv); if (ret != EXIT_SUCCESS) goto cleanup; @@ -385,6 +382,13 @@ int virtTestMain(int argc, testOOM++; virSetErrorFunc(NULL, virtTestErrorFuncQuiet); + virAllocTestInit(); + + /* Run again to count allocs, and ensure it passes :-) */ + ret = (func)(argc, argv); + if (ret != EXIT_SUCCESS) + goto cleanup; + approxAlloc = virAllocTestCount(); testCounter++; if (testDebug) diff --git a/tests/testutilsqemu.c b/tests/testutilsqemu.c index 58707e1..6b6a185 100644 --- a/tests/testutilsqemu.c +++ b/tests/testutilsqemu.c @@ -9,7 +9,7 @@ virCapsPtr testQemuCapsInit(void) { struct utsname utsname; virCapsPtr caps; virCapsGuestPtr guest; - virCapsGuestMachinePtr *machines; + virCapsGuestMachinePtr *machines = NULL; int nmachines; static const char *const x86_machines[] = { "pc", "isapc" -- 1.6.2.5

2 2

[libvirt] [PATCH] remote_internal.c: don't dereference a NULL "conn"
by Jim Meyering 03 Sep '09

03 Sep '09

Here's the test just before the else-if in the patch below: if (conn && conn->driver && STREQ (conn->driver->name, "remote")) { So, in the else-branch, "conn" is guaranteed to be NULL. And dereferenced. This may be only a theoretical risk, but if so, the test of "conn" above should be changed to an assertion, and/or the parameter should get the nonnull attribute. >From a1b1d36d96f6b50ddf514539af85da20ca671bf5 Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Wed, 2 Sep 2009 11:54:38 +0200 Subject: [PATCH] remote_internal.c: don't dereference a NULL "conn" * src/remote_internal.c (remoteDevMonOpen): Avoid NULL-dereference. --- 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 ea50c11..141fef9 100644 --- a/src/remote_internal.c +++ b/src/remote_internal.c @@ -5148,7 +5148,7 @@ remoteDevMonOpen(virConnectPtr conn, conn->devMonPrivateData = priv; remoteDriverUnlock(priv); return VIR_DRV_OPEN_SUCCESS; - } else if (conn->networkDriver && + } else if (conn && conn->networkDriver && STREQ (conn->networkDriver->name, "remote")) { struct private_data *priv = conn->networkPrivateData; remoteDriverLock(priv); -- 1.6.4.2.395.ge3d52

5 9

[libvirt] [PATCH] test.c: don't use undefined local, "def"
by Jim Meyering 03 Sep '09

03 Sep '09

another real bug: >From fd41bd47252654f84684830e02ecbeb3d26e7dee Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Thu, 3 Sep 2009 12:12:19 +0200 Subject: [PATCH] test.c: don't use undefined local, "def" * src/test.c (testOpenVolumesForPool): Upon early virAsprintf or virXPathNodeSet failure, "goto error" would take us to virStorageVolDefFree(def), but with "def" not defined. Initialize it to NULL. --- src/test.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/src/test.c b/src/test.c index 7c8f85b..895cce1 100644 --- a/src/test.c +++ b/src/test.c @@ -435,7 +435,7 @@ static int testOpenVolumesForPool(virConnectPtr conn, char *vol_xpath; int i, ret, func_ret = -1; xmlNodePtr *vols = NULL; - virStorageVolDefPtr def; + virStorageVolDefPtr def = NULL; /* Find storage volumes */ if (virAsprintf(&vol_xpath, "/node/pool[%d]/volume", poolidx) < 0) { -- 1.6.4.2.395.ge3d52

2 1

[libvirt] Support for XCI (Xen Client)
by Andreas Sommer 03 Sep '09

03 Sep '09

Hi, I just wondered whether the Xen Client project (http://www.xen.org/products/xci.html) has been tested, or is supported by libvirt at all? I'm not sure if it has a different Xen API, but it definitely has different domain configuration files - that's why I ask. Best regards

2 1

[libvirt] [PATCH] Support for defining/dumping video device in VirtualBox
by Pritesh Kothari 03 Sep '09

03 Sep '09

Hi All, I have added support for defining/dumping video device in VirtualBox. The patch for the same is attached here. Also this patch depends on the earlier patch "Support for 3d Acceleration in video tag". Regards, Pritesh

3 3

[libvirt] [PATCH] Support for 3d Acceleration in video tag
by Pritesh Kothari 03 Sep '09

03 Sep '09

Hi All, I have added support for 3d Acceleration in the video tag, the patch for the same is attached here. Regards, Pritesh

3 5

[libvirt] [PATCH] Fix bugs in virDomainMigrate v2 code.
by Chris Lalancette 03 Sep '09

03 Sep '09

Paolo Bonzini points out that in my refactoring of the code for virDomainMigrate(), I added a check for the return value from virDomainMigratePerform(). The problem is that we don't want to exit if we fail, we actually want to go on and do virDomainMigrateFinish2() with a non-0 return code to clean things up. Remove the check. While reproducing this issue, I also noticed that we wouldn't always properly propagate an error message. In particular, I found that if you blocked off the migration ports (with iptables) and then tried the migration, it would actually fail but we would get no failure output from Qemu. Therefore, we would think we succeeded, and leave a huge mess behind us. Execute the monitor command "info migrate", and look for a failure string in there as well. Signed-off-by: Chris Lalancette <clalance(a)redhat.com> --- src/libvirt.c | 2 -- src/qemu_driver.c | 15 +++++++++++++++ 2 files changed, 15 insertions(+), 2 deletions(-) diff --git a/src/libvirt.c b/src/libvirt.c index ca8e003..5c1efb6 100644 --- a/src/libvirt.c +++ b/src/libvirt.c @@ -2962,8 +2962,6 @@ virDomainMigrateVersion2 (virDomainPtr domain, */ ret = domain->conn->driver->domainMigratePerform (domain, cookie, cookielen, uri, flags, dname, bandwidth); - if (ret == -1) - goto done; /* In version 2 of the migration protocol, we pass the * status code from the sender to the destination host, diff --git a/src/qemu_driver.c b/src/qemu_driver.c index eb22940..772f2f9 100644 --- a/src/qemu_driver.c +++ b/src/qemu_driver.c @@ -6956,6 +6956,21 @@ qemudDomainMigratePerform (virDomainPtr dom, goto cleanup; } + /* it is also possible that the migrate didn't fail initially, but + * rather failed later on. Check the output of "info migrate" + */ + VIR_FREE(info); + if (qemudMonitorCommand(vm, "info migrate", &info) < 0) { + qemudReportError (dom->conn, dom, NULL, VIR_ERR_OPERATION_FAILED, + "%s", _("could not get info about migration")); + goto cleanup; + } + if (strstr(info, "fail") != NULL) { + qemudReportError (dom->conn, dom, NULL, VIR_ERR_OPERATION_FAILED, + _("migrate failed: %s"), info); + goto cleanup; + } + /* Clean up the source domain. */ qemudShutdownVMDaemon (dom->conn, driver, vm); paused = 0; -- 1.6.0.6

2 2

[libvirt] virsh vcpuinfo always returns CPU=0
by Shi Jin 02 Sep '09

02 Sep '09

Hi there, I wonder what does the CPU entry stands for in the output of vcpuinfo. I guess it is the CPU utilization rate, right? For both Ubuntu 9.04 (virsh version 0.6.1) and RHEL 5.4 (version 0.6.3), I always get the returned CPU value to be 0, no matter how busy the CPU can be and how many VCPU is used. For example, onnecting to uri: qemu:///system VCPU: 0 CPU: 0 State: running CPU Affinity: yyyyyyyyyyyyyyyy VCPU: 1 CPU: 0 State: running CPU Affinity: yyyyyyyyyyyyyyyy Is this a problem and how do I fix it? Thank you very much. -- Shi Jin, PhD

2 1