February 2013 - Devel - libvirt List Archives

[libvirt] [PATCH 0/2] network: add force attribute for dhcp options
by Laine Stump 25 Feb '13

25 Feb '13

The firs patch just fixes unexpected behavior in virNetworkDHCPOptionDefParseXML found when testing the second patch. The second patch adds the "force" attribute I mentioned in the review of the patch that added support for specifying dhcp options in network config. While documenting this new attribute, I noticed that I'd neglected to see that original patch didn't include documentation, so I added full documentation for <option> in order to have a place to add the documentation for "force='yes'". I'm also working on a way to allow specification of option names rather than numbers, but didn't know if I could get it in before the freeze, so I'm sending this patch now. Laine Stump (2): conf: use VIR_EXPAND instead of VIR_REALLOC in virNetworkDHCPDefParseXML network: add force attribute for dhcp options docs/formatnetwork.html.in | 21 ++++++++++++++++++++ src/conf/network_conf.c | 32 +++++++++++++++++++++--------- src/conf/network_conf.h | 3 ++- src/network/bridge_driver.c | 3 ++- tests/networkxml2confdata/nat-network.conf | 2 +- tests/networkxml2confdata/nat-network.xml | 2 +- 6 files changed, 50 insertions(+), 13 deletions(-) -- 1.7.11.7

3 5

[libvirt] [PATCH tck] Avoid multicast MAC addrs in tests
by Daniel P. Berrange 25 Feb '13

25 Feb '13

From: "Daniel P. Berrange" <berrange(a)redhat.com> Certain MAC addrs are used for multicast and recent libvirt will refuse to start VMs using those addrs. Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com> --- scripts/domain/215-nic-hotplug-many.t | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/scripts/domain/215-nic-hotplug-many.t b/scripts/domain/215-nic-hotplug-many.t index 5b935cc..8b4afbd 100644 --- a/scripts/domain/215-nic-hotplug-many.t +++ b/scripts/domain/215-nic-hotplug-many.t @@ -47,9 +47,9 @@ diag "Creating a new transient domain"; my $dom; ok_domain(sub { $dom = $conn->create_domain($xml) }, "created transient domain object"); -my $mac1 = "01:11:22:33:44:55"; -my $mac2 = "02:11:22:33:44:55"; -my $mac3 = "03:11:22:33:44:55"; +my $mac1 = "02:11:22:33:44:51"; +my $mac2 = "02:11:22:33:44:52"; +my $mac3 = "02:11:22:33:44:53"; my $model = "virtio"; my $netxml1 = <<EOF; -- 1.8.1.2

2 1

[libvirt] [PATCH] Fix crash changing CDROM media
by Daniel P. Berrange 25 Feb '13

25 Feb '13

From: "Daniel P. Berrange" <berrange(a)redhat.com> This change tried to fix a crash with changing CDROM media but failed to actually do so commit d0172d2b1b5d865aaa042070d7c2d00effb2ff8c Author: Osier Yang <jyang(a)redhat.com> Date: Tue Feb 19 20:27:45 2013 +0800 qemu: Remove the shared disk entry if the operation is ejecting or updating It was still accessing disk->src, when the entire 'disk' object has been free'd already. Even if it weren't free'd, accessing the 'src' value of virDomainDiskDef is not allowed without first validating disk->type is file or block. Just remove the broken code entirely. Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com> --- src/qemu/qemu_driver.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index 1e96915..8dae8f9 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -5778,13 +5778,14 @@ qemuDomainAttachDeviceDiskLive(virConnectPtr conn, dev->data.disk = tmp; ret = qemuDomainChangeEjectableMedia(driver, vm, disk, orig_disk, false); + /* 'disk' must not be accessed now - it has been free'd. + * 'orig_disk' now points to the new disk, while 'dev_copy' + * now points to the old disk */ /* Need to remove the shared disk entry for the original disk src * if the operation is either ejecting or updating. */ - if (ret == 0 && - orig_disk->src && - STRNEQ_NULLABLE(orig_disk->src, disk->src)) + if (ret == 0) ignore_value(qemuRemoveSharedDisk(driver, dev_copy->data.disk, vm->def->name)); break; -- 1.8.1.2

2 1

[libvirt] [PATCH] qemu: Refactor qemuDomainSetMemoryParameters
by Peter Krempa 25 Feb '13

25 Feb '13

The new TypedParam helper APIs allow to simplify this function significantly. This patch integrates the fix in 75e5bec97b3045e4f926248d5c742f8a50d0f9 by correctly ordering the setting functions instead of reordering the parameters. --- src/qemu/qemu_driver.c | 149 ++++++++++++++++++------------------------------- 1 file changed, 54 insertions(+), 95 deletions(-) diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index 23499ef..f2f5872 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -7140,15 +7140,15 @@ qemuDomainSetMemoryParameters(virDomainPtr dom, unsigned int flags) { virQEMUDriverPtr driver = dom->conn->privateData; - int i; virDomainDefPtr persistentDef = NULL; virCgroupPtr group = NULL; virDomainObjPtr vm = NULL; - virTypedParameterPtr hard_limit = NULL; - virTypedParameterPtr swap_hard_limit = NULL; - int hard_limit_index = 0; - int swap_hard_limit_index = 0; - unsigned long long val = 0; + unsigned long long swap_hard_limit; + unsigned long long memory_hard_limit; + unsigned long long memory_soft_limit; + bool set_swap_hard_limit = false; + bool set_memory_hard_limit = false; + bool set_memory_soft_limit = false; virQEMUDriverConfigPtr cfg = NULL; int ret = -1; int rc; @@ -7167,13 +7167,9 @@ qemuDomainSetMemoryParameters(virDomainPtr dom, NULL) < 0) return -1; - vm = virDomainObjListFindByUUID(driver->domains, dom->uuid); - if (vm == NULL) { - virReportError(VIR_ERR_INTERNAL_ERROR, - _("No such domain %s"), dom->uuid); - goto cleanup; - } + if (!(vm = qemuDomObjFromDomain(dom))) + return -1; cfg = virQEMUDriverGetConfig(driver); @@ -7198,110 +7194,73 @@ qemuDomainSetMemoryParameters(virDomainPtr dom, } } - for (i = 0; i < nparams; i++) { - if (STREQ(params[i].field, VIR_DOMAIN_MEMORY_HARD_LIMIT)) { - hard_limit = &params[i]; - hard_limit_index = i; - } else if (STREQ(params[i].field, VIR_DOMAIN_MEMORY_SWAP_HARD_LIMIT)) { - swap_hard_limit = &params[i]; - swap_hard_limit_index = i; - } - } +#define VIR_GET_LIMIT_PARAMETER(PARAM, VALUE) \ + if ((rc = virTypedParamsGetULLong(params, nparams, PARAM, &VALUE) < 0)) \ + goto cleanup; \ + \ + if (rc == 1) \ + set_ ## VALUE = true; + + VIR_GET_LIMIT_PARAMETER(VIR_DOMAIN_MEMORY_SWAP_HARD_LIMIT, swap_hard_limit) + VIR_GET_LIMIT_PARAMETER(VIR_DOMAIN_MEMORY_HARD_LIMIT, memory_hard_limit) + VIR_GET_LIMIT_PARAMETER(VIR_DOMAIN_MEMORY_SOFT_LIMIT, memory_soft_limit) + +#undef VIR_GET_LIMIT_PARAMETER + /* It will fail if hard limit greater than swap hard limit anyway */ - if (swap_hard_limit && - hard_limit && - (hard_limit->value.ul > swap_hard_limit->value.ul)) { + if (set_swap_hard_limit && set_memory_hard_limit && + (memory_hard_limit > swap_hard_limit)) { virReportError(VIR_ERR_INVALID_ARG, "%s", _("hard limit must be lower than swap hard limit")); goto cleanup; } - if (flags & VIR_DOMAIN_AFFECT_LIVE) { - /* Get current swap hard limit */ - rc = virCgroupGetMemSwapHardLimit(group, &val); - if (rc != 0) { + if (set_swap_hard_limit) { + if (flags & VIR_DOMAIN_AFFECT_LIVE && + (rc = virCgroupSetMemSwapHardLimit(group, swap_hard_limit)) < 0) { virReportSystemError(-rc, "%s", - _("unable to get swap hard limit")); + _("unable to set memory swap_hard_limit tunable")); goto cleanup; } - /* Swap hard_limit and swap_hard_limit to ensure the setting - * could succeed if both of them are provided. - */ - if (swap_hard_limit && hard_limit) { - virTypedParameter param; - - if (swap_hard_limit->value.ul > val) { - if (hard_limit_index < swap_hard_limit_index) { - param = params[hard_limit_index]; - params[hard_limit_index] = params[swap_hard_limit_index]; - params[swap_hard_limit_index] = param; - } - } else { - if (hard_limit_index > swap_hard_limit_index) { - param = params[hard_limit_index]; - params[hard_limit_index] = params[swap_hard_limit_index]; - params[swap_hard_limit_index] = param; - } - } - } + if (flags & VIR_DOMAIN_AFFECT_CONFIG) + persistentDef->mem.swap_hard_limit = swap_hard_limit; } - ret = 0; - for (i = 0; i < nparams; i++) { - virTypedParameterPtr param = &params[i]; - - if (STREQ(param->field, VIR_DOMAIN_MEMORY_HARD_LIMIT)) { - if (flags & VIR_DOMAIN_AFFECT_LIVE) { - rc = virCgroupSetMemoryHardLimit(group, param->value.ul); - if (rc != 0) { - virReportSystemError(-rc, "%s", - _("unable to set memory hard_limit tunable")); - ret = -1; - } - } + if (set_memory_hard_limit) { + if (flags & VIR_DOMAIN_AFFECT_LIVE && + (rc = virCgroupSetMemoryHardLimit(group, memory_hard_limit)) < 0) { + virReportSystemError(-rc, "%s", + _("unable to set memory hard_limit tunable")); + goto cleanup; + } - if (flags & VIR_DOMAIN_AFFECT_CONFIG) { - persistentDef->mem.hard_limit = param->value.ul; - } - } else if (STREQ(param->field, VIR_DOMAIN_MEMORY_SOFT_LIMIT)) { - if (flags & VIR_DOMAIN_AFFECT_LIVE) { - rc = virCgroupSetMemorySoftLimit(group, param->value.ul); - if (rc != 0) { - virReportSystemError(-rc, "%s", - _("unable to set memory soft_limit tunable")); - ret = -1; - } - } + if (flags & VIR_DOMAIN_AFFECT_CONFIG) + persistentDef->mem.hard_limit = memory_hard_limit; + } - if (flags & VIR_DOMAIN_AFFECT_CONFIG) { - persistentDef->mem.soft_limit = param->value.ul; - } - } else if (STREQ(param->field, VIR_DOMAIN_MEMORY_SWAP_HARD_LIMIT)) { - if (flags & VIR_DOMAIN_AFFECT_LIVE) { - rc = virCgroupSetMemSwapHardLimit(group, param->value.ul); - if (rc != 0) { - virReportSystemError(-rc, "%s", - _("unable to set swap_hard_limit tunable")); - ret = -1; - } - } - if (flags & VIR_DOMAIN_AFFECT_CONFIG) { - persistentDef->mem.swap_hard_limit = param->value.ul; - } + if (set_memory_soft_limit) { + if (flags & VIR_DOMAIN_AFFECT_LIVE && + (rc = virCgroupSetMemorySoftLimit(group, memory_soft_limit)) < 0) { + virReportSystemError(-rc, "%s", + _("unable to set memory soft_limit tunable")); + goto cleanup; } - } - if (flags & VIR_DOMAIN_AFFECT_CONFIG) { - if (virDomainSaveConfig(cfg->configDir, persistentDef) < 0) - ret = -1; + if (flags & VIR_DOMAIN_AFFECT_CONFIG) + persistentDef->mem.soft_limit = memory_soft_limit; } + if (flags & VIR_DOMAIN_AFFECT_CONFIG && + virDomainSaveConfig(cfg->configDir, persistentDef) < 0) + goto cleanup; + + ret = 0; + cleanup: virCgroupFree(&group); - if (vm) - virObjectUnlock(vm); + virObjectUnlock(vm); virObjectUnref(caps); virObjectUnref(cfg); return ret; -- 1.8.1.1

2 2

[libvirt] [PATCHv2 0/5] Add API to allow TCP connection tunelling
by Peter Krempa 25 Feb '13

25 Feb '13

This series adds ability for the qemu driver to tunnel connections to TCP ports from the host. This is useful for enabling remote VNC/SPICE sessions without the need to configure SSH tunnels or portforwards and without the need to open the ports for public. There's also an advantage for tools such as virt-viewer that have to guess the remote connection parameters and you have to hope that nothing is in your way. With spice/VNC clients that have support for read/write callbacks, this would allow also direct connection without an intermediate socket. The API and tunelling works but there's no (stable and good working) client for this API. I hacked up a dirty netcat-like terminal into virsh for testing purposes (see patch 5/5) but that isn't what I'd like to see. The client should be able to open a listening socket and when a client connects to it, it opens a stream and connects it to the remote host. For the client there are two options: 1) do all the stuff in virsh: + one tool to rule them all - i'd like to daemonize it and I don't know if that's okay in virsh 2) add a new tool "virtunnel": + less virsh pollution - separate tool ... As nobody responded, I'd like to re-ask for someones opinion on this. (note: this is my personal effort, I'm annoyed of opening ssh tunnels to remote displays on my server and I don't want to open the ports to public. ) After this it would be great to add support for this to virt-viewer. I will have a look at that later. ---- Diff to v1: - fixed error reporting in 2/5 - documented limitation to "localhost" in 3/5 - fixed possible segfault in 4/5 --- Peter Krempa (5): api: Add API to allow TCP tunneling through streams to the host fdstream: Add support for TCP connections of streams qemu: Add configuration options to enable TCP tunelling qemu: Implement virNodeTunnelTcp for the qemu driver NOT_TO_BE_APPLIED_UPSTREAM: quick and dirty virsh client to test the stuff include/libvirt/libvirt.h.in | 11 +++++++ src/driver.h | 8 +++++ src/fdstream.c | 73 ++++++++++++++++++++++++++++++++++++++++++++ src/fdstream.h | 5 +++ src/libvirt.c | 67 ++++++++++++++++++++++++++++++++++++++++ src/libvirt_private.syms | 1 + src/libvirt_public.syms | 1 + src/qemu/qemu.conf | 16 ++++++++++ src/qemu/qemu_conf.c | 26 ++++++++++++++++ src/qemu/qemu_conf.h | 13 ++++++++ src/qemu/qemu_driver.c | 48 +++++++++++++++++++++++++++++ src/remote/remote_driver.c | 1 + src/remote/remote_protocol.x | 9 +++++- src/remote_protocol-structs | 6 ++++ src/rpc/gendispatch.pl | 1 + tools/console.c | 66 +++++++++++++++++++++------------------ tools/console.h | 9 ++---- tools/virsh-domain.c | 17 +++++++++-- tools/virsh-host.c | 60 ++++++++++++++++++++++++++++++++++++ 19 files changed, 397 insertions(+), 41 deletions(-) -- 1.8.0

2 7

[libvirt] [PATCH 0/2] Using client-id for DHCPv6 host identification
by Gene Czarcinski 25 Feb '13

25 Feb '13

The first patch is a trivial fix to the network schema since the name in dhcp-host is really optional. The second patch adds the capability to use the client id to improve the reliability of having a system use a fixed IPv6 address. Gene Czarcinski (2): Trivial fix: in dhcp-host the name is optional use client id for IPv6 DHCP host definition docs/formatnetwork.html.in | 18 ++++++- docs/schemas/basictypes.rng | 59 ++++++++++++++++++++++ docs/schemas/network.rng | 7 ++- src/conf/network_conf.c | 29 +++++++++-- src/conf/network_conf.h | 1 + src/network/bridge_driver.c | 3 +- src/util/virdnsmasq.c | 20 ++++++-- src/util/virdnsmasq.h | 1 + tests/networkxml2confdata/dhcp6-nat-network.xml | 5 +- tests/networkxml2confdata/dhcp6-network.xml | 5 +- .../dhcp6host-routed-network.xml | 5 +- 11 files changed, 138 insertions(+), 15 deletions(-) -- 1.8.1.2

3 11

[libvirt] Versioning/Compatibility in the XML schemas
by Pieter Hollants 25 Feb '13

25 Feb '13

Out of curiosity, seeing that we doing modifications to the XML schema all the time, there does not seem to be some sort of version identifier in the XML schemas. If I take the recently merged patch for DHCP options, for instance, suppose virt-manager would add support for it in its GUI. How would such a network XML definition be interpreted by older libvirt versions? Is it supposed to ignore elements it doesn't know or do we have an implicit requirement of "always use the latest versions together"? -- Pieter Hollants <pieter(a)hollants.com>

2 1

[libvirt] [PATCH v2 resend 0/4]use VIR_ENUM_DECL for domain NIC model and add usb-net
by Guannan Ren 25 Feb '13

25 Feb '13

v1 to v2 *removed i8* NIC models support for QEMU/KVM *updated docs *rebased This patchset uses VIR_ENUM_DECL and VIR_ENUM_IMPL macros for NIC models and changes the related codes to use them. All of NIC models supported by various hypervisors comprise of the enum, so the problem is that we need to do further checking in hypervisor-specific implementation. Fortunately, VMX and Vbox did this checking very well already. In 4/4, a similar job is performed for QEMU/KVM. And, adds usb-net support. The vendorId and productID of is 0525:a4a2. The corresponding item in usb-ids is as follows: 0525 Netchip Technology, Inc. a4a2 Linux-USB Ethernet/RNDIS Gadget In my opinion, using usb-net directly is good enough, so I keep it. Any other idea is welcome. Libvirt XML sample: <devices> <interface type='user'> <mac address='52:54:00:32:6a:91'/> <model type='usb-net'/> <alias name='net1'/> <address type='usb' bus='0' port='1'/> </interface> </devices> qemu commandline: qemu ${other_vm_args} -netdev user,id=hostnet1 \ -device usb-net,netdev=hostnet1,id=net1,\ mac=52:54:00:32:6a:91,bus=usb.0,port=1 Guannan Ren(4) [PATCH v2 resend 1/4] add domain NIC model enum macro [PATCH v2 resend 2/4] net: use virDomainNICModelType{From|To}String functions [PATCH v2 resend 3/4] qemu: add NIC model checking for qemu hypervisor [PATCH v2 resend 4/4] qemu: add usb-net support docs/formatdomain.html.in | 29 +++++++++++------------------ src/conf/domain_conf.c | 78 ++++++++++++++++++++++++++++++++++++++++++++++++++++++------------------------ src/conf/domain_conf.h | 32 +++++++++++++++++++++++++++++++- src/libvirt_private.syms | 2 ++ src/libxl/libxl_conf.c | 5 +++-- src/parallels/parallels_driver.c | 2 +- src/qemu/qemu_command.c | 41 +++++++++++++++++++++++++++-------------- src/qemu/qemu_driver.c | 2 +- src/qemu/qemu_hotplug.c | 10 ++++++---- src/qemu/qemu_process.c | 14 +++++++------- src/vbox/vbox_tmpl.c | 57 +++++++++++++++++++++++++++------------------------------ src/vmx/vmx.c | 31 ++++++++++++++++--------------- src/xenxs/xen_sxpr.c | 21 +++++++++++---------- src/xenxs/xen_xm.c | 21 +++++++++++---------- 14 files changed, 208 insertions(+), 137 deletions(-)

1 4

[libvirt] [PATCH] interface: Fix udev backend bridge device display
by Doug Goldstein 25 Feb '13

25 Feb '13

The bridge device was showing the vnet devices created for the domains as connected to the bridge. libvirt should only show host devices when trying to get the interface definition rather than the domain devices as well. --- Honestly this method sucks. But it makes the code path work and doesn't result in brokenness. I was really thinking of sscanf() but I don't really care to store the values. Suggestions? --- src/interface/interface_backend_udev.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) diff --git a/src/interface/interface_backend_udev.c b/src/interface/interface_backend_udev.c index bd83545..dca85b3 100644 --- a/src/interface/interface_backend_udev.c +++ b/src/interface/interface_backend_udev.c @@ -24,7 +24,9 @@ #include <libudev.h> #include "virerror.h" +#include "c-ctype.h" #include "datatypes.h" +#include "domain_conf.h" #include "interface_driver.h" #include "interface_conf.h" #include "viralloc.h" @@ -527,6 +529,16 @@ udevIfaceBridgeScanDirFilter(const struct dirent *entry) if (STREQ(entry->d_name, ".") || STREQ(entry->d_name, "..")) return 0; + /* Omit the domain interfaces from the list of bridge attached + * devices. All we can do is check for the device name matching + * vnet%d. Improvements to this check are welcome. + */ + if (strlen(entry->d_name) >= 5) { + if (STRPREFIX(entry->d_name, VIR_NET_GENERATED_PREFIX) && + c_isdigit(entry->d_name[4])) + return 0; + } + return 1; } -- 1.7.12.4

2 1

[libvirt] [PATCH] docs: add forward mode='hostdev' example
by Laine Stump 24 Feb '13

24 Feb '13

Also rearrange examples so that all "managed" networks (those with bridges created by libvirt) are together. --- docs/formatnetwork.html.in | 52 +++++++++++++++++++++++++++++++++------------- 1 file changed, 37 insertions(+), 15 deletions(-) diff --git a/docs/formatnetwork.html.in b/docs/formatnetwork.html.in index 4b4c47b..46b7270 100644 --- a/docs/formatnetwork.html.in +++ b/docs/formatnetwork.html.in @@ -332,7 +332,7 @@ </p> <pre> ... - <forward mode='passthrough'> + <forward mode='passthrough' managed='yes'> <pf dev='eth0'/> </forward> ... @@ -863,6 +863,27 @@ </ip> </network></pre> + <h3><a name="examplesNoGateway">Network config with no gateway addresses</a></h3> + + <p> + A valid network definition can contain no IPv4 or IPv6 addresses. + Such a definition can be used for a "very private" or "very + isolated" network since it will not be possible to communicate + with the virtualization host via this network. However, this + virtual network interface can be used for communication between + virtual guest systems. This works for IPv4 + and <span class="since">(Since 1.0.1)</span> IPv6. However, + the <code>ipv6</code> attribute must be set to "yes" for + functional guest-to-guest IPv6 communication. + </p> + + <pre> + <network ipv6='yes'> + <name>nogw</name> + <bridge name="virbr2" stp="on" delay="0"/> + </network> + </pre> + <h3><a name="examplesBridge">Using an existing host bridge</a></h3> <p> @@ -916,25 +937,26 @@ </forward> </network></pre> - <h3><a name="examplesNoGateway">Network config with no gateway addresses</a></h3> + <h3><a name="examplesPCIPassthrough">Assigning SR-IOV Virtual Functions via PCI Passthrough</a></h3> <p> - A valid network definition can contain no IPv4 or IPv6 addresses. Such a definition - can be used for a "very private" or "very isolated" network since it will not be - possible to communicate with the virtualization host via this network. However, - this virtual network interface can be used for communication between virtual guest - systems. This works for IPv4 and <span class="since">(Since 1.0.1)</span> IPv6. - However, the new ipv6='yes' must be added for guest-to-guest IPv6 - communication. + <span class="since">Since 0.9.10, requires a host with working + IOMMU</span> This example shows how to use a libvirt network to + assign individual SR-IOV Virtual Functions (VF) to guests using + PCI passthrough. Only the Physical Function (PF) of the SR-IOV + network adapter needs to be listed in the network definition; a + list of all VFs will be automatically derived from that, and VFs + will be allocated to guest domains as requested. </p> <pre> - <network ipv6='yes'> - <name>nogw</name> - <uuid>7a3b7497-1ec7-8aef-6d5c-38dff9109e93</uuid> - <bridge name="virbr2" stp="on" delay="0" /> - <mac address='00:16:3E:5D:C7:9E'/> - </network></pre> + <network> + <name>passthrough</name> + <forward mode='hostdev' managed='yes'> + <pf dev='eth3'/> + </forward> + </network> + </pre> </body> </html> -- 1.7.11.7

1 0