- Devel - libvirt List Archives

[libvirt] [PATCH 0/2] Add non-FreeBSD guest support to Bhyve driver.
by Conrad Meyer 29 Oct '14

29 Oct '14

2 6

[libvirt] [PATCH v3 0/7] hotplug: Fix libvirtd crash on qemu-attached guest
by John Ferlan 29 Oct '14

29 Oct '14

v2 is here: http://www.redhat.com/archives/libvir-list/2014-September/msg01575.html As Michal pointed out in his review - other devices could have the same issue - so take care of each of them separately (I already knew the answer to my last question...) Totally different approach this time - rather than error out, follow the example I from the Controller Detach code which will generate an alias for the device (although it did miss one minor check to see if it was already set leading to a potential memory leak since the alias code would overwrite whatever was there). Note: In a way 1/7 was already ACK'd - I just hadn't separated it yet for a push and wanted to keep these closer together when/if they were pushed. Whether 7/7 is now necessary is debateable - I keep it only for completeness and environment setup in much the same way the start code handles aliases. John Ferlan (7): virsh: Adjust the text in man page regarding qemu-attach hotplug: Check for alias in controller detach hotplug: Check for alias in disk detach hotplug: Check for alias in hostdev detach hotplug: Check for alias in chrdev detach hotplug: Check for alias in net detach qemu-attach: Assign device aliases src/qemu/qemu_driver.c | 3 +++ src/qemu/qemu_hotplug.c | 28 +++++++++++++++++++++++++++- tools/virsh.pod | 5 +++-- 3 files changed, 33 insertions(+), 3 deletions(-) -- 1.9.3

2 11

[libvirt] [PATCH] remote: fix jump depends on uninitialised value
by Jincheng Miao 29 Oct '14

29 Oct '14

Currently remote driver only initializes partial fields of remote_connect_get_all_domain_stats_args. But xdr_array() will check the uninitialised field 'doms_val'. For safty reason, memset all fields of args is better. Fix the following error from valgrind, like: ==30515== 1 errors in context 1 of 3: ==30515== Conditional jump or move depends on uninitialised value(s) ==30515== at 0x85E9402: xdr_array (xdr_array.c:88) ==30515== by 0x4FD8FC9: xdr_remote_connect_get_all_domain_stats_args (remote_protocol.c:6473) ==30515== by 0x4FE72F2: virNetMessageEncodePayload (virnetmessage.c:350) ==30515== by 0x4FDD21C: virNetClientProgramCall (virnetclientprogram.c:326) ==30515== by 0x4FB4D01: callFull.isra.2 (remote_driver.c:6667) ==30515== by 0x4FCBD45: call (remote_driver.c:6689) ==30515== by 0x4FCBD45: remoteConnectGetAllDomainStats (remote_driver.c:7793) ==30515== by 0x4FA0E75: virConnectGetAllDomainStats (libvirt.c:21678) ==30515== by 0x147FD1: cmdDomstats (virsh-domain-monitor.c:2148) ==30515== by 0x13006B: vshCommandRun (virsh.c:1915) ==30515== by 0x12A9E1: main (virsh.c:3699) Signed-off-by: Jincheng Miao <jmiao(a)redhat.com> --- src/remote/remote_driver.c | 2 ++ 1 files changed, 2 insertions(+), 0 deletions(-) diff --git a/src/remote/remote_driver.c b/src/remote/remote_driver.c index 067f2d0..b8d90ed 100644 --- a/src/remote/remote_driver.c +++ b/src/remote/remote_driver.c @@ -7779,6 +7779,8 @@ remoteConnectGetAllDomainStats(virConnectPtr conn, virDomainStatsRecordPtr elem = NULL; virDomainStatsRecordPtr *tmpret = NULL; + memset(&args, 0, sizeof(args)); + if (ndoms) { if (VIR_ALLOC_N(args.doms.doms_val, ndoms) < 0) goto cleanup; -- 1.7.1

2 1

[libvirt] [PATCH] virsh: Improve the job type reported of virsh cmd blockcommit
by Shanzhi Yu 29 Oct '14

29 Oct '14

When begin a blockcommit job(active) with virsh, it will report "Block Commit started"(it really donesn't matter much), but for more preciese it should report "Active Block Commit started" Signed-off-by: Shanzhi Yu <shyu(a)redhat.com> --- tools/virsh-domain.c | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) diff --git a/tools/virsh-domain.c b/tools/virsh-domain.c index 94ae3d3..3416e31 100644 --- a/tools/virsh-domain.c +++ b/tools/virsh-domain.c @@ -1723,9 +1723,15 @@ cmdBlockCommit(vshControl *ctl, const vshCmd *cmd) goto cleanup; if (!blocking) { - vshPrint(ctl, "%s", _("Block Commit started")); - ret = true; - goto cleanup; + if (active) { + vshPrint(ctl, "%s", _("Active Block Commit started")); + ret = true; + goto cleanup; + } else { + vshPrint(ctl, "%s", _("Block Commit started")); + ret = true; + goto cleanup; + } } while (blocking) { -- 1.9.3

2 1

[libvirt] [PATCH 0/2] v1.1 bugfix: support dhcp network interfaces
by Gene Czarcinski 29 Oct '14

29 Oct '14

v1.1 adds some documentation changes. Support for a network such as -N dhcp,source=default was not working in that dhclient was not being started. Although I am not sure what the real problem is, the solution is to use g_spawn_sync() instead of g_spawn_async() to start /sbin/dhclient. The second patch addes "-v" to the dhclient arguments to improve debugging info. The dhclient into will be in /var/log/messages the Secure Contrainer host system and not in the container itself. Gene Czarcinski (2): v1.1 for dhclient use g_spawn_sync() v1.1 add -v to dhclient parameter arguments libvirt-sandbox/libvirt-sandbox-init-common.c | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) -- 1.9.3

3 5

[libvirt] [PATCH] Renamed internal __mon_yday into __vir_mon_yday to avoid conflicts
by Cédric Bosdonnat 28 Oct '14

28 Oct '14

libc has another constant with the same name, which leads to redefinition error when building against static libvirt --- src/util/virtime.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/src/util/virtime.c b/src/util/virtime.c index acbec41..7b3ec44 100644 --- a/src/util/virtime.c +++ b/src/util/virtime.c @@ -105,7 +105,7 @@ int virTimeFieldsNowRaw(struct tm *fields) #define DIV(a, b) ((a) / (b) - ((a) % (b) < 0)) #define LEAPS_THRU_END_OF(y) (DIV (y, 4) - DIV (y, 100) + DIV (y, 400)) -const unsigned short int __mon_yday[2][13] = { +const unsigned short int __vir_mon_yday[2][13] = { /* Normal years. */ { 0, 31, 59, 90, 120, 151, 181, 212, 243, 273, 304, 334, 365 }, /* Leap years. */ @@ -160,7 +160,7 @@ void virTimeFieldsThen(unsigned long long when, struct tm *fields) fields->tm_year = y - 1900; fields->tm_yday = days; - ip = __mon_yday[is_leap_year(y)]; + ip = __vir_mon_yday[is_leap_year(y)]; for (y = 11; days < (long int) ip[y]; --y) continue; days -= ip[y]; -- 1.8.4.5

3 2

[libvirt] [PATCH] qemu_agent: Produce more readable error messages
by Michal Privoznik 28 Oct '14

28 Oct '14

Not every error message from qemu-ga has to have the 'class' field filled out. For instance, I've seen this error message lately: qemuAgentCheckError:1047 : unable to execute QEMU agent command \ {"execute":"guest-set-time"}: \ {"error":{"desc":"Invalid parameter type, expected: integer"}} However, this got translated into rather generic error message: internal error: unable to execute QEMU agent command 'guest-set-time': unknown QEMU command error So we've dropped better error message in favor of a generic one. This is due to our code which expects 'class' which is not present here. Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com> --- src/qemu/qemu_agent.c | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) diff --git a/src/qemu/qemu_agent.c b/src/qemu/qemu_agent.c index fe38f6d..9f673bf 100644 --- a/src/qemu/qemu_agent.c +++ b/src/qemu/qemu_agent.c @@ -1009,14 +1009,12 @@ static const char * qemuAgentStringifyError(virJSONValuePtr error) { const char *klass = virJSONValueObjectGetString(error, "class"); - const char *detail = NULL; + const char *detail = virJSONValueObjectGetString(error, "desc"); /* The QMP 'desc' field is usually sufficient for our generic - * error reporting needs. + * error reporting needs. However, if not present, translate + * the class into something readable. */ - if (klass) - detail = virJSONValueObjectGetString(error, "desc"); - if (!detail) detail = qemuAgentStringifyErrorClass(klass); -- 2.0.4

2 2

[libvirt] [PATCH python v2] fix libvirt headers list
by Dmitry Guryanov 28 Oct '14

28 Oct '14

Since libvirt.h has been split out, generator.py should be fixed accordingly. So add full list of header files. Signed-off-by: Dmitry Guryanov <dguryanov(a)parallels.com> --- Changes in v2: * fix list format generator.py | 26 ++++++++++++++++++-------- 1 file changed, 18 insertions(+), 8 deletions(-) diff --git a/generator.py b/generator.py index 01ab441..c66c6d4 100755 --- a/generator.py +++ b/generator.py @@ -29,6 +29,21 @@ import xml.sax debug = 0 onlyOverrides = False +libvirt_headers = [ + "libvirt", + "libvirt-domain", + "libvirt-domain-snapshot", + "libvirt-event", + "libvirt-host", + "libvirt-interface", + "libvirt-network", + "libvirt-nodedev", + "libvirt-nwfilter", + "libvirt-secret", + "libvirt-storage", + "libvirt-stream", +] + def getparser(): # Attach parser to an unmarshalling object. return both objects. target = docParser() @@ -112,9 +127,7 @@ class docParser(xml.sax.handler.ContentHandler): self.function_return_field = attrs['field'] elif tag == 'enum': # enums come from header files, hence virterror.h - if (attrs['file'] == "libvirt" or - attrs['file'] == "virterror" or - attrs['file'] == "virerror"): + if attrs['file'] in libvirt_headers + ["virerror", "virterror"]: enum(attrs['type'],attrs['name'],attrs['value']) elif attrs['file'] == "libvirt-lxc": lxc_enum(attrs['type'],attrs['name'],attrs['value']) @@ -127,11 +140,8 @@ class docParser(xml.sax.handler.ContentHandler): if tag == 'function': # fuctions come from source files, hence 'virerror.c' if self.function is not None: - if (self.function_module == "libvirt" or - self.function_module == "event" or - self.function_module == "virevent" or - self.function_module == "virerror" or - self.function_module == "virterror"): + if self.function_module in libvirt_headers + \ + ["event", "virevent", "virerror", "virterror"]: function(self.function, self.function_descr, self.function_return, self.function_args, self.function_file, self.function_module, -- 1.9.3

2 1

[libvirt] [PATCH python] fix libvirt headers list
by Dmitry Guryanov 28 Oct '14

28 Oct '14

Since libvirt.h has been splitted out, generator.py should be fixed accordingly. So add full list of header files. Signed-off-by: Dmitry Guryanov <dguryanov(a)parallels.com> --- generator.py | 24 ++++++++++++++++-------- 1 file changed, 16 insertions(+), 8 deletions(-) diff --git a/generator.py b/generator.py index 01ab441..5ecbed5 100755 --- a/generator.py +++ b/generator.py @@ -29,6 +29,19 @@ import xml.sax debug = 0 onlyOverrides = False +libvirt_headers = ["libvirt", + "libvirt-domain", + "libvirt-domain-snapshot", + "libvirt-event", + "libvirt-host", + "libvirt-interface", + "libvirt-network", + "libvirt-nodedev", + "libvirt-nwfilter", + "libvirt-secret", + "libvirt-storage", + "libvirt-stream",] + def getparser(): # Attach parser to an unmarshalling object. return both objects. target = docParser() @@ -112,9 +125,7 @@ class docParser(xml.sax.handler.ContentHandler): self.function_return_field = attrs['field'] elif tag == 'enum': # enums come from header files, hence virterror.h - if (attrs['file'] == "libvirt" or - attrs['file'] == "virterror" or - attrs['file'] == "virerror"): + if attrs['file'] in libvirt_headers + ["virerror", "virterror"]: enum(attrs['type'],attrs['name'],attrs['value']) elif attrs['file'] == "libvirt-lxc": lxc_enum(attrs['type'],attrs['name'],attrs['value']) @@ -127,11 +138,8 @@ class docParser(xml.sax.handler.ContentHandler): if tag == 'function': # fuctions come from source files, hence 'virerror.c' if self.function is not None: - if (self.function_module == "libvirt" or - self.function_module == "event" or - self.function_module == "virevent" or - self.function_module == "virerror" or - self.function_module == "virterror"): + if self.function_module in libvirt_headers + \ + ["event", "virevent", "virerror", "virterror"]: function(self.function, self.function_descr, self.function_return, self.function_args, self.function_file, self.function_module, -- 1.9.3

2 3

[libvirt] [PATCH 00/11] RFC: refactor internal driver architecture
by Daniel P. Berrange 28 Oct '14

28 Oct '14

Currently we have a bit of a craz setup for activating drivers for virConnectPtr. We iterate over the primary hypervisor drivers until one matches and activates. This bit is fine. Then we do the same for the network, storage, interface, etc secondary drivers. This is where the fun starts. In the case of the stateless drivers that run outside of libvirtd, they always want the secondary drivers to match the primary driver. Unfortunately if they don't register any dummy secondary driver they'll get given the reomte driver instead. The probing of secondary drivers is really unhelpful here. In the case of stateful drivers that run inside of libvirtd, there is (almost) always only one driver instance that they want to work with. So at best the probing of secondary drivers is a waste of time. With the increasing feature set though we're getting tight dependancies from the hypervisor driver to the secondary driver. eg QEMU depends on the specific network driver we have because it calls various internal APIs it exposes. We also have circular dependancy problems during libvirtd startup: https://www.redhat.com/archives/libvir-list/2014-January/msg01143.html There the storage driver needs a virConnectPtr instance in order to talk to the secret driver. At the same time the storage driver must run before the hypervisor driver because the hypervisor driver needs to use it for autostart of VMs. This is a horrible mess. The solution I think is to remove the concept of probing for secondary drivers entirely. Instead I suggest creating a struct virDriver { ... } which contains a pointer to virHypervisorDriver, virNetworkDriver, virInterfaceDriver, etc. now virConnectPtr will only reference the single virDriverPtr and when we open a hypervisor driver, we immediately get a full set of APIs for everything. This way all the hypervisor drivers will always know exactly what secondary driver they are working with. We'll be able to remove all the xxxPrivateData elements from the virConnectPtr. The stateless drivers can just access the main privateData. The stateful drivers can avoid any use of privateData at all - just access their global static variable with the driverState. This won't quite solve the circular dependancy problem in that mail thread above. By removing the directly reliance on the virConnectPtr driver table though, we'll be able to update the code more easily so it can fetch secrets directly and thus the storage driver wont need to do virConnectOpen("qemu:///system"). The patches that follow illustrate the first part of cleaning up use of the xxxPrivateData fields. This isn't complete but I wanted to show the direction it would go before I spend the time todo the full job. Incidentally our code is buggy as hell when it comes to using the xxxPrivateData fields. The PHyp driver actually abuses the networkPrivateData field for all its APIs. Alot of code wil directly access the global state driver state already ignoring the privateData fields already. So they are really more pain than they are worth and leading to buggy code. Daniel P. Berrange (11): Clean up remote driver connection open code Update remote driver to always use privateData Update ESX driver to always use privateData Update Hyper-V driver to always use privateData Move phyp internal info out of the header file Remove abuse of networkPrivateData in phyp driver Update Parallels driver to always use privateData Update Test driver to always use privateData Remove use of storagePrivateData from storage driver Remove use of networkPrivateData from network driver Remove use of networkPrivateData from netcf driver src/esx/esx_device_monitor.c | 6 +- src/esx/esx_interface_driver.c | 16 +- src/esx/esx_network_driver.c | 20 +- src/esx/esx_nwfilter_driver.c | 6 +- src/esx/esx_secret_driver.c | 6 +- src/esx/esx_storage_backend_iscsi.c | 26 +- src/esx/esx_storage_backend_vmfs.c | 36 +-- src/esx/esx_storage_driver.c | 44 ++-- src/hyperv/hyperv_device_monitor.c | 6 +- src/hyperv/hyperv_interface_driver.c | 6 +- src/hyperv/hyperv_network_driver.c | 6 +- src/hyperv/hyperv_nwfilter_driver.c | 6 +- src/hyperv/hyperv_secret_driver.c | 6 +- src/hyperv/hyperv_storage_driver.c | 6 +- src/interface/interface_backend_netcf.c | 71 ++---- src/network/bridge_driver.c | 435 ++++++++++++++------------------ src/parallels/parallels_network.c | 2 - src/parallels/parallels_storage.c | 12 +- src/parallels/parallels_utils.h | 1 + src/phyp/phyp_driver.c | 225 ++++++++--------- src/phyp/phyp_driver.h | 53 ---- src/remote/remote_driver.c | 153 +++-------- src/rpc/gendispatch.pl | 28 +- src/storage/storage_driver.c | 194 +++++++------- src/test/test_driver.c | 24 +- 25 files changed, 537 insertions(+), 857 deletions(-) -- 2.1.0

3 24