- Devel - libvirt List Archives

[libvirt] [PATCH] rpc: wait longer for session daemon to start up
by Cole Robinson 17 Mar '16

17 Mar '16

https://bugzilla.redhat.com/show_bug.cgi?id=1271183 We only wait .5 seconds for the session daemon to start up and present its socket, which isn't sufficient for many users. Bump up the sleep interval and retry amount so we wait for a total of 5 seconds. --- danpb suggests dropping the reverting this: commit be78814ae07f092d9c4e71fd82dd1947aba2f029 Author: Michal Privoznik <mprivozn(a)redhat.com> Date: Thu Apr 2 14:41:17 2015 +0200 virNetSocketNewConnectUNIX: Use flocks when spawning a daemon Prior to that we didn't need the retry logic at all... but that's a bit more involved and boxes users are suffering with this issue in the meantime src/rpc/virnetsocket.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/src/rpc/virnetsocket.c b/src/rpc/virnetsocket.c index b0d5b1c..d909b94 100644 --- a/src/rpc/virnetsocket.c +++ b/src/rpc/virnetsocket.c @@ -614,7 +614,7 @@ int virNetSocketNewConnectUNIX(const char *path, char *lockpath = NULL; int lockfd = -1; int fd = -1; - int retries = 100; + int retries = 500; virSocketAddr localAddr; virSocketAddr remoteAddr; char *rundir = NULL; @@ -707,7 +707,7 @@ int virNetSocketNewConnectUNIX(const char *path, daemonLaunched = true; } - usleep(5000); + usleep(10000); } localAddr.len = sizeof(localAddr.data); -- 2.5.0

4 4

[libvirt] [PATCH] virlog: Fix build breaker with "comparison between signed and unsigned"
by Erik Skultety 16 Mar '16

16 Mar '16

Refactor series 0b231195 worked with virLogDestination type which, depending on the compiler, might be (and probably will be) an unsigned data type. However, virEnumFromString may return -1 in case of an error. So, when enum happens to be unsigned, some compilers will naturally complain about foo: 'if (foo < 0)' --- I pushed the patch under build breaker rule. src/util/virlog.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/util/virlog.c b/src/util/virlog.c index 591d38e..007fc65 100644 --- a/src/util/virlog.c +++ b/src/util/virlog.c @@ -1088,7 +1088,7 @@ virLogParseOutput(const char *src) char *abspath = NULL; size_t count = 0; virLogPriority prio; - virLogDestination dest; + int dest; bool isSUID = virIsSUID(); if (!src) -- 2.4.3

1 0

[libvirt] [PATCH v2 0/7] vz: add disk and controller check in domainPostParse phase
by Maxim Nestratov 16 Mar '16

16 Mar '16

Changes from v1 =============== A new patch moving prlsdkCheckDiskUnsupportedParams to vz_utils.c added. Commit messages reworded. Minor formatting issues fixed. Maxim Nestratov (1): vz: move prlsdkCheckDiskUnsupportedParams to vz_utils.c Mikhail Feoktistov (6): vz: save vz version in connection structure vz: add vzCapabilities to connection structure vz: check supported disk format and bus vz: report correct disk format in domainGetXMLDesc vz: check supported controllers vz: set default SCSI model src/vz/vz_driver.c | 61 +++-------- src/vz/vz_sdk.c | 203 +++++++--------------------------- src/vz/vz_sdk.h | 2 +- src/vz/vz_utils.c | 314 +++++++++++++++++++++++++++++++++++++++++++++++++++++ src/vz/vz_utils.h | 24 ++++ 5 files changed, 397 insertions(+), 207 deletions(-) -- 2.4.3

2 9

[libvirt] Where does 'rx_drop' in 'domifstat' come from?
by Yaniv Kaul 16 Mar '16

16 Mar '16

Any idea why I'd see drops on an interface? mini@ykaul-mini:~/ovirt-system-tests$ sudo virsh domifstat 111 vnet1 vnet1 rx_bytes 25488 vnet1 rx_packets 387 vnet1 rx_errs 0 *vnet1 rx_drop 1424* vnet1 tx_bytes 5751 vnet1 tx_packets 39 vnet1 tx_errs 0 vnet1 tx_drop 0 I have several others VMs in the same host, pretty much identically configured, all are fine, all on the same bridge: mini@ykaul-mini:~/ovirt-system-tests$ sudo brctl show bridge name bridge id STP enabled interfaces 5e85-930e31b 8000.525400247745 yes 5e85-93031b-nic vnet0 vnet1 vnet2 vnet3 vnet4 virbr0 8000.5254005e7e4b yes virbr0-nic Network (host0 is the problematic one): virsh # net-dumpxml 5e85-930e31b9b9 <network connections='5'> <name>5e85-930e31b9b9</name> <uuid>6f52f5a8-5322-4b57-890e-d4d5a0a7ed50</uuid> <forward mode='nat'> <nat> <port start='1024' end='65535'/> </nat> </forward> <bridge name='5e85-930e31b' stp='on' delay='0'/> <mac address='52:54:00:24:77:45'/> <dns forwardPlainNames='yes'/> <ip address='192.168.200.1' netmask='255.255.255.0'> <dhcp> <range start='192.168.200.100' end='192.168.200.254'/> <host mac='54:52:c0:a8:c8:04' name='lago_basic_suite_3_6_host1' ip='192.168.200.4'/> <host mac='54:52:c0:a8:c8:03' name='lago_basic_suite_3_6_engine' ip='192.168.200.3'/> <host mac='54:52:c0:a8:c8:02' name='lago_basic_suite_3_6_storage-iscsi' ip='192.168.200.2'/> <host mac='54:52:c0:a8:c8:05' name='lago_basic_suite_3_6_storage-nfs' ip='192.168.200.5'/> <host mac='54:52:c0:a8:c8:06' name='lago_basic_suite_3_6_host0' ip='192.168.200.6'/> </dhcp> </ip> </network> Domain XML attached. Running libvirt 1.2.18, on latest F23. TIA, Y.

1 0

[libvirt] [PATCH 0/6] vz: add disk and controller check in domainPostParse phase
by Mikhail Feoktistov 16 Mar '16

16 Mar '16

Mikhail Feoktistov (6): vz: save vz version in connection structure vz: add vzCapabilities to connection structure vz: check supported disk format and bus vz: report correct disk format in domainGetXMLDesc vz: check supported controllers vz: set default SCSI model src/vz/vz_driver.c | 61 +++++------------ src/vz/vz_sdk.c | 80 +++++++++++----------- src/vz/vz_sdk.h | 2 +- src/vz/vz_utils.c | 191 +++++++++++++++++++++++++++++++++++++++++++++++++++++ src/vz/vz_utils.h | 24 +++++++ 5 files changed, 274 insertions(+), 84 deletions(-) -- 1.8.3.1

2 9

[libvirt] [Issue]: Regarding client socket getting closed from the server once the lxc container is started
by Rammohan Gowrimoole (rgowrimo) 16 Mar '16

16 Mar '16

Hi Folks, Using libvirt python bindings we are creating an lxc container.Here is the problem that we see sometimes (say 20 % of the time) when we create a new container. 1.container gets created, and also starts.However the we are not able to enter the namepace of the container.It throws an error initPid not available.We see that the using netstat command , socket connection is closed. 2.To get around this problem we have to stop and start the container again.We see that socket under (/var/run/libvirt/*) connection is established and we are able to enter the namespace. Enabled the libvirtd debug logs to debug this issue. For success case we see that for new client connection gets created and is able to handle async incoming events, 2016-03-12 08:18:55.748+0000: 1247: info : virObjectNew:202 : OBJECT_NEW: obj=0x7fed54005460 classname=virLXCMonitor 2016-03-12 08:18:55.748+0000: 1247: debug : virNetSocketNew:159 : localAddr=0x7fed7cd1d170 remoteAddr=0x7fed7cd1d200 fd=28 errfd=-1 pid=0 2016-03-12 08:18:55.749+0000: 1247: info : virObjectNew:202 : OBJECT_NEW: obj=0x7fed54009040 classname=virNetSocket 2016-03-12 08:18:55.749+0000: 1247: info : virNetSocketNew:209 : RPC_SOCKET_NEW: sock=0x7fed54009040 fd=28 errfd=-1 pid=0 localAddr=127.0.0.1;0, remoteAddr=127.0.0.1;0 2016-03-12 08:18:55.749+0000: 1247: info : virObjectNew:202 : OBJECT_NEW: obj=0x7fed54009d10 classname=virNetClient 2016-03-12 08:18:55.749+0000: 1247: info : virNetClientNew:327 : RPC_CLIENT_NEW: client=0x7fed54009d10 sock=0x7fed54009040 2016-03-12 08:18:55.749+0000: 1247: info : virObjectRef:296 : OBJECT_REF: obj=0x7fed54009d10 2016-03-12 08:18:55.749+0000: 1247: info : virObjectRef:296 : OBJECT_REF: obj=0x7fed54009040 2016-03-12 08:18:55.750+0000: 1247: info : virObjectNew:202 : OBJECT_NEW: obj=0x7fed540009a0 classname=virNetClientProgram 2016-03-12 08:18:55.750+0000: 1247: info : virObjectRef:296 : OBJECT_REF: obj=0x7fed540009a0 2016-03-12 08:18:55.750+0000: 1247: info : virObjectRef:296 : OBJECT_REF: obj=0x7fed54005460 2016-03-12 08:18:55.750+0000: 1247: info : virObjectUnref:259 : OBJECT_UNREF: obj=0x7fed5c168eb0 2016-03-12 08:18:55.750+0000: 1247: debug : virLXCProcessCleanInterfaces:475 : Cleared net names: eth0 2016-03-12 08:18:55.750+0000: 1247: info : virObjectUnref:259 : OBJECT_UNREF: obj=0x7fed5c168eb0 2016-03-12 08:18:55.750+0000: 1247: info : virObjectUnref:259 : OBJECT_UNREF: obj=0x7fed5c169600 2016-03-12 08:18:55.755+0000: 1244: debug : virNetClientIncomingEvent:1808 : client=0x7fed54009d10 wantclose=0 2016-03-12 08:18:55.755+0000: 1244: debug : virNetClientIncomingEvent:1816 : Event fired 0x7fed54009040 1 2016-03-12 08:18:55.755+0000: 1244: debug : virNetMessageDecodeLength:151 : Got length, now need 36 total (32 more) 2016-03-12 08:18:55.756+0000: 1244: info : virNetClientCallDispatch:1116 : RPC_CLIENT_MSG_RX: client=0x7fed54009d10 len=36 prog=305402420 vers=1 proc=2 type=2 status=0 serial=1 2016-03-12 08:18:55.756+0000: 1244: debug : virKeepAliveCheckMessage:377 : ka=(nil), client=0x7fed81fc5ed4, msg=0x7fed54009d78 2016-03-12 08:18:55.756+0000: 1244: debug : virNetClientProgramDispatch:220 : prog=305402420 ver=1 type=2 status=0 serial=1 proc=2 2016-03-12 08:18:55.756+0000: 1244: debug : virLXCMonitorHandleEventInit:109 : Event init 1420 For failure case ,we see that the client socket connection is initiated and gets closed immediately after receiving an incoming event.In this case, I don’t see an object for virNetClientProgram being created. Incoming event comes in and since the its unable to find client->prog it bails out and closes the connection. Snaphost of the code, static int virNetClientCallDispatchMessage(virNetClientPtr client) { size_t i; virNetClientProgramPtr prog = NULL; for (i = 0; i < client->nprograms; i++) { if (virNetClientProgramMatches(client->programs[i], &client->msg)) { prog = client->programs[i]; break; } } if (!prog) { VIR_DEBUG("No program found for event with prog=%d vers=%d", client->msg.header.prog, client->msg.header.vers); return -1; } 2016-03-12 08:19:15.935+0000: 1246: info : virObjectRef:296 : OBJECT_REF: obj=0x7fed5c168eb0 2016-03-12 08:19:15.935+0000: 1246: info : virObjectRef:296 : OBJECT_REF: obj=0x7fed82bd7bc0 2016-03-12 08:19:15.935+0000: 1246: info : virObjectNew:202 : OBJECT_NEW: obj=0x7fed82bd8120 classname=virLXCMonitor 2016-03-12 08:19:15.935+0000: 1246: debug : virNetSocketNew:159 : localAddr=0x7fed7d51e170 remoteAddr=0x7fed7d51e200 fd=31 errfd=-1 pid=0 2016-03-12 08:19:15.936+0000: 1246: info : virObjectNew:202 : OBJECT_NEW: obj=0x7fed82bd8660 classname=virNetSocket 2016-03-12 08:19:15.936+0000: 1246: info : virNetSocketNew:209 : RPC_SOCKET_NEW: sock=0x7fed82bd8660 fd=31 errfd=-1 pid=0 localAddr=127.0.0.1;0, remoteAddr=127.0.0.1;0 2016-03-12 08:19:15.936+0000: 1246: info : virObjectNew:202 : OBJECT_NEW: obj=0x7fed82bd8ca0 classname=virNetClient 2016-03-12 08:19:15.936+0000: 1246: info : virNetClientNew:327 : RPC_CLIENT_NEW: client=0x7fed82bd8ca0 sock=0x7fed82bd8660 2016-03-12 08:19:15.936+0000: 1246: info : virObjectRef:296 : OBJECT_REF: obj=0x7fed82bd8ca0 2016-03-12 08:19:15.936+0000: 1246: info : virObjectRef:296 : OBJECT_REF: obj=0x7fed82bd8660 2016-03-12 08:19:15.942+0000: 1244: debug : virNetClientIncomingEvent:1808 : client=0x7fed82bd8ca0 wantclose=0 2016-03-12 08:19:15.942+0000: 1244: debug : virNetClientIncomingEvent:1816 : Event fired 0x7fed82bd8660 1 2016-03-12 08:19:15.942+0000: 1244: debug : virNetMessageDecodeLength:151 : Got length, now need 36 total (32 more) 2016-03-12 08:19:15.942+0000: 1244: info : virNetClientCallDispatch:1116 : RPC_CLIENT_MSG_RX: client=0x7fed82bd8ca0 len=36 prog=305402420 vers=1 proc=2 type=2 status=0 serial=1 2016-03-12 08:19:15.942+0000: 1244: debug : virKeepAliveCheckMessage:377 : ka=(nil), client=0x7fed81fc5ed4, msg=0x7fed82bd8d08 2016-03-12 08:19:15.942+0000: 1244: debug : virNetClientCallDispatchMessage:1008 : No program found for event with prog=305402420 vers=1 2016-03-12 08:19:15.942+0000: 1244: debug : virNetMessageClear:57 : msg=0x7fed82bd8d08 nfds=0 2016-03-12 08:19:15.942+0000: 1244: debug : virNetClientMarkClose:632 : client=0x7fed82bd8ca0, reason=0 2016-03-12 08:19:15.942+0000: 1244: debug : virNetClientCloseLocked:648 : client=0x7fed82bd8ca0, sock=0x7fed82bd8660, reason=0 Here is the snapshot of code , virLXCMonitorPtr virLXCMonitorNew(virDomainObjPtr vm, const char *socketdir, virLXCMonitorCallbacksPtr cb) { virLXCMonitorPtr mon; char *sockpath = NULL; if (virLXCMonitorInitialize() < 0) return NULL; if (!(mon = virObjectLockableNew(virLXCMonitorClass))) return NULL; if (virAsprintf(&sockpath, "%s/%s.sock", socketdir, vm->def->name) < 0) goto error; if (!(mon->client = virNetClientNewUNIX(sockpath, false, NULL))) goto error; if (virNetClientRegisterAsyncIO(mon->client) < 0) goto error; if (!(mon->program = virNetClientProgramNew(VIR_LXC_MONITOR_PROGRAM, VIR_LXC_MONITOR_PROGRAM_VERSION, virLXCMonitorEvents, ARRAY_CARDINALITY(virLXCMonitorEvents), mon))) goto error; if (virNetClientAddProgram(mon->client, mon->program) < 0) goto error; mon->vm = vm; memcpy(&mon->cb, cb, sizeof(mon->cb)); virObjectRef(mon); virNetClientSetCloseCallback(mon->client, virLXCMonitorEOFNotify, mon, virLXCMonitorCloseFreeCallback); Is the problem occurring due to invocation of “virNetClientRegisterAsyncIO" api before the virNetClientAddProgram.Probably once we register for aysnc IO , immediately an event comes in and that thread takes priority and bails out since it does not find the client->prog?Also the client is not retrying to establish a new connection. Please let me any thoughts/comments.Is there any patch already available which has fixed this issue? -Thanks, Rammohan

1 0

[libvirt] [PATCH 00/21] Various (tiny) clean-ups and changes for daemon/admin
by Martin Kletzander 16 Mar '16

16 Mar '16

I wanted to make virAdmConnectListAllServers()'s dispatch function generated by our mighty gendispatch script. And when doing that I stumbled upon various other things, changed what seemed better and at the end I ended up with one huge commit. Because lot of those changes should not be done together in one commit, I split each tiny part so it's nicer to review. However, some might point out that some commits do not belong together. To that I say it was a pain splitting the commit, don't make me split the series as well, please =) This "conflicts" with Erik's series that deals with worker threads tuning. We about know that and we will merge those together, or rather rebase on top of each other based on which version of virAdmConnectLookupServer{,ByName}() you and we like better. Martin Kletzander (21): admin: Make virAdmServerFree() handle NULL gracefully admin: Check for flags properly server: Store server name in server object daemon: Get server name from the server itself virerror: Introduce new error type NO_SERVER daemon: Set error for unknown server name daemon: Properly check for clients Expose virNetServerGetName admin: Do not work with virAdm on the server side Change virNetDaemonGetServerNames to virNetDaemonGetServers admin: Don't use priority for admin APIs virt-admin: Don't leak uri all over the place admin: Be consistent when resetting errors gendispatch: Cluster, don't capture if not needed gendispatch: Be able to generate multi-return values admin: Generate ConnectListServers dispatch helpers gendispatch: Accept server as an argument admin: Add virAdmConnectLookupServer gendispatch: Remember the name of snapshot variable name gendispatch: Support modern listing of more types remote: Generate what's possible daemon/admin.c | 49 +-- daemon/admin_server.c | 36 +-- daemon/admin_server.h | 12 +- daemon/libvirtd.c | 10 +- daemon/remote.c | 599 ------------------------------------ include/libvirt/libvirt-admin.h | 4 + include/libvirt/virterror.h | 1 + po/POTFILES.in | 1 - src/admin/admin_protocol.x | 21 +- src/admin/admin_remote.c | 63 ---- src/admin_protocol-structs | 8 + src/libvirt-admin.c | 48 ++- src/libvirt_admin_private.syms | 2 + src/libvirt_admin_public.syms | 1 + src/libvirt_remote.syms | 2 + src/locking/lock_daemon.c | 5 +- src/logging/log_daemon.c | 5 +- src/lxc/lxc_controller.c | 5 +- src/remote/remote_driver.c | 654 ---------------------------------------- src/remote/remote_protocol.x | 40 +-- src/rpc/gendispatch.pl | 374 +++++++++++++++++------ src/rpc/virnetdaemon.c | 65 ++-- src/rpc/virnetdaemon.h | 3 +- src/rpc/virnetserver.c | 20 +- src/rpc/virnetserver.h | 6 +- src/util/virerror.c | 6 + tests/virnetdaemontest.c | 10 +- tools/virt-admin.c | 8 +- 28 files changed, 509 insertions(+), 1549 deletions(-) -- 2.7.2

5 55

[libvirt] [PATCH] conf: nodedev: Remove inverted match logic
by Andrea Bolognani 16 Mar '16

16 Mar '16

Commit c1276177f832 simplified the match logic; this commit gets rid of the inverted logic to make the code even more straightforward. --- src/conf/node_device_conf.c | 28 ++++++++++++++-------------- 1 file changed, 14 insertions(+), 14 deletions(-) diff --git a/src/conf/node_device_conf.c b/src/conf/node_device_conf.c index 0c9c348..b9b9dfa 100644 --- a/src/conf/node_device_conf.c +++ b/src/conf/node_device_conf.c @@ -1806,22 +1806,22 @@ virNodeDeviceMatch(virNodeDeviceObjPtr devobj, { /* filter by cap type */ if (flags & VIR_CONNECT_LIST_NODE_DEVICES_FILTERS_CAP) { - if (!(MATCH(SYSTEM) || - MATCH(PCI_DEV) || - MATCH(USB_DEV) || - MATCH(USB_INTERFACE) || - MATCH(NET) || - MATCH(SCSI_HOST) || - MATCH(SCSI_TARGET) || - MATCH(SCSI) || - MATCH(STORAGE) || - MATCH(FC_HOST) || - MATCH(VPORTS) || - MATCH(SCSI_GENERIC))) - return false; + if (MATCH(SYSTEM) || + MATCH(PCI_DEV) || + MATCH(USB_DEV) || + MATCH(USB_INTERFACE) || + MATCH(NET) || + MATCH(SCSI_HOST) || + MATCH(SCSI_TARGET) || + MATCH(SCSI) || + MATCH(STORAGE) || + MATCH(FC_HOST) || + MATCH(VPORTS) || + MATCH(SCSI_GENERIC)) + return true; } - return true; + return false; } #undef MATCH -- 2.5.0

2 2

[libvirt] [PATCH] qemu migration: assign addresses in xml passed as parameter
by Nikolay Shirokovskiy 16 Mar '16

16 Mar '16

Usecase: 1. define domain with xml A where addresses are not set 2. migrate with '--xml A' You get ABI incompatibility because we comprare xmls with addresses set and not. This patch assign addresses in parameter xml so the comprasion is fair. Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy(a)virtuozzo.com> --- src/qemu/qemu_migration.c | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/src/qemu/qemu_migration.c b/src/qemu/qemu_migration.c index 1443d96..7cceaa1 100644 --- a/src/qemu/qemu_migration.c +++ b/src/qemu/qemu_migration.c @@ -2992,6 +2992,7 @@ qemuMigrationBeginPhase(virQEMUDriverPtr driver, virDomainDefPtr def = NULL; qemuDomainObjPrivatePtr priv = vm->privateData; virCapsPtr caps = NULL; + virQEMUCapsPtr qemuCaps = NULL; unsigned int cookieFlags = QEMU_MIGRATION_COOKIE_LOCKSTATE; VIR_DEBUG("driver=%p, vm=%p, xmlin=%s, dname=%s," @@ -3105,6 +3106,13 @@ qemuMigrationBeginPhase(virQEMUDriverPtr driver, VIR_DOMAIN_DEF_PARSE_INACTIVE))) goto cleanup; + if (!(qemuCaps = virQEMUCapsCacheLookup(driver->qemuCapsCache, + def->emulator))) + goto cleanup; + + if (qemuDomainAssignAddresses(def, qemuCaps, NULL) < 0) + goto cleanup; + if (!qemuDomainDefCheckABIStability(driver, vm->def, def)) goto cleanup; @@ -3117,6 +3125,7 @@ qemuMigrationBeginPhase(virQEMUDriverPtr driver, qemuMigrationCookieFree(mig); virObjectUnref(caps); virDomainDefFree(def); + virObjectUnref(qemuCaps); return rv; } -- 1.8.3.1

2 2

[libvirt] [PATCH] qemu: Don't access uninitialized memory
by Martin Kletzander 16 Mar '16

16 Mar '16

In qemuConnectDomainXMLToNative() we set up the monitor, but we never memset() it to zeros. Thanks to the introduction of the logfile parameter of chardevs (and the logfile member of the struct), we started checking whether that's non-NULL and that exposed this old error. Signed-off-by: Martin Kletzander <mkletzan(a)redhat.com> --- src/qemu/qemu_driver.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index a0d6596aef7b..e868b9f899e7 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -6950,7 +6950,7 @@ static char *qemuConnectDomainXMLToNative(virConnectPtr conn, { virQEMUDriverPtr driver = conn->privateData; virDomainDefPtr def = NULL; - virDomainChrSourceDef monConfig; + virDomainChrSourceDef monConfig = {0}; virQEMUCapsPtr qemuCaps = NULL; bool monitor_json = false; virCommandPtr cmd = NULL; -- 2.7.3

2 2