- Devel - libvirt List Archives

[libvirt] [PATCHv3 0/4] Various additions to libxl driver
by Markus Groß 31 May '11

31 May '11

This series adds additional functionality to the libxl driver. The detailed changes are noted in each individual patch. The second patch adds core dumping and requires at least Xen 4.1.1, because of this patch for xen: http://lists.xensource.com/archives/html/xen-devel/2011-05/msg01469.html The save/restore functionality requires Xen 4.2 since these patches are needed: http://lists.xensource.com/archives/html/xen-devel/2011-05/msg01314.html http://lists.xensource.com/archives/html/xen-devel/2011-05/msg01528.html To set the maximum memory of a domain this patch from Xen 4.2 is required: http://lists.xensource.com/archives/html/xen-devel/2011-05/msg01717.html Markus Groß (4): Populate domid field of devices for libxenlight Add domainCoreDump to libxl driver Add domainSave/Restore to libxl driver Support maximum and current memory flags in libxl driver src/libxl/libxl_conf.c | 20 ++- src/libxl/libxl_conf.h | 24 ++- src/libxl/libxl_driver.c | 431 +++++++++++++++++++++++++++++++++++++++++----- 3 files changed, 418 insertions(+), 57 deletions(-) -- 1.7.5.2

3 10

[libvirt] ssh from host to guest on default network
by Kay Williams 31 May '11

31 May '11

Hello, I am looking for an reliable, scriptable method to 1) install a kvm guest using the default network and 2) ssh from the host to the guest to perform machine maintenance (e.g. yum update). Ideally, this would be done by providing a host name for the guest during install (e.g. virt-install --name <myname> ...), and then by ssh'ing to the machine using the provided hostname (e.g. ssh <myname>) and a username/password or preconfigured certificate. The trouble I am having is with ssh'ing from the host to guest machine. 1. I can't (easily) use the guest hostname, as the host isn't automatically configured to resolve guest names. 2. I can't (easily) use the guest IP address (e.g. read from /var/lib/libvirt/dnsmasq/default.leases), as the IP addresses are dynamic and can change when the guest is reinstalled (which I am doing a fair amount, at least in testing). 3. I can't (easily) assign the guest a static IP address, as the default network configuration assumes all IP addresses in the range can be dynamically allocated, and thus a statically assigned IP might conflict. I can work around these issues in a variety of unfortunate ways. But I wonder if there is something elegant I am missing. If not, this would be a very nice feature for the future.... Thanks, Kay

1 0

[libvirt] [PATCH] Add support for 'passthru' mode for direct network interfaces
by D. Herrendoerfer 30 May '11

30 May '11

Hi all, starting with kernel 2.6.38 macvtap supports a 'passthru' mode for attaching virtual functions of a SRIOV capable network card directly to a VM. This patch adds the capability to configure such a device. Signed-off-by: Dirk Herrendoerfer <d.herrendoerfer at herrendoerfer.name> diff --git a/docs/schemas/domain.rng b/docs/schemas/domain.rng index 7163c6e..e2a66ae 100644 --- a/docs/schemas/domain.rng +++ b/docs/schemas/domain.rng @@ -2351,7 +2351,7 @@ </define> <define name="bridgeMode"> <data type="string"> - <param name="pattern">(vepa|bridge|private)</param> + <param name="pattern">(vepa|bridge|private|passthru)</param> </data> </define> <define name="addrMAC"> diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c index 6b733d4..104d50d 100644 --- a/src/conf/domain_conf.c +++ b/src/conf/domain_conf.c @@ -376,7 +376,8 @@ VIR_ENUM_IMPL(virDomainSeclabel, VIR_DOMAIN_SECLABEL_LAST, VIR_ENUM_IMPL(virDomainNetdevMacvtap, VIR_DOMAIN_NETDEV_MACVTAP_MODE_LAST, "vepa", "private", - "bridge") + "bridge", + "passthru" ) VIR_ENUM_IMPL(virVirtualPort, VIR_VIRTUALPORT_TYPE_LAST, "none", diff --git a/src/conf/domain_conf.h b/src/conf/domain_conf.h index 6ea30b9..5833d72 100644 --- a/src/conf/domain_conf.h +++ b/src/conf/domain_conf.h @@ -335,6 +335,7 @@ enum virDomainNetdevMacvtapType { VIR_DOMAIN_NETDEV_MACVTAP_MODE_VEPA, VIR_DOMAIN_NETDEV_MACVTAP_MODE_PRIVATE, VIR_DOMAIN_NETDEV_MACVTAP_MODE_BRIDGE, + VIR_DOMAIN_NETDEV_MACVTAP_MODE_PASSTHRU, VIR_DOMAIN_NETDEV_MACVTAP_MODE_LAST, }; diff --git a/src/util/macvtap.c b/src/util/macvtap.c index a7af0cb..1b274cc 100644 --- a/src/util/macvtap.c +++ b/src/util/macvtap.c @@ -473,6 +473,9 @@ macvtapModeFromInt(enum virDomainNetdevMacvtapType mode) case VIR_DOMAIN_NETDEV_MACVTAP_MODE_BRIDGE: return MACVLAN_MODE_BRIDGE; + case VIR_DOMAIN_NETDEV_MACVTAP_MODE_PASSTHRU: + return MACVLAN_MODE_PASSTHRU; + case VIR_DOMAIN_NETDEV_MACVTAP_MODE_VEPA: default: return MACVLAN_MODE_VEPA; @@ -546,7 +549,7 @@ configMacvtapTap(int tapfd, int vnet_hdr) * be NULL if this function is supposed to choose a name * @macaddress: The MAC address for the macvtap device * @linkdev: The interface name of the NIC to connect to the external bridge - * @mode: int describing the mode for 'bridge', 'vepa' or 'private'. + * @mode: int describing the mode for 'bridge', 'vepa', 'private' or 'passthru'. * @vnet_hdr: 1 to enable IFF_VNET_HDR, 0 to disable it * @vmuuid: The UUID of the VM the macvtap belongs to * @virtPortProfile: pointer to object holding the virtual port profile data diff --git a/src/util/macvtap.h b/src/util/macvtap.h index 54205c7..c45dd13 100644 --- a/src/util/macvtap.h +++ b/src/util/macvtap.h @@ -92,6 +92,7 @@ void delMacvtap(const char *ifname, # define MACVTAP_MODE_PRIVATE_STR "private" # define MACVTAP_MODE_VEPA_STR "vepa" # define MACVTAP_MODE_BRIDGE_STR "bridge" +# define MACVTAP_MODE_PASSTHRU_STR "passthru" int vpAssociatePortProfileId(const char *macvtap_ifname, const unsigned char *macvtap_macaddr,

5 10

[libvirt] [PATCH RFC] Add domainSave/Restore to libxl driver
by Markus Groß 30 May '11

30 May '11

This patch adds save/restore functionality to the libxl driver. It is a v2 of this patch: https://www.redhat.com/archives/libvir-list/2011-April/msg00338.html v2: * header is now padded and has a version field * the correct restore function from libxl is used * only create the restore event once in libxlVmStart However I ran into a reproducible segfault. Assume you saved a vm with: # virsh save domU foo.img If you restore the save image, destroy the vm and restore it again, a segfault occurs: # virsh restore foo.img # virsh destroy domU # virsh restore foo.img # segfault If you restart libvirt between the restores no segfault occurs: # virsh restore foo.img # virsh destroy domU # restart libvirt # virsh restore foo.img According to a gdb backtrace a memcpy from the function xc_domain_restore is causing the segfault. Then I saved a vm with the xl tool (which also uses the libxenlight interface) and restored it twice. No segfault occured. So I figured something must went wrong in libvirt. However I was unable to identify the responsible part of code. Now I am hoping that a review will possibly solve this issue. Here is the gdb backtrace: Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 0x42ba1950 (LWP 23813)] 0x00007f1bc17d906b in memcpy () from /lib/libc.so.6 (gdb) bt #0 0x00007f1bc17d906b in memcpy () from /lib/libc.so.6 #1 0x00007f1bc2b49346 in xc_domain_restore () from /usr/lib/libxenguest.so.4.0 #2 0x00007f1bc31a80f3 in ?? () from /usr/lib/libxenlight.so.1.0 #3 0x00007f1bc31a105e in ?? () from /usr/lib/libxenlight.so.1.0 #4 0x000000000049389a in libxlVmStart (driver=0x7f1bb8003e80, vm=0x7f1bb8053400, start_paused=false, restore_fd=20) at libxl/libxl_driver.c:531 #5 0x00000000004945c5 in libxlDomainRestore (conn=<value optimized out>, from=<value optimized out>) at libxl/libxl_driver.c:1745 #6 0x00007f1bc1f85a4b in virDomainRestore (conn=0x20a4d40, from=0x7f1bb804d260 "/root/tt.img") at libvirt.c:2330 #7 0x000000000042c5b8 in remoteDispatchDomainRestore (server=<value optimized out>, client=<value optimized out>, conn=0xaf0, hdr=<value optimized out>, rerr=0x42ba0e20, args=0xffffffff, ret=0x42ba0f00) at remote.c:2616 #8 0x0000000000431cb7 in remoteDispatchClientRequest (server=0x209fe90, client=0x20a4ad0, msg=0x20b1aa0) at dispatch.c:524 #9 0x000000000041dd53 in qemudWorker (data=0x20a8018) at libvirtd.c:1622 #10 0x00007f1bc1cb9fc7 in start_thread () from /lib/libpthread.so.0 #11 0x00007f1bc182b64d in clone () from /lib/libc.so.6 #12 0x0000000000000000 in ?? () Cheers, Markus --- src/libxl/libxl_conf.h | 14 +++ src/libxl/libxl_driver.c | 228 ++++++++++++++++++++++++++++++++++++++++++---- 2 files changed, 225 insertions(+), 17 deletions(-) diff --git a/src/libxl/libxl_conf.h b/src/libxl/libxl_conf.h index 8c87786..e7dd3a1 100644 --- a/src/libxl/libxl_conf.h +++ b/src/libxl/libxl_conf.h @@ -1,5 +1,6 @@ /*---------------------------------------------------------------------------*/ /* Copyright (c) 2011 SUSE LINUX Products GmbH, Nuernberg, Germany. + * Copyright (C) 2011 Univention GmbH. * * This library is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public @@ -17,6 +18,7 @@ * * Authors: * Jim Fehlig <jfehlig(a)novell.com> + * Markus Groß <gross(a)univention.de> */ /*---------------------------------------------------------------------------*/ @@ -85,6 +87,18 @@ struct _libxlDomainObjPrivate { int eventHdl; }; +#define LIBXL_SAVE_MAGIC "libvirt-xml\n \0 \r" +#define LIBXL_SAVE_VERSION 1 + +typedef struct _libxlSavefileHeader libxlSavefileHeader; +typedef libxlSavefileHeader *libxlSavefileHeaderPtr; +struct _libxlSavefileHeader { + char magic[sizeof(LIBXL_SAVE_MAGIC)-1]; + uint32_t version; + uint32_t xmlLen; + /* 24 bytes used, pad up to 64 bytes */ + uint32_t unused[10]; +}; # define libxlError(code, ...) \ virReportErrorHelper(VIR_FROM_LIBXL, code, __FILE__, \ diff --git a/src/libxl/libxl_driver.c b/src/libxl/libxl_driver.c index 247d78e..56a62c9 100644 --- a/src/libxl/libxl_driver.c +++ b/src/libxl/libxl_driver.c @@ -29,6 +29,7 @@ #include <sys/utsname.h> #include <math.h> #include <libxl.h> +#include <fcntl.h> #include "internal.h" #include "logging.h" @@ -60,11 +61,10 @@ static libxlDriverPrivatePtr libxl_driver = NULL; - /* Function declarations */ static int -libxlVmStart(libxlDriverPrivatePtr driver, - virDomainObjPtr vm, bool start_paused); +libxlVmStart(libxlDriverPrivatePtr driver, virDomainObjPtr vm, + bool start_paused, int restore_fd); /* Function definitions */ @@ -188,7 +188,7 @@ libxlAutostartDomain(void *payload, const void *name ATTRIBUTE_UNUSED, virResetLastError(); if (vm->autostart && !virDomainObjIsActive(vm) && - libxlVmStart(driver, vm, false) < 0) { + libxlVmStart(driver, vm, false, -1) < 0) { err = virGetLastError(); VIR_ERROR(_("Failed to autostart VM '%s': %s"), vm->def->name, @@ -378,7 +378,7 @@ static void libxlEventHandler(int watch, break; case SHUTDOWN_reboot: libxlVmReap(driver, vm, 0); - libxlVmStart(driver, vm, 0); + libxlVmStart(driver, vm, 0, -1); break; default: VIR_INFO("Unhandled shutdown_reason %d", info.shutdown_reason); @@ -504,8 +504,8 @@ cleanup: * virDomainObjPtr should be locked on invocation */ static int -libxlVmStart(libxlDriverPrivatePtr driver, - virDomainObjPtr vm, bool start_paused) +libxlVmStart(libxlDriverPrivatePtr driver, virDomainObjPtr vm, + bool start_paused, int restore_fd) { libxl_domain_config d_config; virDomainDefPtr def = vm->def; @@ -524,12 +524,23 @@ libxlVmStart(libxlDriverPrivatePtr driver, //TODO: Balloon dom0 ?? //ret = freemem(&d_config->b_info, &d_config->dm_info); - ret = libxl_domain_create_new(&priv->ctx, &d_config, - NULL, &child_console_pid, &domid); + if (restore_fd < 0) + ret = libxl_domain_create_new(&priv->ctx, &d_config, + NULL, &child_console_pid, &domid); + else + ret = libxl_domain_create_restore(&priv->ctx, &d_config, NULL, + &child_console_pid, &domid, + restore_fd); + if (ret) { - libxlError(VIR_ERR_INTERNAL_ERROR, - _("libxenlight failed to create new domain '%s'"), - d_config.c_info.name); + if (restore_fd < 0) + libxlError(VIR_ERR_INTERNAL_ERROR, + _("libxenlight failed to create new domain '%s'"), + d_config.c_info.name); + else + libxlError(VIR_ERR_INTERNAL_ERROR, + _("libxenlight failed to restore domain '%s'"), + d_config.c_info.name); goto error; } @@ -562,7 +573,9 @@ libxlVmStart(libxlDriverPrivatePtr driver, goto error; event = virDomainEventNewFromObj(vm, VIR_DOMAIN_EVENT_STARTED, - VIR_DOMAIN_EVENT_STARTED_BOOTED); + restore_fd < 0 ? + VIR_DOMAIN_EVENT_STARTED_BOOTED : + VIR_DOMAIN_EVENT_STARTED_RESTORED); libxlDomainEventQueue(driver, event); libxl_domain_config_destroy(&d_config); @@ -1046,7 +1059,8 @@ libxlDomainCreateXML(virConnectPtr conn, const char *xml, goto cleanup; def = NULL; - if (libxlVmStart(driver, vm, (flags & VIR_DOMAIN_START_PAUSED) != 0) < 0) { + if (libxlVmStart(driver, vm, (flags & VIR_DOMAIN_START_PAUSED) != 0, + -1) < 0) { virDomainRemoveInactive(&driver->domains, vm); vm = NULL; goto cleanup; @@ -1566,6 +1580,186 @@ libxlDomainGetInfo(virDomainPtr dom, virDomainInfoPtr info) } static int +libxlDomainSave(virDomainPtr dom, const char * to) +{ + libxlDriverPrivatePtr driver = dom->conn->privateData; + virDomainObjPtr vm; + libxlDomainObjPrivatePtr priv; + virDomainEventPtr event = NULL; + libxlSavefileHeader hdr; + libxl_domain_suspend_info s_info; + char *xml; + uint32_t xml_len; + int fd; + int ret = -1; + + libxlDriverLock(driver); + vm = virDomainFindByUUID(&driver->domains, dom->uuid); + + if (!vm) { + char uuidstr[VIR_UUID_STRING_BUFLEN]; + virUUIDFormat(dom->uuid, uuidstr); + libxlError(VIR_ERR_NO_DOMAIN, + _("No domain with matching uuid '%s'"), uuidstr); + goto cleanup; + } + + if (!virDomainObjIsActive(vm)) { + libxlError(VIR_ERR_OPERATION_INVALID, "%s", _("Domain is not running")); + goto cleanup; + } + + priv = vm->privateData; + + if (vm->state != VIR_DOMAIN_PAUSED) { + memset(&s_info, 0, sizeof(s_info)); + + if ((fd = virFileOpenAs(to, O_CREAT|O_TRUNC|O_WRONLY, S_IRUSR|S_IWUSR, + getuid(), getgid(), 0)) < 0) { + virReportSystemError(-fd, + _("Failed to create domain save file '%s'"), + to); + goto cleanup; + } + + if ((xml = virDomainDefFormat(vm->def, 0)) == NULL) + goto cleanup; + xml_len = strlen(xml) + 1; + + memset(&hdr, 0, sizeof(hdr)); + memcpy(hdr.magic, LIBXL_SAVE_MAGIC, sizeof(hdr.magic)); + hdr.version = LIBXL_SAVE_VERSION; + hdr.xmlLen = xml_len; + + if (safewrite(fd, &hdr, sizeof(hdr)) != sizeof(hdr)) { + libxlError(VIR_ERR_OPERATION_FAILED, + _("Failed to write save file header")); + goto cleanup; + } + + if (safewrite(fd, xml, xml_len) != xml_len) { + libxlError(VIR_ERR_OPERATION_FAILED, + _("Failed to write xml description")); + goto cleanup; + } + + if (libxl_domain_suspend(&priv->ctx, &s_info, dom->id, fd) != 0) { + libxlError(VIR_ERR_INTERNAL_ERROR, + _("Failed to save domain '%d' with libxenlight"), + dom->id); + goto cleanup; + } + + event = virDomainEventNewFromObj(vm, VIR_DOMAIN_EVENT_STOPPED, + VIR_DOMAIN_EVENT_STOPPED_SAVED); + + if (libxlVmReap(driver, vm, 1) != 0) { + libxlError(VIR_ERR_INTERNAL_ERROR, + _("Failed to destroy domain '%d'"), dom->id); + goto cleanup; + } + + if (!vm->persistent) { + virDomainRemoveInactive(&driver->domains, vm); + vm = NULL; + } + ret = 0; + } +cleanup: + VIR_FREE(xml); + if (VIR_CLOSE(fd) < 0) + virReportSystemError(errno, "%s", _("cannot close file")); + if (vm) + virDomainObjUnlock(vm); + if (event) + libxlDomainEventQueue(driver, event); + libxlDriverUnlock(driver); + return ret; +} + +static int +libxlDomainRestore(virConnectPtr conn, const char * from) +{ + libxlDriverPrivatePtr driver = conn->privateData; + virDomainDefPtr def = NULL; + virDomainObjPtr vm = NULL; + libxlSavefileHeader hdr; + char *xml = NULL; + int fd; + int ret = -1; + + libxlDriverLock(driver); + + if ((fd = virFileOpenAs(from, O_RDONLY, 0, getuid(), getgid(), 0)) < 0) { + libxlError(VIR_ERR_OPERATION_FAILED, + "%s", _("cannot read domain image")); + goto cleanup; + } + + if (saferead(fd, &hdr, sizeof(hdr)) != sizeof(hdr)) { + libxlError(VIR_ERR_OPERATION_FAILED, + "%s", _("failed to read libxl header")); + goto cleanup; + } + + if (memcmp(hdr.magic, LIBXL_SAVE_MAGIC, sizeof(hdr.magic))) { + libxlError(VIR_ERR_INVALID_ARG, "%s", _("image magic is incorrect")); + goto cleanup; + } + + if (hdr.version > LIBXL_SAVE_VERSION) { + libxlError(VIR_ERR_OPERATION_FAILED, + _("image version is not supported (%d > %d)"), + hdr.version, LIBXL_SAVE_VERSION); + goto cleanup; + } + + if (hdr.xmlLen <= 0) { + libxlError(VIR_ERR_OPERATION_FAILED, + _("invalid XML length: %d"), hdr.xmlLen); + goto cleanup; + } + + if (VIR_ALLOC_N(xml, hdr.xmlLen) < 0) { + virReportOOMError(); + goto cleanup; + } + + if (saferead(fd, xml, hdr.xmlLen) != hdr.xmlLen) { + libxlError(VIR_ERR_OPERATION_FAILED, "%s", _("failed to read XML")); + goto cleanup; + } + + if (!(def = virDomainDefParseString(driver->caps, xml, + VIR_DOMAIN_XML_INACTIVE))) + goto cleanup; + + if (virDomainObjIsDuplicate(&driver->domains, def, 1) < 0) + goto cleanup; + + if (!(vm = virDomainAssignDef(driver->caps, &driver->domains, def, true))) + goto cleanup; + + def = NULL; + + if ((ret = libxlVmStart(driver, vm, false, fd)) < 0 && + !vm->persistent) { + virDomainRemoveInactive(&driver->domains, vm); + vm = NULL; + } + +cleanup: + VIR_FREE(xml); + virDomainDefFree(def); + if (VIR_CLOSE(fd) < 0) + virReportSystemError(errno, "%s", _("cannot close file")); + if (vm) + virDomainObjUnlock(vm); + libxlDriverUnlock(driver); + return ret; +} + +static int libxlDomainSetVcpusFlags(virDomainPtr dom, unsigned int nvcpus, unsigned int flags) { @@ -2036,7 +2230,7 @@ libxlDomainCreateWithFlags(virDomainPtr dom, goto cleanup; } - ret = libxlVmStart(driver, vm, (flags & VIR_DOMAIN_START_PAUSED) != 0); + ret = libxlVmStart(driver, vm, (flags & VIR_DOMAIN_START_PAUSED) != 0, -1); cleanup: if (vm) @@ -2672,8 +2866,8 @@ static virDriver libxlDriver = { NULL, /* domainSetBlkioParameters */ NULL, /* domainGetBlkioParameters */ libxlDomainGetInfo, /* domainGetInfo */ - NULL, /* domainSave */ - NULL, /* domainRestore */ + libxlDomainSave, /* domainSave */ + libxlDomainRestore, /* domainRestore */ NULL, /* domainCoreDump */ libxlDomainSetVcpus, /* domainSetVcpus */ libxlDomainSetVcpusFlags, /* domainSetVcpusFlags */ -- 1.7.1

2 10

[libvirt] [PATCH] Remove unused 'target' field in virDomainHostdevDef
by Daniel P. Berrange 29 May '11

29 May '11

The virDomainHostdevDef struct contains a 'char *target' field. This is set to 'NULL' when parsing XML and never used / set anywhere else. Clearly it is bogus & unused * src/conf/domain_conf.c, src/conf/domain_conf.h: Remove target from virDomainHostdevDef --- src/conf/domain_conf.c | 2 -- src/conf/domain_conf.h | 1 - 2 files changed, 0 insertions(+), 3 deletions(-) diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c index 00c8058..f2e2aca 100644 --- a/src/conf/domain_conf.c +++ b/src/conf/domain_conf.c @@ -893,7 +893,6 @@ void virDomainHostdevDefFree(virDomainHostdevDefPtr def) if (!def) return; - VIR_FREE(def->target); virDomainDeviceInfoClear(&def->info); VIR_FREE(def); } @@ -4774,7 +4773,6 @@ virDomainHostdevDefParseXML(const xmlNodePtr node, virReportOOMError(); return NULL; } - def->target = NULL; mode = virXMLPropString(node, "mode"); if (mode) { diff --git a/src/conf/domain_conf.h b/src/conf/domain_conf.h index 60e1704..da34770 100644 --- a/src/conf/domain_conf.h +++ b/src/conf/domain_conf.h @@ -792,7 +792,6 @@ struct _virDomainHostdevDef { int dummy; } caps; } source; - char* target; int bootIndex; virDomainDeviceInfo info; /* Guest address */ }; -- 1.7.4.4

2 1

[libvirt] [PATCH 0/3 v2] More graceful handling of QEMU monitor failures
by Daniel P. Berrange 29 May '11

29 May '11

An update of https://www.redhat.com/archives/libvir-list/2011-May/msg00657.html Currently when libvirt has a serious error doing I/O and/or parsing of the QEMU monitor, it will kill off the guest. Application developers have expressed a desire for more graceful handling of this scenario. In particular to allow the guest OS to continue to run, without any further monitor interactons, and then kill/restart it at a time which is convenient to the guest admin/apps. New in this posting: - Change the name of the event to 'CONTROL_ERROR' and use the generic callback, instead of passing a type field - Add python dispatch code - Add 3rd patch to improve qemu monitor error reporting

3 6

[libvirt] [PATCH] Fix virExecWithHook Prototype
by Richard Laager 29 May '11

29 May '11

This was necessary to get libvirt to build on Solaris 11 Express and seems correct (as it makes this match the definition in util.c): diff --git a/src/util/util.h b/src/util/util.h index d320c40..31ec089 100644 --- a/src/util/util.h +++ b/src/util/util.h @@ -70,7 +70,7 @@ int virExecDaemonize(const char *const*argv, int virExecWithHook(const char *const*argv, const char *const*envp, const fd_set *keepfd, - int *retpid, + pid_t *retpid, int infd, int *outfd, int *errfd,

2 1

[libvirt] [PATCH] nwfilter: reorder locks
by Stefan Berger 27 May '11

27 May '11

This patch reorders the locks for the nwfilter updates and the access the nwfilter objects. In the case that the IP address learning thread was instantiating filters while an update happened, the previous order lead to a deadlock. I am also adding a text file describing the locking order of the nwfilter subsystem. Signed-off-by: Stefan Berger <stefanb(a)us.ibm.com> --- src/conf/nwfilter_conf.c | 9 ++++---- src/nwfilter/locking.txt | 45 +++++++++++++++++++++++++++++++++++++++++ src/nwfilter/nwfilter_driver.c | 4 +++ 3 files changed, 54 insertions(+), 4 deletions(-) Index: libvirt-acl/src/conf/nwfilter_conf.c =================================================================== --- libvirt-acl.orig/src/conf/nwfilter_conf.c +++ libvirt-acl/src/conf/nwfilter_conf.c @@ -2440,15 +2440,13 @@ virNWFilterTestUnassignDef(virConnectPtr { int rc = 0; - virNWFilterLockFilterUpdates(); - nwfilter->wantRemoved = 1; /* trigger the update on VMs referencing the filter */ if (virNWFilterTriggerVMFilterRebuild(conn)) rc = 1; nwfilter->wantRemoved = 0; - virNWFilterUnlockFilterUpdates(); + return rc; } @@ -2480,8 +2478,9 @@ virNWFilterObjAssignDef(virConnectPtr co return NULL; } + virNWFilterLockFilterUpdates(); + if ((nwfilter = virNWFilterObjFindByName(nwfilters, def->name))) { - virNWFilterLockFilterUpdates(); nwfilter->newDef = def; /* trigger the update on VMs referencing the filter */ if (virNWFilterTriggerVMFilterRebuild(conn)) { @@ -2498,6 +2497,8 @@ virNWFilterObjAssignDef(virConnectPtr co return nwfilter; } + virNWFilterUnlockFilterUpdates(); + if (VIR_ALLOC(nwfilter) < 0) { virReportOOMError(); return NULL; Index: libvirt-acl/src/nwfilter/nwfilter_driver.c =================================================================== --- libvirt-acl.orig/src/nwfilter/nwfilter_driver.c +++ libvirt-acl/src/nwfilter/nwfilter_driver.c @@ -372,6 +372,8 @@ nwfilterUndefine(virNWFilterPtr obj) { nwfilterDriverLock(driver); virNWFilterCallbackDriversLock(); + virNWFilterLockFilterUpdates(); + nwfilter = virNWFilterObjFindByUUID(&driver->nwfilters, obj->uuid); if (!nwfilter) { virNWFilterReportError(VIR_ERR_NO_NWFILTER, @@ -399,6 +401,8 @@ cleanup: if (nwfilter) virNWFilterObjUnlock(nwfilter); + virNWFilterUnlockFilterUpdates(); + virNWFilterCallbackDriversUnlock(); nwfilterDriverUnlock(driver); return ret; Index: libvirt-acl/src/nwfilter/locking.txt =================================================================== --- /dev/null +++ libvirt-acl/src/nwfilter/locking.txt @@ -0,0 +1,45 @@ +NWfilter locks: +--------------- + +The following NWFilter functions grab locks: + +- nwfilterDriverLock(driverState) + - protects driverState +- virNWFilterCallbackDriversLock() + - lock access to callback drivers +- virNWFilterLockFilterUpdates() + - prevent modification of any filter in use by any VM +- virNWFilterObjLock() + called by + - virNWFilterObjFindByUUID(), + - virNWFilterObjFindByName(), + - virNWFilterObjAssignDef() + and the functions return with the lock held + + +NWFilter-Lock ordering: +----------------------- + +Order in which the NWFilter lock functions have to be called: + +1. nwFilterDriverLock +2. virNWFilterCallbackDriversLock +3. virNWFilterLockFilterUpdates +4. virNWFilterObjLock + +Other locks: +------------ + +Other relevant locks to consider are related to hypervisor drivers. The +following locking orders are good with the qemu_driver as an example: + +qemu_driver - qemu_domain - filter object +qemu_driver - filter object - qemu_domain + +The qemu_driver serves as a general lock for all subsequent locks to a +qemu_domain. This works because the qemu_driver will always be locked +for a qemu_domain, thus only one lock-sequence of either (qemu_domain, +filter object) or (filter object, qemu_domain) can be active at any time. + +Reference: +https://www.redhat.com/archives/libvir-list/2010-October/msg00152.html

2 3

[libvirt] [PATCH] interface: reformat error logs
by Laine Stump 27 May '11

27 May '11

It was suggested during review of a different patch that the libvirt interface driver API's should have "netcf:" in their log messages. This patch eliminates that from all interface driver API functions, and also eliminates the extra " - " in the case that netcf returns no details in its error info (which *never* happens at present, but could happen sometime in the future. --- src/interface/netcf_driver.c | 79 +++++++++++++++++++++++++----------------- 1 files changed, 47 insertions(+), 32 deletions(-) diff --git a/src/interface/netcf_driver.c b/src/interface/netcf_driver.c index b5b4f79..dd15c9d 100644 --- a/src/interface/netcf_driver.c +++ b/src/interface/netcf_driver.c @@ -107,8 +107,9 @@ static struct netcf_if *interfaceDriverGetNetcfIF(struct netcf *ncf, virInterfac int errcode = ncf_error(ncf, &errmsg, &details); if (errcode != NETCF_NOERROR) { interfaceReportError(netcf_to_vir_err(errcode), - _("couldn't find interface named '%s' (netcf: %s - %s)"), - ifinfo->name, errmsg, details ? details : ""); + _("couldn't find interface named '%s': %s%s%s"), + ifinfo->name, errmsg, details ? " - " : "", + details ? details : ""); } else { interfaceReportError(VIR_ERR_NO_INTERFACE, _("couldn't find interface named '%s'"), @@ -188,8 +189,8 @@ static int interfaceNumOfInterfaces(virConnectPtr conn) const char *errmsg, *details; int errcode = ncf_error(driver->netcf, &errmsg, &details); interfaceReportError(netcf_to_vir_err(errcode), - _("failed to get number of interfaces on host (netcf: %s - %s)"), - errmsg, details ? details : ""); + _("failed to get number of interfaces on host: %s%s%s"), + errmsg, details ? " - " : "", details ? details : ""); } interfaceDriverUnlock(driver); @@ -208,8 +209,9 @@ static int interfaceListInterfaces(virConnectPtr conn, char **const names, int n const char *errmsg, *details; int errcode = ncf_error(driver->netcf, &errmsg, &details); interfaceReportError(netcf_to_vir_err(errcode), - _("failed to list host interfaces (netcf: %s - %s)"), - errmsg, details ? details : ""); + _("failed to list host interfaces: %s%s%s"), + errmsg, details ? " - " : "", + details ? details : ""); } interfaceDriverUnlock(driver); @@ -228,8 +230,9 @@ static int interfaceNumOfDefinedInterfaces(virConnectPtr conn) const char *errmsg, *details; int errcode = ncf_error(driver->netcf, &errmsg, &details); interfaceReportError(netcf_to_vir_err(errcode), - _("failed to get number of defined interfaces on host (netcf: %s - %s)"), - errmsg, details ? details : ""); + _("failed to get number of defined interfaces on host: %s%s%s"), + errmsg, details ? " - " : "", + details ? details : ""); } interfaceDriverUnlock(driver); @@ -248,8 +251,9 @@ static int interfaceListDefinedInterfaces(virConnectPtr conn, char **const names const char *errmsg, *details; int errcode = ncf_error(driver->netcf, &errmsg, &details); interfaceReportError(netcf_to_vir_err(errcode), - _("failed to list host defined interfaces (netcf: %s - %s)"), - errmsg, details ? details : ""); + _("failed to list host defined interfaces: %s%s%s"), + errmsg, details ? " - " : "", + details ? details : ""); } interfaceDriverUnlock(driver); @@ -271,8 +275,9 @@ static virInterfacePtr interfaceLookupByName(virConnectPtr conn, int errcode = ncf_error(driver->netcf, &errmsg, &details); if (errcode != NETCF_NOERROR) { interfaceReportError(netcf_to_vir_err(errcode), - _("couldn't find interface named '%s' (netcf: %s - %s)"), - name, errmsg, details ? details : ""); + _("couldn't find interface named '%s': %s%s%s"), + name, errmsg, + details ? " - " : "", details ? details : ""); } else { interfaceReportError(VIR_ERR_NO_INTERFACE, _("couldn't find interface named '%s'"), name); @@ -303,8 +308,9 @@ static virInterfacePtr interfaceLookupByMACString(virConnectPtr conn, const char *errmsg, *details; int errcode = ncf_error(driver->netcf, &errmsg, &details); interfaceReportError(netcf_to_vir_err(errcode), - _("couldn't find interface with MAC address '%s' (netcf: %s - %s)"), - macstr, errmsg, details ? details : ""); + _("couldn't find interface with MAC address '%s': %s%s%s"), + macstr, errmsg, details ? " - " : "", + details ? details : ""); goto cleanup; } if (niface == 0) { @@ -353,8 +359,9 @@ static char *interfaceGetXMLDesc(virInterfacePtr ifinfo, const char *errmsg, *details; int errcode = ncf_error(driver->netcf, &errmsg, &details); interfaceReportError(netcf_to_vir_err(errcode), - _("could not get interface XML description (netcf: %s - %s)"), - errmsg, details ? details : ""); + _("could not get interface XML description: %s%s%s"), + errmsg, details ? " - " : "", + details ? details : ""); goto cleanup; } @@ -407,8 +414,9 @@ static virInterfacePtr interfaceDefineXML(virConnectPtr conn, const char *errmsg, *details; int errcode = ncf_error(driver->netcf, &errmsg, &details); interfaceReportError(netcf_to_vir_err(errcode), - _("could not get interface XML description (netcf: %s - %s)"), - errmsg, details ? details : ""); + _("could not get interface XML description: %s%s%s"), + errmsg, details ? " - " : "", + details ? details : ""); goto cleanup; } @@ -440,8 +448,9 @@ static int interfaceUndefine(virInterfacePtr ifinfo) { const char *errmsg, *details; int errcode = ncf_error(driver->netcf, &errmsg, &details); interfaceReportError(netcf_to_vir_err(errcode), - _("failed to undefine interface %s (netcf: %s - %s)"), - ifinfo->name, errmsg, details ? details : ""); + _("failed to undefine interface %s: %s%s%s"), + ifinfo->name, errmsg, details ? " - " : "", + details ? details : ""); goto cleanup; } @@ -471,8 +480,9 @@ static int interfaceCreate(virInterfacePtr ifinfo, const char *errmsg, *details; int errcode = ncf_error(driver->netcf, &errmsg, &details); interfaceReportError(netcf_to_vir_err(errcode), - _("failed to create (start) interface %s (netcf: %s - %s)"), - ifinfo->name, errmsg, details ? details : ""); + _("failed to create (start) interface %s: %s%s%s"), + ifinfo->name, errmsg, details ? " - " : "", + details ? details : ""); goto cleanup; } @@ -502,8 +512,9 @@ static int interfaceDestroy(virInterfacePtr ifinfo, const char *errmsg, *details; int errcode = ncf_error(driver->netcf, &errmsg, &details); interfaceReportError(netcf_to_vir_err(errcode), - _("failed to destroy (stop) interface %s (netcf: %s - %s)"), - ifinfo->name, errmsg, details ? details : ""); + _("failed to destroy (stop) interface %s: %s%s%s"), + ifinfo->name, errmsg, details ? " - " : "", + details ? details : ""); goto cleanup; } @@ -532,8 +543,9 @@ static int interfaceIsActive(virInterfacePtr ifinfo) const char *errmsg, *details; int errcode = ncf_error(driver->netcf, &errmsg, &details); interfaceReportError(netcf_to_vir_err(errcode), - _("failed to get status of interface %s (netcf: %s - %s)"), - ifinfo->name, errmsg, details ? details : ""); + _("failed to get status of interface %s: %s%s%s"), + ifinfo->name, errmsg, details ? " - " : "", + details ? details : ""); goto cleanup; } @@ -560,8 +572,9 @@ static int interfaceChangeBegin(virConnectPtr conn, unsigned int flags) const char *errmsg, *details; int errcode = ncf_error(driver->netcf, &errmsg, &details); interfaceReportError(netcf_to_vir_err(errcode), - _("failed to begin transaction: %s%s%s)"), - errmsg, details ? " - " : "", details ? details : ""); + _("failed to begin transaction: %s%s%s"), + errmsg, details ? " - " : "", + details ? details : ""); } interfaceDriverUnlock(driver); @@ -582,8 +595,9 @@ static int interfaceChangeCommit(virConnectPtr conn, unsigned int flags) const char *errmsg, *details; int errcode = ncf_error(driver->netcf, &errmsg, &details); interfaceReportError(netcf_to_vir_err(errcode), - _("failed to commit transaction: %s%s%s)"), - errmsg, details ? " - " : "", details ? details : ""); + _("failed to commit transaction: %s%s%s"), + errmsg, details ? " - " : "", + details ? details : ""); } interfaceDriverUnlock(driver); @@ -604,8 +618,9 @@ static int interfaceChangeRollback(virConnectPtr conn, unsigned int flags) const char *errmsg, *details; int errcode = ncf_error(driver->netcf, &errmsg, &details); interfaceReportError(netcf_to_vir_err(errcode), - _("failed to rollback transaction: %s%s%s)"), - errmsg, details ? " - " : "", details ? details : ""); + _("failed to rollback transaction: %s%s%s"), + errmsg, details ? " - " : "", + details ? details : ""); } interfaceDriverUnlock(driver); -- 1.7.3.4

2 2

Re: [libvirt] [Qemu-devel] [RFC] live snapshot, live merge, live block migration
by Dor Laor 27 May '11

27 May '11

On 05/20/2011 03:19 PM, Stefan Hajnoczi wrote: > I'm interested in what the API for snapshots would look like. > Specifically how does user software do the following: > 1. Create a snapshot > 2. Delete a snapshot > 3. List snapshots > 4. Access data from a snapshot There are plenty of options there: - Run a (unrelated) VM and hotplug the snapshot as additional disk - Use v2v (libguestfs) - Boot the VM w/ RO - Plenty more > 5. Restore a VM from a snapshot > 6. Get the dirty blocks list (for incremental backup) It might be needed for additional proposes like efficient delta sync across sites or any other storage operation (dedup, etc) > > We've discussed image format-level approaches but I think the scope of > the API should cover several levels at which snapshots are > implemented: > 1. Image format - image file snapshot (Jes, Jagane) > 2. Host file system - ext4 and btrfs snapshots > 3. Storage system - LVM or SAN volume snapshots > > It will be hard to take advantage of more efficient host file system > or storage system snapshots if they are not designed in now. I agree but it can also be a chicken and the egg problem. Actually 1/2/3/5 are already working today regardless of live snapshots. > Is anyone familiar enough with the libvirt storage APIs to draft an > extension that adds snapshot support? I will take a stab at it if no > one else want to try it. I added libvirt-list and Ayal Baron from vdsm. What you're asking is even beyond snapshots, it's the whole management of VM images. Doing the above operations is simple but for enterprise virtualization solution you'll need to lock the NFS/SAN images, handle failures of VM/SAN/Mgmt, keep the snapshots info in mgmt DB, etc. Today it is managed by a combination of rhev-m/vdsm and libvirt. I agree it would have been nice to get such common single entry point interface. > > Stefan

3 3