- Devel - libvirt List Archives

[libvirt] [PATCH 2/2] libxl: support pci passthrough
by Chunyan Liu 11 Jan '13

11 Jan '13

Support PCI passthrough in libxl driver. Signed-off-by: Chunyan Liu <cyliu(a)suse.com> --- src/libxl/libxl_conf.c | 44 ++++++++++++++++++++++++++++++++++++++++++++ src/libxl/libxl_conf.h | 5 ++++- src/libxl/libxl_driver.c | 16 ++++++++++++++++ 3 files changed, 64 insertions(+), 1 deletions(-) diff --git a/src/libxl/libxl_conf.c b/src/libxl/libxl_conf.c index 2705e65..223f7cb 100644 --- a/src/libxl/libxl_conf.c +++ b/src/libxl/libxl_conf.c @@ -814,6 +814,46 @@ error: return -1; } +static int +libxlMakePciList(virDomainDefPtr def, libxl_domain_config *d_config) +{ + virDomainHostdevDefPtr *l_hostdevs = def->hostdevs; + int nhostdevs = def->nhostdevs; + libxl_device_pci *x_pcidevs; + int i; + + if (nhostdevs == 0) + return 0; + + if (VIR_ALLOC_N(x_pcidevs, nhostdevs) < 0) { + virReportOOMError(); + return -1; + } + + d_config->num_pcidevs = 0; + d_config->pcidevs = x_pcidevs; + + for (i = 0 ; i < nhostdevs ; i++) { + virDomainHostdevDefPtr hostdev = l_hostdevs[i]; + libxl_device_pci pcidev; + + if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS) + continue; + if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI) + continue; + + pcidev.domain = hostdev->source.subsys.u.pci.domain; + pcidev.bus = hostdev->source.subsys.u.pci.bus; + pcidev.dev = hostdev->source.subsys.u.pci.slot; + pcidev.func = hostdev->source.subsys.u.pci.function; + + d_config->pcidevs[d_config->num_pcidevs] = pcidev; + d_config->num_pcidevs ++; + } + + return 0; +} + virCapsPtr libxlMakeCapabilities(libxl_ctx *ctx) { @@ -863,6 +903,10 @@ libxlBuildDomainConfig(libxlDriverPrivatePtr driver, goto error; } + if (libxlMakePciList(def, d_config) < 0) { + goto error; + } + d_config->on_reboot = def->onReboot; d_config->on_poweroff = def->onPoweroff; d_config->on_crash = def->onCrash; diff --git a/src/libxl/libxl_conf.h b/src/libxl/libxl_conf.h index c8808a1..704d8b8 100644 --- a/src/libxl/libxl_conf.h +++ b/src/libxl/libxl_conf.h @@ -35,7 +35,7 @@ # include "capabilities.h" # include "configmake.h" # include "virbitmap.h" - +# include "pci.h" # define LIBXL_VNC_PORT_MIN 5900 # define LIBXL_VNC_PORT_MAX 65535 @@ -76,6 +76,9 @@ struct _libxlDriverPrivate { char *stateDir; char *libDir; char *saveDir; + + pciDeviceList *activePciHostdevs; + pciDeviceList *inactivePciHostdevs; }; typedef struct _libxlDomainObjPrivate libxlDomainObjPrivate; diff --git a/src/libxl/libxl_driver.c b/src/libxl/libxl_driver.c index a956188..4670fee 100644 --- a/src/libxl/libxl_driver.c +++ b/src/libxl/libxl_driver.c @@ -42,6 +42,7 @@ #include "libxl.h" #include "libxl_driver.h" #include "libxl_conf.h" +#include "libxl_hostdev.h" #include "xen_xm.h" #include "virtypedparam.h" #include "viruri.h" @@ -514,6 +515,8 @@ libxlVmReap(libxlDriverPrivatePtr driver, return -1; } + libxlDomainReAttachHostDevices(driver, vm->def); + libxlVmCleanup(driver, vm, reason); return 0; } @@ -756,6 +759,10 @@ libxlVmStart(libxlDriverPrivatePtr driver, virDomainObjPtr vm, VIR_FREE(managed_save_path); } + VIR_DEBUG("Preparing host PCI devices"); + if (libxlPrepareHostDevices(driver, vm->def) < 0) + goto error; + libxl_domain_config_init(&d_config); if (libxlBuildDomainConfig(driver, vm->def, &d_config) < 0) @@ -931,6 +938,8 @@ libxlShutdown(void) VIR_FREE(libxl_driver->stateDir); VIR_FREE(libxl_driver->libDir); VIR_FREE(libxl_driver->saveDir); + pciDeviceListFree(libxl_driver->activePciHostdevs); + pciDeviceListFree(libxl_driver->inactivePciHostdevs); virDomainEventStateFree(libxl_driver->domainEventState); @@ -1077,6 +1086,12 @@ libxlStartup(bool privileged, libxl_driver->caps->privateDataAllocFunc = libxlDomainObjPrivateAlloc; libxl_driver->caps->privateDataFreeFunc = libxlDomainObjPrivateFree; + if ((libxl_driver->activePciHostdevs = pciDeviceListNew()) == NULL) + goto error; + + if ((libxl_driver->inactivePciHostdevs = pciDeviceListNew()) == NULL) + goto error; + /* Load running domains first. */ if (virDomainLoadAllConfigs(libxl_driver->caps, &libxl_driver->domains, @@ -1549,6 +1564,7 @@ libxlDomainShutdownFlags(virDomainPtr dom, unsigned int flags) /* vm is marked shutoff (or removed from domains list if not persistent) * in shutdown event handler. */ + libxlDomainReAttachHostDevices(driver, vm->def); ret = 0; cleanup: -- 1.7.3.4

1 0

[libvirt] [PATCH 1/2] libxl: add APIs to prepare/release host pci device
by Chunyan Liu 11 Jan '13

11 Jan '13

Handle preparing and releasing host pci devices for PCI passthrough. Preparing host pci devices will be used before vm is started, while releasing (or, reattaching) will be used after vm is shut off. Code refers to qemu driver, so add Red Hat copyright. Signed-off-by: Chunyan Liu <cyliu(a)suse.com> --- po/POTFILES.in | 1 + src/Makefile.am | 1 + src/libxl/libxl_hostdev.c | 600 +++++++++++++++++++++++++++++++++++++++++++++ src/libxl/libxl_hostdev.h | 44 ++++ 4 files changed, 646 insertions(+), 0 deletions(-) create mode 100644 src/libxl/libxl_hostdev.c create mode 100644 src/libxl/libxl_hostdev.h diff --git a/po/POTFILES.in b/po/POTFILES.in index 95619f9..a2b8d6b 100644 --- a/po/POTFILES.in +++ b/po/POTFILES.in @@ -66,6 +66,7 @@ src/lxc/lxc_monitor.c src/lxc/lxc_process.c src/libxl/libxl_driver.c src/libxl/libxl_conf.c +src/libxl/libxl_hostdev.c src/network/bridge_driver.c src/node_device/node_device_driver.c src/node_device/node_device_hal.c diff --git a/src/Makefile.am b/src/Makefile.am index da571c7..d7149b5 100644 --- a/src/Makefile.am +++ b/src/Makefile.am @@ -531,6 +531,7 @@ XENAPI_DRIVER_SOURCES = \ LIBXL_DRIVER_SOURCES = \ libxl/libxl_conf.c libxl/libxl_conf.h \ + libxl/libxl_hostdev.c libxl/libxl_hostdev.h \ libxl/libxl_driver.c libxl/libxl_driver.h UML_DRIVER_SOURCES = \ diff --git a/src/libxl/libxl_hostdev.c b/src/libxl/libxl_hostdev.c new file mode 100644 index 0000000..19c6d82 --- /dev/null +++ b/src/libxl/libxl_hostdev.c @@ -0,0 +1,600 @@ +/*---------------------------------------------------------------------------*/ +/* Copyright (C) 2006-2012 Red Hat, Inc. + * Copyright (c) 2013 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 + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this library. If not, see + * <http://www.gnu.org/licenses/>. + * + * Authors: + * Daniel P. Berrange <berrange(a)redhat.com> + * Chunyan Liu <cyliu(a)suse.com> + */ +/*---------------------------------------------------------------------------*/ + +#include <config.h> + +#include "libxl_hostdev.h" +#include "logging.h" +#include "virterror_internal.h" +#include "memory.h" +#include "pci.h" +#include "virnetdev.h" + +#define VIR_FROM_THIS VIR_FROM_LIBXL + +static pciDeviceList * +libxlGetActivePciHostDeviceList(libxlDriverPrivatePtr driver, + virDomainHostdevDefPtr *hostdevs, + int nhostdevs) +{ + pciDeviceList *list; + int i; + + if (!(list = pciDeviceListNew())) + return NULL; + + for (i = 0 ; i < nhostdevs ; i++) { + virDomainHostdevDefPtr hostdev = hostdevs[i]; + pciDevice *dev; + pciDevice *activeDev; + + if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS) + continue; + if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI) + continue; + + dev = pciGetDevice(hostdev->source.subsys.u.pci.domain, + hostdev->source.subsys.u.pci.bus, + hostdev->source.subsys.u.pci.slot, + hostdev->source.subsys.u.pci.function); + if (!dev) { + pciDeviceListFree(list); + return NULL; + } + + if ((activeDev = pciDeviceListFind(driver->activePciHostdevs, dev))) { + if (pciDeviceListAdd(list, activeDev) < 0) { + pciFreeDevice(dev); + pciDeviceListFree(list); + return NULL; + } + } + + pciFreeDevice(dev); + } + + return list; +} + +static pciDeviceList * +libxlGetPciHostDeviceList(virDomainHostdevDefPtr *hostdevs, int nhostdevs) +{ + pciDeviceList *list; + int i; + + if (!(list = pciDeviceListNew())) + return NULL; + + for (i = 0 ; i < nhostdevs ; i++) { + virDomainHostdevDefPtr hostdev = hostdevs[i]; + pciDevice *dev; + + if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS) + continue; + if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI) + continue; + + dev = pciGetDevice(hostdev->source.subsys.u.pci.domain, + hostdev->source.subsys.u.pci.bus, + hostdev->source.subsys.u.pci.slot, + hostdev->source.subsys.u.pci.function); + if (!dev) { + pciDeviceListFree(list); + return NULL; + } + + if (pciDeviceListAdd(list, dev) < 0) { + pciFreeDevice(dev); + pciDeviceListFree(list); + return NULL; + } + + pciDeviceSetManaged(dev, hostdev->managed); + } + + return list; +} + +void libxlReattachPciDevice(pciDevice *dev, libxlDriverPrivatePtr driver) +{ + /* If the device is not managed and was attached to guest + * successfully, it must have been inactive. + */ + if (!pciDeviceGetManaged(dev)) { + pciDeviceListAdd(driver->inactivePciHostdevs, dev); + return; + } + + if (pciReAttachDevice(dev, driver->activePciHostdevs, + driver->inactivePciHostdevs, "pciback") < 0) { + virErrorPtr err = virGetLastError(); + VIR_ERROR(_("Failed to re-attach PCI device: %s"), + err ? err->message : _("unknown error")); + virResetError(err); + } +} + +static int +libxlDomainHostdevPciSysfsPath(virDomainHostdevDefPtr hostdev, char **sysfs_path) +{ + struct pci_config_address config_address; + + config_address.domain = hostdev->source.subsys.u.pci.domain; + config_address.bus = hostdev->source.subsys.u.pci.bus; + config_address.slot = hostdev->source.subsys.u.pci.slot; + config_address.function = hostdev->source.subsys.u.pci.function; + + return pciConfigAddressToSysfsFile(&config_address, sysfs_path); +} + +static int +libxlDomainHostdevIsVirtualFunction(virDomainHostdevDefPtr hostdev) +{ + char *sysfs_path = NULL; + int ret = -1; + + if (libxlDomainHostdevPciSysfsPath(hostdev, &sysfs_path) < 0) + return ret; + + ret = pciDeviceIsVirtualFunction(sysfs_path); + + VIR_FREE(sysfs_path); + + return ret; +} + +static int +libxlDomainHostdevNetDevice(virDomainHostdevDefPtr hostdev, char **linkdev, + int *vf) +{ + int ret = -1; + char *sysfs_path = NULL; + + if (libxlDomainHostdevPciSysfsPath(hostdev, &sysfs_path) < 0) + return ret; + + if (pciDeviceIsVirtualFunction(sysfs_path) == 1) { + if (pciDeviceGetVirtualFunctionInfo(sysfs_path, linkdev, + vf) < 0) + goto cleanup; + } else { + if (pciDeviceNetName(sysfs_path, linkdev) < 0) + goto cleanup; + *vf = -1; + } + + ret = 0; + +cleanup: + VIR_FREE(sysfs_path); + + return ret; +} + +static int +libxlDomainHostdevNetConfigVirtPortProfile(const char *linkdev, int vf, + virNetDevVPortProfilePtr virtPort, + const virMacAddrPtr macaddr, + const unsigned char *uuid, + int associate) +{ + int ret = -1; + + if (!virtPort) + return ret; + + switch (virtPort->virtPortType) { + case VIR_NETDEV_VPORT_PROFILE_NONE: + case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH: + case VIR_NETDEV_VPORT_PROFILE_8021QBG: + case VIR_NETDEV_VPORT_PROFILE_LAST: + virReportError(VIR_ERR_CONFIG_UNSUPPORTED, + _("virtualport type %s is " + "currently not supported on interfaces of type " + "hostdev"), + virNetDevVPortTypeToString(virtPort->virtPortType)); + break; + + case VIR_NETDEV_VPORT_PROFILE_8021QBH: + if (associate) + ret = virNetDevVPortProfileAssociate(NULL, virtPort, macaddr, + linkdev, vf, uuid, + VIR_NETDEV_VPORT_PROFILE_OP_CREATE, false); + else + ret = virNetDevVPortProfileDisassociate(NULL, virtPort, + macaddr, linkdev, vf, + VIR_NETDEV_VPORT_PROFILE_OP_DESTROY); + break; + } + + return ret; +} + +static int +libxlDomainHostdevNetConfigReplace(virDomainHostdevDefPtr hostdev, + const unsigned char *uuid, + char *stateDir) +{ + char *linkdev = NULL; + virNetDevVlanPtr vlan; + virNetDevVPortProfilePtr virtPort; + int ret = -1; + int vf = -1; + int vlanid = -1; + int port_profile_associate = 1; + int isvf; + + isvf = libxlDomainHostdevIsVirtualFunction(hostdev); + if (isvf <= 0) { + virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s", + _("Interface type hostdev is currently supported on" + " SR-IOV Virtual Functions only")); + return ret; + } + + if (libxlDomainHostdevNetDevice(hostdev, &linkdev, &vf) < 0) + return ret; + + vlan = virDomainNetGetActualVlan(hostdev->parent.data.net) + virtPort = virDomainNetGetActualVirtPortProfile( + hostdev->parent.data.net) + if (virtPort) { + if (vlan) { + virReportError(VIR_ERR_CONFIG_UNSUPPORTED, + _("direct setting of the vlan tag is not allowed " + "for hostdev devices using %s mode"), + virNetDevVPortTypeToString(virtPort->virtPortType)); + goto cleanup; + } + ret = libxlDomainHostdevNetConfigVirtPortProfile(linkdev, vf, + virtPort, &hostdev->parent.data.net->mac, uuid, + port_profile_associate); + } else { + /* Set only mac and vlan */ + if (vlan) { + if (vlan->nTags != 1 || vlan->trunk) { + virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s", + _("vlan trunking is not supported " + "by SR-IOV network devices")); + goto cleanup; + } + if (vf == -1) { + virReportError(VIR_ERR_CONFIG_UNSUPPORTED, + _("vlan can only be set for SR-IOV VFs, but " + "%s is not a VF"), linkdev); + goto cleanup; + } + vlanid = vlan->tag[0]; + } else if (vf >= 0) { + vlanid = 0; /* assure any current vlan tag is reset */ + } + + ret = virNetDevReplaceNetConfig(linkdev, vf, + &hostdev->parent.data.net->mac, + vlanid, stateDir); + } +cleanup: + VIR_FREE(linkdev); + return ret; +} + +static int +libxlDomainHostdevNetConfigRestore(virDomainHostdevDefPtr hostdev, + char *stateDir) +{ + char *linkdev = NULL; + virNetDevVPortProfilePtr virtPort; + int ret = -1; + int vf = -1; + int port_profile_associate = 0; + int isvf; + + isvf = libxlDomainHostdevIsVirtualFunction(hostdev); + if (isvf <= 0) { + virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s", + _("Interface type hostdev is currently supported on" + " SR-IOV Virtual Functions only")); + return ret; + } + + if (libxlDomainHostdevNetDevice(hostdev, &linkdev, &vf) < 0) + return ret; + + virtPort = virDomainNetGetActualVirtPortProfile( + hostdev->parent.data.net) + if (virtPort) + ret = libxlDomainHostdevNetConfigVirtPortProfile(linkdev, vf, virtPort, + &hostdev->parent.data.net->mac, NULL, + port_profile_associate); + else + ret = virNetDevRestoreNetConfig(linkdev, vf, stateDir); + + VIR_FREE(linkdev); + + return ret; +} + +void libxlDomainReAttachHostdevPciDevices(libxlDriverPrivatePtr driver, + const char *name, + virDomainHostdevDefPtr *hostdevs, + int nhostdevs) +{ + pciDeviceList *pcidevs; + int i; + + if (!(pcidevs = libxlGetActivePciHostDeviceList(driver, + hostdevs, + nhostdevs))) { + virErrorPtr err = virGetLastError(); + VIR_ERROR(_("Failed to allocate pciDeviceList: %s"), + err ? err->message : _("unknown error")); + virResetError(err); + return; + } + + /* Again 4 loops; mark all devices as inactive before reset + * them and reset all the devices before re-attach. + * Attach mac and port profile parameters to devices + */ + for (i = 0; i < pciDeviceListCount(pcidevs); i++) { + pciDevice *dev = pciDeviceListGet(pcidevs, i); + pciDevice *activeDev = NULL; + + /* Never delete the dev from list driver->activePciHostdevs + * if it's used by other domain. + */ + activeDev = pciDeviceListFind(driver->activePciHostdevs, dev); + if (activeDev && + STRNEQ_NULLABLE(name, pciDeviceGetUsedBy(activeDev))) { + pciDeviceListSteal(pcidevs, dev); + continue; + } + + /* pciDeviceListFree() will take care of freeing the dev. */ + pciDeviceListSteal(driver->activePciHostdevs, dev); + } + + /* + * For SRIOV net host devices, unset mac and port profile before + * reset and reattach device + */ + for (i = 0; i < nhostdevs; i++) { + virDomainHostdevDefPtr hostdev = hostdevs[i]; + if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS) + continue; + if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI) + continue; + if (hostdev->parent.type == VIR_DOMAIN_DEVICE_NET && + hostdev->parent.data.net) { + libxlDomainHostdevNetConfigRestore(hostdev, driver->stateDir); + } + } + + for (i = 0; i < pciDeviceListCount(pcidevs); i++) { + pciDevice *dev = pciDeviceListGet(pcidevs, i); + if (pciResetDevice(dev, driver->activePciHostdevs, + driver->inactivePciHostdevs) < 0) { + virErrorPtr err = virGetLastError(); + VIR_ERROR(_("Failed to reset PCI device: %s"), + err ? err->message : _("unknown error")); + virResetError(err); + } + } + + for (i = 0; i < pciDeviceListCount(pcidevs); i++) { + pciDevice *dev = pciDeviceListGet(pcidevs, i); + libxlReattachPciDevice(dev, driver); + } + + pciDeviceListFree(pcidevs); +} + +void libxlDomainReAttachHostDevices(libxlDriverPrivatePtr driver, + virDomainDefPtr def) +{ + if (!def->nhostdevs) + return; + + libxlDomainReAttachHostdevPciDevices(driver, def->name, def->hostdevs, + def->nhostdevs); +} + +int libxlPrepareHostdevPCIDevices(libxlDriverPrivatePtr driver, + const char *name, + const unsigned char *uuid, + virDomainHostdevDefPtr *hostdevs, + int nhostdevs) +{ + pciDeviceList *pcidevs; + int last_processed_hostdev_vf = -1; + int i; + int ret = -1; + + if (!(pcidevs = libxlGetPciHostDeviceList(hostdevs, nhostdevs))) + return -1; + + /* Loop 1: validate that non-managed device isn't in use */ + + for (i = 0; i < pciDeviceListCount(pcidevs); i++) { + pciDevice *dev = pciDeviceListGet(pcidevs, i); + pciDevice *other; + + /* The device is in use by other active domain if + * the dev is in list driver->activePciHostdevs. + */ + if ((other = pciDeviceListFind(driver->activePciHostdevs, dev))) { + const char *other_name = pciDeviceGetUsedBy(other); + + if (other_name) + virReportError(VIR_ERR_OPERATION_INVALID, + _("PCI device %s is in use by domain %s"), + pciDeviceGetName(dev), other_name); + else + virReportError(VIR_ERR_OPERATION_INVALID, + _("PCI device %s is already in use"), + pciDeviceGetName(dev)); + goto cleanup; + } + } + + /* Loop 2: detach managed devices */ + for (i = 0; i < pciDeviceListCount(pcidevs); i++) { + pciDevice *dev = pciDeviceListGet(pcidevs, i); + if (pciDeviceGetManaged(dev) && + pciDettachDevice(dev, driver->activePciHostdevs, NULL, "pciback") < 0) + goto reattachdevs; + } + + /* Loop 3: Now that all the PCI hostdevs have been detached, we + * can safely reset them */ + for (i = 0; i < pciDeviceListCount(pcidevs); i++) { + pciDevice *dev = pciDeviceListGet(pcidevs, i); + if (pciResetDevice(dev, driver->activePciHostdevs, + driver->inactivePciHostdevs) < 0) + goto reattachdevs; + } + + /* Loop 4: For SRIOV network devices, Now that we have detached the + * the network device, set the netdev config */ + for (i = 0; i < nhostdevs; i++) { + virDomainHostdevDefPtr hostdev = hostdevs[i]; + if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS) + continue; + if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI) + continue; + if (hostdev->parent.type == VIR_DOMAIN_DEVICE_NET && + hostdev->parent.data.net) { + if (libxlDomainHostdevNetConfigReplace(hostdev, uuid, + driver->stateDir) < 0) { + goto resetvfnetconfig; + } + } + last_processed_hostdev_vf = i; + } + + /* Loop 5: Now mark all the devices as active */ + for (i = 0; i < pciDeviceListCount(pcidevs); i++) { + pciDevice *dev = pciDeviceListGet(pcidevs, i); + if (pciDeviceListAdd(driver->activePciHostdevs, dev) < 0) { + pciFreeDevice(dev); + goto inactivedevs; + } + } + + /* Loop 6: Now remove the devices from inactive list. */ + for (i = 0; i < pciDeviceListCount(pcidevs); i++) { + pciDevice *dev = pciDeviceListGet(pcidevs, i); + pciDeviceListDel(driver->inactivePciHostdevs, dev); + } + + /* Loop 7: Now set the used_by_domain of the device in + * driver->activePciHostdevs as domain name. + */ + for (i = 0; i < pciDeviceListCount(pcidevs); i++) { + pciDevice *dev, *activeDev; + + dev = pciDeviceListGet(pcidevs, i); + activeDev = pciDeviceListFind(driver->activePciHostdevs, dev); + + pciDeviceSetUsedBy(activeDev, name); + } + + /* Loop 8: Now set the original states for hostdev def */ + for (i = 0; i < nhostdevs; i++) { + pciDevice *dev; + pciDevice *pcidev; + virDomainHostdevDefPtr hostdev = hostdevs[i]; + + if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS) + continue; + if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI) + continue; + + dev = pciGetDevice(hostdev->source.subsys.u.pci.domain, + hostdev->source.subsys.u.pci.bus, + hostdev->source.subsys.u.pci.slot, + hostdev->source.subsys.u.pci.function); + + /* original states "unbind_from_stub", "remove_slot", + * "reprobe" were already set by pciDettachDevice in + * loop 2. + */ + if ((pcidev = pciDeviceListFind(pcidevs, dev))) { + hostdev->origstates.states.pci.unbind_from_stub = + pciDeviceGetUnbindFromStub(pcidev); + hostdev->origstates.states.pci.remove_slot = + pciDeviceGetRemoveSlot(pcidev); + hostdev->origstates.states.pci.reprobe = + pciDeviceGetReprobe(pcidev); + } + + pciFreeDevice(dev); + } + + /* Loop 9: Now steal all the devices from pcidevs */ + while (pciDeviceListCount(pcidevs) > 0) { + pciDevice *dev = pciDeviceListGet(pcidevs, 0); + pciDeviceListSteal(pcidevs, dev); + } + + ret = 0; + goto cleanup; + +inactivedevs: + /* Only steal all the devices from driver->activePciHostdevs. We will + * free them in pciDeviceListFree(). + */ + while (pciDeviceListCount(pcidevs) > 0) { + pciDevice *dev = pciDeviceListGet(pcidevs, 0); + pciDeviceListSteal(driver->activePciHostdevs, dev); + } + +resetvfnetconfig: + for (i = 0; i < last_processed_hostdev_vf; i++) { + virDomainHostdevDefPtr hostdev = hostdevs[i]; + if (hostdev->parent.type == VIR_DOMAIN_DEVICE_NET && + hostdev->parent.data.net) { + libxlDomainHostdevNetConfigRestore(hostdev, driver->stateDir); + } + } + +reattachdevs: + for (i = 0; i < pciDeviceListCount(pcidevs); i++) { + pciDevice *dev = pciDeviceListGet(pcidevs, i); + pciReAttachDevice(dev, driver->activePciHostdevs, NULL, "pciback"); + } + +cleanup: + pciDeviceListFree(pcidevs); + return ret; +} + +int +libxlPrepareHostDevices(libxlDriverPrivatePtr driver, + virDomainDefPtr def) +{ + return libxlPrepareHostdevPCIDevices(driver, def->name, def->uuid, + def->hostdevs, def->nhostdevs); +} diff --git a/src/libxl/libxl_hostdev.h b/src/libxl/libxl_hostdev.h new file mode 100644 index 0000000..e5a5058 --- /dev/null +++ b/src/libxl/libxl_hostdev.h @@ -0,0 +1,44 @@ +/*---------------------------------------------------------------------------*/ +/* Copyright (c) 2013 SUSE LINUX Products GmbH, Nuernberg, Germany. + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this library. If not, see + * <http://www.gnu.org/licenses/>. + * + * Authors: + * Chunyan Liu <cyliu(a)suse.com> + */ +/*---------------------------------------------------------------------------*/ + +#ifndef __LIBXL_HOSTDEV_H__ +# define __LIBXL_HOSTDEV_H__ + +# include "libxl_conf.h" +# include "domain_conf.h" + +int libxlPrepareHostdevPCIDevices(libxlDriverPrivatePtr driver, + const char *name, + const unsigned char *uuid, + virDomainHostdevDefPtr *hostdevs, + int nhostdevs); +int libxlPrepareHostDevices(libxlDriverPrivatePtr driver, + virDomainDefPtr def); +void libxlReattachPciDevice(pciDevice *dev, libxlDriverPrivatePtr driver); +void libxlDomainReAttachHostdevPciDevices(libxlDriverPrivatePtr driver, + const char *name, + virDomainHostdevDefPtr *hostdevs, + int nhostdevs); +void libxlDomainReAttachHostDevices(libxlDriverPrivatePtr driver, + virDomainDefPtr def); + +#endif /* __LIBXL_HOSTDEV_H__ */ -- 1.7.3.4

1 0

[libvirt] [PATCH qom-cpu 0/7] disable kvm_mmu + -cpu "enforce" fixes (v3)
by Eduardo Habkost 11 Jan '13

11 Jan '13

Changes on v3: - Patches 3-9 from v2 are now already on qom-cpu tree - Remove CONFIG_KVM #ifdefs by declaring fake KVM_* #defines on sysemu/kvm.h - Refactor code that uses the feature word arrays (to make it easier to add a new feature name array) - Add feature name array for CPUID leaf 0xC0000001 Changes on v2: - Now both the kvm_mmu-disable and -cpu "enforce" changes are on the same series - Coding style fixes Git tree for reference: git://github.com/ehabkost/qemu-hacks.git cpu-enforce-all.v3 https://github.com/ehabkost/qemu-hacks/tree/cpu-enforce-all.v3 The changes are a bit intrusive, but: - The longer we take to make "enforce" strict as it should (and make libvirt finally use it), more users will have VMs with migration-unsafe unpredictable guest ABIs. For this reason, I would like to get this into QEMU 1.4. - The changes in this series should affect only users that are already using the "enforce" flag, and I believe whoever is using the "enforce" flag really want the strict behavior introduced by this series. Eduardo Habkost (7): kvm: Add fake KVM constants to avoid #ifdefs on KVM-specific code target-i386: Don't set any KVM flag by default if KVM is disabled target-i386: Disable kvm_mmu by default target-i386/cpu: Introduce FeatureWord typedefs target-i386: kvm_check_features_against_host(): Use feature_word_info target-i386/cpu.c: Add feature name array for ext4_features target-i386: check/enforce: Check all feature words include/sysemu/kvm.h | 14 ++++ target-i386/cpu.c | 193 ++++++++++++++++++++++++++++++++------------------- target-i386/cpu.h | 15 ++++ 3 files changed, 150 insertions(+), 72 deletions(-) -- 1.7.11.7

4 15

[libvirt] [PATCH] build: require avahi instead of avahi-libs on rhel5.
by Yufang Zhang 11 Jan '13

11 Jan '13

On rhel5, libs of avahi are packaged into avahi instead of avahi-libs. Actually, there is no avahi-libs package shipped with rhel5. This patch fixes this by requiring avahi on rhel5. --- libvirt.spec.in | 4 ++++ 1 files changed, 4 insertions(+), 0 deletions(-) diff --git a/libvirt.spec.in b/libvirt.spec.in index d095eac..3362d3f 100644 --- a/libvirt.spec.in +++ b/libvirt.spec.in @@ -615,8 +615,12 @@ Requires: module-init-tools # for /sbin/ip & /sbin/tc Requires: iproute %if %{with_avahi} +%if 0%{?rhel} == 5 +Requires: avahi +%else Requires: avahi-libs %endif +%endif %if %{with_network} Requires: dnsmasq >= 2.41 Requires: radvd -- 1.7.4.1

2 1

[libvirt] another tarball tweak
by Eric Blake 10 Jan '13

10 Jan '13

I tried sending this twice yesterday, but either it's stuck in some long queue, or got eaten along the way by an overactive spam filter, as it hasn't seemed to hit the list yet. Hopefully by sending by hand instead of my normal 'git send-email', I can see it get through. I'll push this under the build-breaker rule (yes, it's a bit of a stretch, since extra files don't really hurt, but I want to avoid the situation of a tarball being generated differently merely because of different configure options). From daa886b635e4a4f46782c8ac6274de1623e12f67 Mon Sep 17 00:00:00 2001 From: Eric Blake <eblake(a)redhat.com> Date: Wed, 9 Jan 2013 15:07:30 -0700 Subject: [PATCH] maint: don't distribute generated .def files I ran 'make dist' in the directory left over from ./autobuild.sh (which was configured for a mingw cross build); the resulting tarball had more files than 'make dist' on a normal Linux build. I traced it to the fact that we were distributing a generated file, but only when configure said the end user had to generate the file in the first place. In the process, I noticed that we had some difference in symbol file names; I added a comment explaining why the difference exists (after first trying to normalize the names and hitting VPATH build failures). * configure.ac (LIBVIRT_QEMU_SYMBOL_FILE): Add some comments. * src/Makefile.am (EXTRA_DIST): No need to ship a generated file; particularly since which file is built depends on configure results. --- configure.ac | 2 ++ src/Makefile.am | 1 - 2 files changed, 2 insertions(+), 1 deletion(-) diff --git a/configure.ac b/configure.ac index 5acba9b..dd28401 100644 --- a/configure.ac +++ b/configure.ac @@ -2676,6 +2676,8 @@ CYGWIN_EXTRA_LIBADD= CYGWIN_EXTRA_PYTHON_LIBADD= MINGW_EXTRA_LDFLAGS= WIN32_EXTRA_CFLAGS= +dnl libvirt.syms is generated in builddir, but libvirt_qemu.syms is in git; +dnl hence the asymmetric naming of these two symbol files. LIBVIRT_SYMBOL_FILE=libvirt.syms LIBVIRT_QEMU_SYMBOL_FILE='$(srcdir)/libvirt_qemu.syms' MSCOM_LIBS= diff --git a/src/Makefile.am b/src/Makefile.am index da571c7..8a4731b 100644 --- a/src/Makefile.am +++ b/src/Makefile.am @@ -1564,7 +1564,6 @@ libvirt_qemu_la_LDFLAGS = $(VERSION_SCRIPT_FLAGS)$(LIBVIRT_QEMU_SYMBOL_FILE) \ $(AM_LDFLAGS) libvirt_qemu_la_CFLAGS = $(AM_CFLAGS) libvirt_qemu_la_LIBADD = libvirt.la $(CYGWIN_EXTRA_LIBADD) -EXTRA_DIST += $(LIBVIRT_QEMU_SYMBOL_FILE) lockdriverdir = $(libdir)/libvirt/lock-driver lockdriver_LTLIBRARIES = -- 1.8.0.2 -- Eric Blake eblake redhat com +1-919-301-3266 Libvirt virtualization library http://libvirt.org

1 0

[libvirt] [PATCH 0/2] spec file cleanups
by Eric Blake 10 Jan '13

10 Jan '13

While working on recent tarball issues, I noticed the spec file was hard to read. In developing the second patch for my own sanity, I found some cleanups that I separated out to the first patch. I don't know if you want to apply the second patch, and even if you do, it might be nice if 'cppi' could be enhanced to enforce indentation rules in spec files the same way it does for preprocessor rules in .c files (maybe even by using a judicious tr to turn %if into #if of a temporary file, so that existing cppi could do it); without enforcement, I'm afraid the second patch will be quickly fall back into hard-to-read status. Eric Blake (2): spec: remove redundant %if spec: indent %if to make it easier to see conditions libvirt.spec.in | 864 ++++++++++++++++++++++++++++---------------------------- 1 file changed, 428 insertions(+), 436 deletions(-) -- 1.8.0.2

4 9

[libvirt] [PATCH] docs: fix typo in isa-serial additions
by Laine Stump 10 Jan '13

10 Jan '13

This was preventing make rpm from completing. --- Pushed under the build breaker rule. docs/formatdomain.html.in | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docs/formatdomain.html.in b/docs/formatdomain.html.in index 90f8abe..bb0b199 100644 --- a/docs/formatdomain.html.in +++ b/docs/formatdomain.html.in @@ -3693,7 +3693,7 @@ qemu-kvm -net nic,model=? /dev/null specifies the port number. Ports are numbered starting from 0. There are usually 0, 1 or 2 serial ports. There is also an optional <code>type</code> attribute <span class="since">since 1.0.2</span> - which has two choices for its value, one is< code>isa-serial</code>, + which has two choices for its value, one is <code>isa-serial</code>, the other is <code>usb-serial</code>. If <code>type</code> is missing, <code>isa-serial</code> will be used by default. For <code>usb-serial</code> an optional sub-element <code><address></code> with -- 1.7.11.7

1 0

[libvirt] [PATCH] util: reduce syscalls for virGetDeviceID
by Eric Blake 10 Jan '13

10 Jan '13

There's no need to do lots of readlink() calls to canonicalize a name if we're only going to use stat() on it, since stat() already chases symlinks. * src/util/virutil.c (virGetDeviceID): Let stat() do the symlink chasing. --- src/util/virutil.c | 15 +++------------ 1 file changed, 3 insertions(+), 12 deletions(-) diff --git a/src/util/virutil.c b/src/util/virutil.c index 47ab17f..7c54bea 100644 --- a/src/util/virutil.c +++ b/src/util/virutil.c @@ -1,7 +1,7 @@ /* * virutil.c: common, generic utility functions * - * Copyright (C) 2006-2012 Red Hat, Inc. + * Copyright (C) 2006-2013 Red Hat, Inc. * Copyright (C) 2006 Daniel P. Berrange * Copyright (C) 2006, 2007 Binary Karma * Copyright (C) 2006 Shuveb Hussain @@ -3135,27 +3135,18 @@ int virGetDeviceID(const char *path, int *maj, int *min) { struct stat sb; - char *canonical_path = NULL; - if (virFileResolveLink(path, &canonical_path) < 0) + if (stat(path, &sb) < 0) return -errno; - if (stat(canonical_path, &sb) < 0) { - VIR_FREE(canonical_path); - return -errno; - } - - if (!S_ISBLK(sb.st_mode)) { - VIR_FREE(canonical_path); + if (!S_ISBLK(sb.st_mode)) return -EINVAL; - } if (maj) *maj = major(sb.st_rdev); if (min) *min = minor(sb.st_rdev); - VIR_FREE(canonical_path); return 0; } #else -- 1.8.0.2

2 2

[libvirt] [PATCH] pass stub driver name instead of pciFindStubDriver
by Chunyan Liu 10 Jan '13

10 Jan '13

Pass stub driver name directly to pciDettachDevice and pciReAttachDevice to fit for different libvirt drivers. For example, qemu driver prefers pci-stub, but Xen prefers pciback. Signed-off-by: Chunyan Liu <cyliu(a)suse.com> --- src/qemu/qemu_driver.c | 4 +- src/qemu/qemu_hostdev.c | 6 ++-- src/util/virpci.c | 62 +++++++++++++++------------------------------- src/util/virpci.h | 6 +++- src/xen/xen_driver.c | 4 +- 5 files changed, 31 insertions(+), 51 deletions(-) diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index 06b0d28..3427d3f 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -10022,7 +10022,7 @@ qemuNodeDeviceDettach(virNodeDevicePtr dev) in_inactive_list = pciDeviceListFind(driver->inactivePciHostdevs, pci); if (pciDettachDevice(pci, driver->activePciHostdevs, - driver->inactivePciHostdevs) < 0) + driver->inactivePciHostdevs, "pci-stub") < 0) goto out; ret = 0; @@ -10067,7 +10067,7 @@ qemuNodeDeviceReAttach(virNodeDevicePtr dev) qemuDriverLock(driver); if (pciReAttachDevice(pci, driver->activePciHostdevs, - driver->inactivePciHostdevs) < 0) + driver->inactivePciHostdevs, "pci-stub") < 0) goto out; ret = 0; diff --git a/src/qemu/qemu_hostdev.c b/src/qemu/qemu_hostdev.c index 174d125..896ad9b 100644 --- a/src/qemu/qemu_hostdev.c +++ b/src/qemu/qemu_hostdev.c @@ -458,7 +458,7 @@ int qemuPrepareHostdevPCIDevices(virQEMUDriverPtr driver, for (i = 0; i < pciDeviceListCount(pcidevs); i++) { pciDevice *dev = pciDeviceListGet(pcidevs, i); if (pciDeviceGetManaged(dev) && - pciDettachDevice(dev, driver->activePciHostdevs, NULL) < 0) + pciDettachDevice(dev, driver->activePciHostdevs, NULL, "pci-stub") < 0) goto reattachdevs; } @@ -574,7 +574,7 @@ resetvfnetconfig: reattachdevs: for (i = 0; i < pciDeviceListCount(pcidevs); i++) { pciDevice *dev = pciDeviceListGet(pcidevs, i); - pciReAttachDevice(dev, driver->activePciHostdevs, NULL); + pciReAttachDevice(dev, driver->activePciHostdevs, NULL, "pci-stub"); } cleanup: @@ -830,7 +830,7 @@ void qemuReattachPciDevice(pciDevice *dev, virQEMUDriverPtr driver) } if (pciReAttachDevice(dev, driver->activePciHostdevs, - driver->inactivePciHostdevs) < 0) { + driver->inactivePciHostdevs, "pci-stub") < 0) { virErrorPtr err = virGetLastError(); VIR_ERROR(_("Failed to re-attach PCI device: %s"), err ? err->message : _("unknown error")); diff --git a/src/util/virpci.c b/src/util/virpci.c index 4b26452..fb305e2 100644 --- a/src/util/virpci.c +++ b/src/util/virpci.c @@ -855,57 +855,35 @@ pciDeviceFile(char **buffer, const char *device, const char *file) return 0; } - -static const char * -pciFindStubDriver(void) +static int +pciProbeStubDriver(const char *driver) { char *drvpath = NULL; int probed = 0; recheck: - if (pciDriverDir(&drvpath, "pci-stub") < 0) { - return NULL; - } - - if (virFileExists(drvpath)) { - VIR_FREE(drvpath); - return "pci-stub"; - } - - if (pciDriverDir(&drvpath, "pciback") < 0) { - return NULL; - } - - if (virFileExists(drvpath)) { + if (pciDriverDir(&drvpath, driver) == 0 && virFileExists(drvpath)) { + /* driver already loaded, return */ VIR_FREE(drvpath); - return "pciback"; + return 0; } VIR_FREE(drvpath); if (!probed) { - const char *const stubprobe[] = { MODPROBE, "pci-stub", NULL }; - const char *const backprobe[] = { MODPROBE, "pciback", NULL }; - + const char *const probecmd[] = { MODPROBE, driver, NULL }; probed = 1; - /* - * Probing for pci-stub will succeed regardless of whether - * on native or Xen kernels. - * On Xen though, we want to prefer pciback, so probe - * for that first, because that will only work on Xen - */ - if (virRun(backprobe, NULL) < 0 && - virRun(stubprobe, NULL) < 0) { + if (virRun(probecmd, NULL) < 0 ) { char ebuf[1024]; - VIR_WARN("failed to load pci-stub or pciback drivers: %s", + VIR_WARN("failed to load driver %s: %s", driver, virStrerror(errno, ebuf, sizeof(ebuf))); - return NULL; + return -1; } goto recheck; } - return NULL; + return -1; } static int @@ -1149,12 +1127,12 @@ cleanup: int pciDettachDevice(pciDevice *dev, pciDeviceList *activeDevs, - pciDeviceList *inactiveDevs) + pciDeviceList *inactiveDevs, + const char *driver) { - const char *driver = pciFindStubDriver(); - if (!driver) { - virReportError(VIR_ERR_INTERNAL_ERROR, "%s", - _("cannot find any PCI stub module")); + if (pciProbeStubDriver(driver) < 0) { + virReportError(VIR_ERR_INTERNAL_ERROR, + _("Failed to load PCI stub module %s"), driver); return -1; } @@ -1179,12 +1157,12 @@ pciDettachDevice(pciDevice *dev, int pciReAttachDevice(pciDevice *dev, pciDeviceList *activeDevs, - pciDeviceList *inactiveDevs) + pciDeviceList *inactiveDevs, + const char *driver) { - const char *driver = pciFindStubDriver(); - if (!driver) { - virReportError(VIR_ERR_INTERNAL_ERROR, "%s", - _("cannot find any PCI stub module")); + if (pciProbeStubDriver(driver) < 0) { + virReportError(VIR_ERR_INTERNAL_ERROR, + _("Failed to load PCI stub module %s"), driver); return -1; } diff --git a/src/util/virpci.h b/src/util/virpci.h index 115da4e..bf7da01 100644 --- a/src/util/virpci.h +++ b/src/util/virpci.h @@ -44,10 +44,12 @@ void pciFreeDevice (pciDevice *dev); const char *pciDeviceGetName (pciDevice *dev); int pciDettachDevice (pciDevice *dev, pciDeviceList *activeDevs, - pciDeviceList *inactiveDevs); + pciDeviceList *inactiveDevs, + const char *driver); int pciReAttachDevice (pciDevice *dev, pciDeviceList *activeDevs, - pciDeviceList *inactiveDevs); + pciDeviceList *inactiveDevs, + const char *driver); int pciResetDevice (pciDevice *dev, pciDeviceList *activeDevs, pciDeviceList *inactiveDevs); diff --git a/src/xen/xen_driver.c b/src/xen/xen_driver.c index 559025e..2795ebc 100644 --- a/src/xen/xen_driver.c +++ b/src/xen/xen_driver.c @@ -2113,7 +2113,7 @@ xenUnifiedNodeDeviceDettach(virNodeDevicePtr dev) if (!pci) return -1; - if (pciDettachDevice(pci, NULL, NULL) < 0) + if (pciDettachDevice(pci, NULL, NULL, "pciback") < 0) goto out; ret = 0; @@ -2203,7 +2203,7 @@ xenUnifiedNodeDeviceReAttach(virNodeDevicePtr dev) goto out; } - if (pciReAttachDevice(pci, NULL, NULL) < 0) + if (pciReAttachDevice(pci, NULL, NULL, "pciback") < 0) goto out; ret = 0; -- 1.7.3.4

3 2

[libvirt] Attaching an "Internal Address"
by Varun Bhatnagar 10 Jan '13

10 Jan '13

Hi, I am trying to attach an internal address to my domain. So for that I am defining the following syntax in my xml file: but it is giving the following error: Tried giving interface type as network but that is adding Host-only type interface. Can anyone please suggest what can be the possible solution. I think it is some problem with the source network. Do I have to define any internal network before in any interface? If yes then how? Any help will be appreciated. Thanks in advance. Regards, Varun =====-----=====-----===== Notice: The information contained in this e-mail message and/or attachments to it may contain confidential or privileged information. If you are not the intended recipient, any dissemination, use, review, distribution, printing or copying of the information contained in this e-mail message and/or attachments to it are strictly prohibited. If you have received this communication in error, please notify us by reply e-mail or telephone and immediately and permanently delete the message and any attachments. Thank you

2 1