[libvirt] [RFC PATCH v3] qemu: Add suport for pci-assign.configfd option

This allows libvirt to open the PCI device sysfs config file prior to dropping privileges so qemu can access the full config space. Without this, a de-privileged qemu can only access the first 64 bytes of config space. Signed-off-by: Alex Williamson <alex.williamson(a)redhat.com&gt; --- Note: this is still dependent on the configfd option being accepted into qemu/kvm. v2: - fix qemuargs, two separate args (thanks Chris) - don't hardcode pci segment - error message of open() fail v3: - add hotplug support - incorporate Daniel's comments src/qemu/qemu_conf.c | 98 +++++++++++++++++++++++++++++++++++++++++++++++- src/qemu/qemu_conf.h | 6 ++- src/qemu/qemu_driver.c | 28 +++++++++++++- 3 files changed, 128 insertions(+), 4 deletions(-) diff --git a/src/qemu/qemu_conf.c b/src/qemu/qemu_conf.c index 2755545..926ce48 100644 --- a/src/qemu/qemu_conf.c +++ b/src/qemu/qemu_conf.c @@ -1350,6 +1350,48 @@ fail: return -1; } +static void qemudParsePCIDeviceStrs(const char *qemu, unsigned long long *flags) +{ + const char *const qemuarg[] = { qemu, "-device", "pci-assign,?", NULL }; + const char *const qemuenv[] = { "LC_ALL=C", NULL }; + pid_t child; + int status; + int newstderr = -1; + + if (virExec(qemuarg, qemuenv, NULL, + &child, -1, NULL, &newstderr, VIR_EXEC_CLEAR_CAPS) < 0) + return; + + char *pciassign = NULL; + enum { MAX_PCI_OUTPUT_SIZE = 1024*4 }; + int len = virFileReadLimFD(newstderr, MAX_PCI_OUTPUT_SIZE, &pciassign); + if (len < 0) { + virReportSystemError(errno, + _("Unable to read %s pci-assign device output"), + qemu); + goto cleanup; + } + + if (strstr(pciassign, "pci-assign.configfd")) + *flags |= QEMUD_CMD_FLAG_PCI_CONFIGFD; + +cleanup: + VIR_FREE(pciassign); + close(newstderr); +rewait: + if (waitpid(child, &status, 0) != child) { + if (errno == EINTR) + goto rewait; + + VIR_ERROR(_("Unexpected exit status from qemu %d pid %lu"), + WEXITSTATUS(status), (unsigned long)child); + } + if (WEXITSTATUS(status) != 0) { + VIR_WARN("Unexpected exit status '%d', qemu probably failed", + WEXITSTATUS(status)); + } +} + int qemudExtractVersionInfo(const char *qemu, unsigned int *retversion, unsigned long long *retflags) { @@ -1383,6 +1425,9 @@ int qemudExtractVersionInfo(const char *qemu, &version, &is_kvm, &kvm_version) == -1) goto cleanup2; + if (flags & QEMUD_CMD_FLAG_DEVICE) + qemudParsePCIDeviceStrs(qemu, &flags); + if (retversion) *retversion = version; if (retflags) @@ -2892,8 +2937,33 @@ error: } +int +qemudOpenPCIConfig(virDomainHostdevDefPtr dev) +{ + char *path = NULL; + int configfd = -1; + + if (virAsprintf(&path, "/sys/bus/pci/devices/%04x:%02x:%02x.%01x/config", + dev->source.subsys.u.pci.domain, + dev->source.subsys.u.pci.bus, + dev->source.subsys.u.pci.slot, + dev->source.subsys.u.pci.function) < 0) { + virReportOOMError(); + return -1; + } + + configfd = open(path, O_RDWR, 0); + + if (configfd < 0) + virReportSystemError(errno, _("Failed opening %s"), path); + + VIR_FREE(path); + + return configfd; +} + char * -qemuBuildPCIHostdevDevStr(virDomainHostdevDefPtr dev) +qemuBuildPCIHostdevDevStr(virDomainHostdevDefPtr dev, const char *configfd) { virBuffer buf = VIR_BUFFER_INITIALIZER; @@ -2903,6 +2973,8 @@ qemuBuildPCIHostdevDevStr(virDomainHostdevDefPtr dev) dev->source.subsys.u.pci.slot, dev->source.subsys.u.pci.function); virBufferVSprintf(&buf, ",id=%s", dev->info.alias); + if (configfd && *configfd) + virBufferVSprintf(&buf, ",configfd=%s", configfd); if (qemuBuildDeviceAddressStr(&buf, &dev->info) < 0) goto error; @@ -4605,8 +4677,30 @@ int qemudBuildCommandLine(virConnectPtr conn, if (hostdev->mode == VIR_DOMAIN_HOSTDEV_MODE_SUBSYS && hostdev->source.subsys.type == VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI) { if (qemuCmdFlags & QEMUD_CMD_FLAG_DEVICE) { + char *configfd_name = NULL; + if (qemuCmdFlags & QEMUD_CMD_FLAG_PCI_CONFIGFD) { + int configfd = qemudOpenPCIConfig(hostdev); + + if (configfd >= 0) { + if (virAsprintf(&configfd_name, "%d", configfd) < 0) { + close(configfd); + virReportOOMError(); + goto no_memory; + } + + if (VIR_REALLOC_N(*vmfds, (*nvmfds)+1) < 0) { + VIR_FREE(configfd_name); + close(configfd); + goto no_memory; + } + + (*vmfds)[(*nvmfds)++] = configfd; + } + } ADD_ARG_LIT("-device"); - if (!(devstr = qemuBuildPCIHostdevDevStr(hostdev))) + devstr = qemuBuildPCIHostdevDevStr(hostdev, configfd_name); + VIR_FREE(configfd_name); + if (!devstr) goto error; ADD_ARG(devstr); } else if (qemuCmdFlags & QEMUD_CMD_FLAG_PCIDEVICE) { diff --git a/src/qemu/qemu_conf.h b/src/qemu/qemu_conf.h index 8fd8d79..d5ce4ce 100644 --- a/src/qemu/qemu_conf.h +++ b/src/qemu/qemu_conf.h @@ -89,6 +89,7 @@ enum qemud_cmd_flags { QEMUD_CMD_FLAG_NO_HPET = (1LL << 33), /* -no-hpet flag is supported */ QEMUD_CMD_FLAG_NO_KVM_PIT = (1LL << 34), /* -no-kvm-pit-reinjection supported */ QEMUD_CMD_FLAG_TDF = (1LL << 35), /* -tdf flag (user-mode pit catchup) */ + QEMUD_CMD_FLAG_PCI_CONFIGFD = (1LL << 36), /* pci-assign.configfd */ }; /* Main driver state */ @@ -245,7 +246,10 @@ char * qemuBuildSoundDevStr(virDomainSoundDefPtr sound); /* Legacy, pre device support */ char * qemuBuildPCIHostdevPCIDevStr(virDomainHostdevDefPtr dev); /* Current, best practice */ -char * qemuBuildPCIHostdevDevStr(virDomainHostdevDefPtr dev); +char * qemuBuildPCIHostdevDevStr(virDomainHostdevDefPtr dev, + const char *configfd); + +int qemudOpenPCIConfig(virDomainHostdevDefPtr dev); /* Current, best practice */ char * qemuBuildChrChardevStr(virDomainChrDefPtr dev); diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index afdc718..52f2374 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -7655,6 +7655,8 @@ static int qemudDomainAttachHostPciDevice(struct qemud_driver *driver, pciDevice *pci; int ret; char *devstr = NULL; + int configfd = -1; + char *configfd_name = NULL; if (VIR_REALLOC_N(vm->def->hostdevs, vm->def->nhostdevs+1) < 0) { virReportOOMError(); @@ -7685,8 +7687,26 @@ static int qemudDomainAttachHostPciDevice(struct qemud_driver *driver, goto error; if (qemuDomainPCIAddressEnsureAddr(priv->pciaddrs, &hostdev->info) < 0) goto error; + if (qemuCmdFlags & QEMUD_CMD_FLAG_PCI_CONFIGFD) { + configfd = qemudOpenPCIConfig(hostdev); + if (configfd >= 0) { + if (virAsprintf(&configfd_name, "fd-%s", + hostdev->info.alias) < 0) { + virReportOOMError(); + goto error; + } + + qemuDomainObjEnterMonitorWithDriver(driver, vm); + if (qemuMonitorSendFileHandle(priv->mon, configfd_name, + configfd) < 0) { + qemuDomainObjExitMonitorWithDriver(driver, vm); + goto error; + } + qemuDomainObjExitMonitorWithDriver(driver, vm); + } + } - if (!(devstr = qemuBuildPCIHostdevDevStr(hostdev))) + if (!(devstr = qemuBuildPCIHostdevDevStr(hostdev, configfd_name))) goto error; qemuDomainObjEnterMonitorWithDriver(driver, vm); @@ -7710,6 +7730,9 @@ static int qemudDomainAttachHostPciDevice(struct qemud_driver *driver, vm->def->hostdevs[vm->def->nhostdevs++] = hostdev; VIR_FREE(devstr); + VIR_FREE(configfd_name); + if (configfd >= 0) + close(configfd); return 0; @@ -7721,6 +7744,9 @@ error: VIR_FREE(devstr); pciDeviceListDel(driver->activePciHostdevs, pci); + VIR_FREE(configfd_name); + if (configfd >= 0) + close(configfd); return -1; }