February 2015 - Devel - libvirt List Archives

[libvirt] [PATCH] SRIOV NIC offload feature discovery
by James Chapman 23 Feb '15

23 Feb '15

Adding functionality to libvirt that will allow it query the ethtool interface for the availability of certain NIC HW offload features --- src/conf/device_conf.h | 6 ++ src/conf/node_device_conf.c | 7 ++ src/conf/node_device_conf.h | 2 + src/libvirt_private.syms | 1 + src/node_device/node_device_udev.c | 4 ++ src/util/virnetdev.c | 134 +++++++++++++++++++++++++++++++++++++ src/util/virnetdev.h | 12 +++- 7 files changed, 165 insertions(+), 1 deletion(-) diff --git a/src/conf/device_conf.h b/src/conf/device_conf.h index 7256cdc..091f2f0 100644 --- a/src/conf/device_conf.h +++ b/src/conf/device_conf.h @@ -62,6 +62,12 @@ struct _virInterfaceLink { unsigned int speed; /* link speed in Mbits per second */ }; +typedef struct _virDevFeature virDevFeature; +typedef virDevFeature *virDevFeaturePtr; +struct _virDevFeature { + char *name; /* device feature */ +}; + int virDevicePCIAddressIsValid(virDevicePCIAddressPtr addr); int virDevicePCIAddressParseXML(xmlNodePtr node, diff --git a/src/conf/node_device_conf.c b/src/conf/node_device_conf.c index a728a00..7f4dbfe 100644 --- a/src/conf/node_device_conf.c +++ b/src/conf/node_device_conf.c @@ -437,6 +437,12 @@ char *virNodeDeviceDefFormat(const virNodeDeviceDef *def) virBufferEscapeString(&buf, "<address>%s</address>\n", data->net.address); virInterfaceLinkFormat(&buf, &data->net.lnk); + if (data->net.features) { + for (i = 0; i < data->net.nfeatures; i++) { + virBufferAsprintf(&buf, "<feature name='%s'/>\n", + data->net.features[i].name); + } + } if (data->net.subtype != VIR_NODE_DEV_CAP_NET_LAST) { const char *subtyp = virNodeDevNetCapTypeToString(data->net.subtype); @@ -1679,6 +1685,7 @@ void virNodeDevCapsDefFree(virNodeDevCapsDefPtr caps) case VIR_NODE_DEV_CAP_NET: VIR_FREE(data->net.ifname); VIR_FREE(data->net.address); + VIR_FREE(data->net.features); break; case VIR_NODE_DEV_CAP_SCSI_HOST: VIR_FREE(data->scsi_host.wwnn); diff --git a/src/conf/node_device_conf.h b/src/conf/node_device_conf.h index fd5d179..918523a 100644 --- a/src/conf/node_device_conf.h +++ b/src/conf/node_device_conf.h @@ -141,6 +141,8 @@ struct _virNodeDevCapsDef { char *ifname; virInterfaceLink lnk; virNodeDevNetCapType subtype; /* LAST -> no subtype */ + size_t nfeatures; + virDevFeaturePtr features; } net; struct { unsigned int host; diff --git a/src/libvirt_private.syms b/src/libvirt_private.syms index c07a561..1d165a9 100644 --- a/src/libvirt_private.syms +++ b/src/libvirt_private.syms @@ -1659,6 +1659,7 @@ virNetDevAddRoute; virNetDevClearIPAddress; virNetDevDelMulti; virNetDevExists; +virNetDevGetFeatures; virNetDevGetIndex; virNetDevGetIPv4Address; virNetDevGetLinkInfo; diff --git a/src/node_device/node_device_udev.c b/src/node_device/node_device_udev.c index 03c7a0b..349733f 100644 --- a/src/node_device/node_device_udev.c +++ b/src/node_device/node_device_udev.c @@ -719,6 +719,10 @@ static int udevProcessNetworkInterface(struct udev_device *device, if (virNetDevGetLinkInfo(data->net.ifname, &data->net.lnk) < 0) goto out; + if (virNetDevGetFeatures(data->net.ifname, &data->net.features, + &data->net.nfeatures) < 0) + goto out; + ret = 0; out: diff --git a/src/util/virnetdev.c b/src/util/virnetdev.c index 2a0eb43..e513c85 100644 --- a/src/util/virnetdev.c +++ b/src/util/virnetdev.c @@ -2728,3 +2728,137 @@ int virNetDevGetRxFilter(const char *ifname, *filter = fil; return ret; } + +#if defined(SIOCETHTOOL) && defined(HAVE_STRUCT_IFREQ) + +/** + * _virNetDevFeatureAvailable + * This function checks for the availability of a network device feature + * + * @ifname: name of the interface + * @cmd: reference to an ethtool command structure + * + * Returns 0 on success, -1 on failure. + */ +int +_virNetDevFeatureAvailable(const char *ifname, struct ethtool_value *cmd) +{ + int ret = -1; + int sock = -1; + virIfreq ifr; + + sock = socket(AF_LOCAL, SOCK_DGRAM, 0); + if (sock < 0) { + virReportSystemError(errno, "%s", _("Cannot open control socket")); + goto cleanup; + } + + strcpy(ifr.ifr_name, ifname); + ifr.ifr_data = (void*) cmd; + + if (ioctl(sock, SIOCETHTOOL, &ifr) != 0) { + /* Privileged command, no error */ + if (errno == EPERM || errno == EINVAL) { + virReportSystemError(errno, "%s", _("ioctl")); + /* Some kernels dont support named feature, no error */ + } else if (errno == EOPNOTSUPP) { + virReportSystemError(errno, "%s", _("Warning")); + } else { + virReportSystemError(errno, "%s", _("Error")); + goto cleanup; + } + } + + ret = cmd->data > 0 ? 1: 0; + cleanup: + if (sock) + VIR_FORCE_CLOSE(sock); + + return ret; +} + + +/** + * virNetDevGetFeatures: + * This function gets the nic offloads features available for ifname + * + * @ifname: name of the interface + * @features: network device feature structures + * @nfeatures: number of features available + * + * Returns 0 on success, -1 on failure. + */ +int +virNetDevGetFeatures(const char *ifname, + virDevFeaturePtr *features, + size_t *nfeatures) +{ + int ret = -1; + size_t i = -1; + size_t j = -1; + struct ethtool_value cmd; + + struct elem{ + const char *name; + const int cmd; + }; + /* legacy ethtool getters */ + struct elem cmds[] = { + {"rx", ETHTOOL_GRXCSUM}, + {"tx", ETHTOOL_GTXCSUM }, + {"sg", ETHTOOL_GSG}, + {"tso", ETHTOOL_GTSO}, + {"gso", ETHTOOL_GGSO}, + {"gro", ETHTOOL_GGRO}, + }; + /* ethtool masks */ + struct elem flags[] = { + {"lro", ETH_FLAG_LRO}, + {"rxvlan", ETH_FLAG_RXVLAN}, + {"txvlan", ETH_FLAG_TXVLAN}, + {"ntuple", ETH_FLAG_NTUPLE}, + {"rxhash", ETH_FLAG_RXHASH}, + }; + + for (i = 0; i < (sizeof(cmds)/sizeof(struct elem)); i++) { + cmd.cmd = cmds[i].cmd; + if (_virNetDevFeatureAvailable(ifname, &cmd)) { + if (VIR_EXPAND_N(*features, *nfeatures, 1) < 0) + goto cleanup; + if ((ret = VIR_STRDUP((*features)[i].name, cmds[i].name)) != 1) + goto cleanup; + } + } + + cmd.cmd = ETHTOOL_GFLAGS; + for (j = 0; j < (sizeof(flags)/sizeof(struct elem)); j++) { + if (_virNetDevFeatureAvailable(ifname, &cmd)) { + if (cmd.data & (flags[j].cmd)) { + if (VIR_EXPAND_N(*features, *nfeatures, 1) < 0) + goto cleanup; + if ((ret = VIR_STRDUP((*features)[i++].name, flags[j].name)) != 1) + goto cleanup; + } + } + } + + ret = 0; + cleanup: + + return ret; + +} +#else +int +virNetDevGetFeatures(const char *ifname, + virDevFeaturePtr *features, + size_t *nfeatures) +{ + VIR_DEBUG("Getting network device features on %s is not implemented on this platform", + ifname); + *features = NULL; + *nfeatures = 0; + + return 0; +} +#endif diff --git a/src/util/virnetdev.h b/src/util/virnetdev.h index de8b480..3847c68 100644 --- a/src/util/virnetdev.h +++ b/src/util/virnetdev.h @@ -31,6 +31,8 @@ # include "virpci.h" # include "device_conf.h" +# include <linux/ethtool.h> +typedef struct ethtool_cmd virEthCmd; # ifdef HAVE_STRUCT_IFREQ typedef struct ifreq virIfreq; # else @@ -182,9 +184,17 @@ int virNetDevGetVirtualFunctionInfo(const char *vfname, char **pfname, int *vf) ATTRIBUTE_NONNULL(1); +int virNetDevGetFeatures(const char *ifname, + virDevFeaturePtr *features, + size_t *nfeatures) + ATTRIBUTE_NONNULL(1); + +int _virNetDevFeatureAvailable(const char *ifname, struct ethtool_value *cmd) + ATTRIBUTE_NONNULL(1) ATTRIBUTE_RETURN_CHECK; + int virNetDevGetLinkInfo(const char *ifname, virInterfaceLinkPtr lnk) - ATTRIBUTE_NONNULL(1); + ATTRIBUTE_NONNULL(1) ATTRIBUTE_RETURN_CHECK; virNetDevRxFilterPtr virNetDevRxFilterNew(void) ATTRIBUTE_RETURN_CHECK; -- 2.1.0

4 4

[libvirt] Exact meaning of "nativeMode" attribute in vlan tags
by Laine Stump 23 Feb '15

23 Feb '15

You'd think that I would know this, since I'm the person who reviewed jrobson's patch adding support for the nativeMode attribute to the vlan tag element. But you'd be wrong. Here is what the config looks like: <vlan trunk='yes'> <tag id='42' nativeMode='untagged'/> <tag id='47'/> </vlan> I understand that trunk='yes' means that packets with any of the tags listed in a <tag> subelement can be sent out this port (and the tag will *not* be removed), and likewise packets arriving into the bridge from the port are allowed to have any of the listed tags (and, again, no tag will be removed). But what exactly do nativeMode='untagged' and nativeMode='tagged' mean? As I understand it, (nativeMode='untagged'|nativeMode='tagged') means that packets (arriving from|sent to) the port (without a tag/with that tag) will be (tagged|untagged). Can someone who fully understands this please select A or B for each of the 4 parenthesized items (in as many permutations as make sense). I guess that in one of the modes, untagged packets going in one direction or the other will be tagged, and vice versa, I just don't know which direction does which, and for which mode, and don't want to guess. (I'm asking this because I want to implement identical functionality for standard Linux host bridges - I want to make sure there are no surprises for people switching between OVS and Linux host bridge implementations).

3 2

[libvirt] [PATCH v2 0/2] Python bindings for IOThreads
by John Ferlan 23 Feb '15

23 Feb '15

v1 here: http://www.redhat.com/archives/libvir-list/2015-February/msg00684.html Changes in v2: * Return an empty list when there are no IOThreads found * Fix alloc/cleanup logic to match review comments * Used "PyObject *error = NULL" like the GetCPUStats code in order to return either NULL for Py*_New() and PyTuple_SetItem() method errors or VIR_PY_NONE for other exceptions * Only make the Py_XDECREF calls on objects which don't get consumed by PyTuple_SetItem * Change name of method from "getIOThreadsInfo" to "ioThreadsInfo" John Ferlan (2): Support virDomainGetIOThreadsInfo and virDomainIOThreadsInfoFree Support virDomainSetIOThreads generator.py | 6 ++ libvirt-override-api.xml | 14 ++++ libvirt-override.c | 180 +++++++++++++++++++++++++++++++++++++++++++++++ sanitytest.py | 5 ++ 4 files changed, 205 insertions(+) -- 2.1.0

1 2

[libvirt] Plan for next release
by Daniel Veillard 23 Feb '15

23 Feb '15

Oops, Feb is really short and I got caught, I think if we want a release by beginning of March, i.e. next week we need to enter freeze ASAP. So I'm suggesting to start the freeze tomorrow, for a release on Monday if everything goes well ! Hope this suits everybody, thanks, Daniel -- Daniel Veillard | Open Source and Standards, Red Hat veillard(a)redhat.com | libxml Gnome XML XSLT toolkit http://xmlsoft.org/ http://veillard.com/ | virtualization library http://libvirt.org/

1 0

[libvirt] [PATCH] qemu: Fix AAVMF/OVMF #define names
by Cole Robinson 21 Feb '15

21 Feb '15

The AAVMF and OVMF names were swapped. Reorder the one usage where it matters so behavior doesn't change. --- src/qemu/qemu_conf.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/src/qemu/qemu_conf.c b/src/qemu/qemu_conf.c index c6b083c..2cf3905 100644 --- a/src/qemu/qemu_conf.c +++ b/src/qemu/qemu_conf.c @@ -149,10 +149,10 @@ virQEMUDriverConfigLoaderNVRAMParse(virQEMUDriverConfigPtr cfg, } -#define VIR_QEMU_OVMF_LOADER_PATH "/usr/share/AAVMF/AAVMF_CODE.fd" -#define VIR_QEMU_OVMF_NVRAM_PATH "/usr/share/AAVMF/AAVMF_VARS.fd" -#define VIR_QEMU_AAVMF_LOADER_PATH "/usr/share/OVMF/OVMF_CODE.fd" -#define VIR_QEMU_AAVMF_NVRAM_PATH "/usr/share/OVMF/OVMF_VARS.fd" +#define VIR_QEMU_OVMF_LOADER_PATH "/usr/share/OVMF/OVMF_CODE.fd" +#define VIR_QEMU_OVMF_NVRAM_PATH "/usr/share/OVMF/OVMF_VARS.fd" +#define VIR_QEMU_AAVMF_LOADER_PATH "/usr/share/AAVMF/AAVMF_CODE.fd" +#define VIR_QEMU_AAVMF_NVRAM_PATH "/usr/share/AAVMF/AAVMF_VARS.fd" virQEMUDriverConfigPtr virQEMUDriverConfigNew(bool privileged) { @@ -313,10 +313,10 @@ virQEMUDriverConfigPtr virQEMUDriverConfigNew(bool privileged) goto error; cfg->nloader = 2; - if (VIR_STRDUP(cfg->loader[0], VIR_QEMU_OVMF_LOADER_PATH) < 0 || - VIR_STRDUP(cfg->nvram[0], VIR_QEMU_OVMF_NVRAM_PATH) < 0 || - VIR_STRDUP(cfg->loader[1], VIR_QEMU_AAVMF_LOADER_PATH) < 0 || - VIR_STRDUP(cfg->nvram[1], VIR_QEMU_AAVMF_NVRAM_PATH) < 0) + if (VIR_STRDUP(cfg->loader[0], VIR_QEMU_AAVMF_LOADER_PATH) < 0 || + VIR_STRDUP(cfg->nvram[0], VIR_QEMU_AAVMF_NVRAM_PATH) < 0 || + VIR_STRDUP(cfg->loader[1], VIR_QEMU_OVMF_LOADER_PATH) < 0 || + VIR_STRDUP(cfg->nvram[1], VIR_QEMU_OVMF_NVRAM_PATH) < 0) goto error; #endif -- 2.1.0

2 2

[libvirt] [PATCH 0/3] Fix memory ABI stability check issues
by Peter Krempa 20 Feb '15

20 Feb '15

Note that this series applies only on top of the NUMA config unification series: http://www.redhat.com/archives/libvir-list/2015-February/msg00532.html Please see individual patches for explanation Peter Krempa (3): conf: ABI: Hugepage backing definition is not guest ABI conf: ABI: Memballoon setting is not guest ABI conf: numa: Check ABI stability of NUMA configuration src/conf/domain_conf.c | 31 ++----------------------------- src/conf/numa_conf.c | 37 +++++++++++++++++++++++++++++++++++++ src/conf/numa_conf.h | 3 +++ src/libvirt_private.syms | 1 + 4 files changed, 43 insertions(+), 29 deletions(-) -- 2.2.2

3 9

[libvirt] [PATCH v2] qemu: Check for negative port values in network drive configuration
by Erik Skultety 20 Feb '15

20 Feb '15

We interpret port values as signed int (convert them from char *), so if anegative value is provided in network disk's configuration, we accept it as valid, however there's an 'unknown cause' error raised later. This error is only accidental because we return the port value in the return code. This patch adds just a minor tweak to the already existing check so we reject negative values the same way as we reject non-numerical strings. Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1163553 --- src/qemu/qemu_command.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/src/qemu/qemu_command.c b/src/qemu/qemu_command.c index 743d6f0..6941b5a 100644 --- a/src/qemu/qemu_command.c +++ b/src/qemu/qemu_command.c @@ -2951,10 +2951,10 @@ static int qemuNetworkDriveGetPort(int protocol, const char *port) { - int ret = 0; - + unsigned ret = 0; if (port) { - if (virStrToLong_i(port, NULL, 10, &ret) < 0) { + if (virStrToLong_uip(port, NULL, 10, &ret) < 0 || + (int) ret < 0) { virReportError(VIR_ERR_INTERNAL_ERROR, _("failed to parse port number '%s'"), port); -- 1.9.3

2 1

[libvirt] [PATCH 00/24] Move all NUMA related configuration into one structure
by Peter Krempa 20 Feb '15

20 Feb '15

Due to historical madne^Wreasons the NUMA configuration is split between /domain/cpu and /domain/numatune. Internally the data was also split into two places. We can't do anything about the external representation but we certainly can store all the definitions in one place internally. This series does that. Peter Krempa (24): conf: Move numatune_conf to numa_conf conf: Move NUMA cell parsing code from cpu conf to numa conf conf: Refactor virDomainNumaDefCPUParseXML conf: numa: Don't duplicate NUMA cell cpumask conf: Move NUMA cell formatter to numa_conf conf: numa: Rename virDomainNumatune to virDomainNuma conf: Move enum virMemAccess to the NUMA code and rename it conf: numa: Recalculate rather than remember total NUMA cpu count conf: numa: Improve error message in case a numa node doesn't have cpus conf: numa: Reformat virDomainNumatuneParseXML conf: numa: Refactor logic in virDomainNumatuneParseXML conf: numa: Format <numatune> XML only if necessary conf: Separate helper for creating domain objects conf: Allocate domain definition with the new helper conf: numa: Always allocate the NUMA config conf: numa: Avoid re-allocation of the NUMA conf numa: conf: Tweak parameters of virDomainNumatuneSet conf: numa: Don't pass double pointer to virDomainNumatuneParseXML qemu: command: Unify retrieval of NUMA cell count in qemuBuildNumaArgStr conf: numa: Add helper to get guest NUMA node count and refactor users conf: numa: Add accesor for the NUMA node cpu mask conf: numa: Add accessor to NUMA node's memory access mode conf: numa: Add setter/getter for NUMA node memory size conf: Move all NUMA configuration to virDomainNuma po/POTFILES.in | 2 +- src/Makefile.am | 2 +- src/conf/cpu_conf.c | 151 +------- src/conf/cpu_conf.h | 25 +- src/conf/domain_conf.c | 59 ++- src/conf/domain_conf.h | 11 +- src/conf/{numatune_conf.c => numa_conf.c} | 429 +++++++++++++++------ src/conf/{numatune_conf.h => numa_conf.h} | 77 ++-- src/cpu/cpu.c | 2 +- src/libvirt_private.syms | 13 +- src/lxc/lxc_cgroup.c | 4 +- src/lxc/lxc_controller.c | 6 +- src/lxc/lxc_native.c | 4 +- src/openvz/openvz_conf.c | 2 +- src/parallels/parallels_sdk.c | 4 +- src/phyp/phyp_driver.c | 2 +- src/qemu/qemu_cgroup.c | 12 +- src/qemu/qemu_command.c | 52 +-- src/qemu/qemu_driver.c | 20 +- src/qemu/qemu_process.c | 4 +- src/vbox/vbox_common.c | 8 +- src/vmx/vmx.c | 2 +- src/xen/xen_hypervisor.c | 8 +- src/xen/xend_internal.c | 4 +- src/xen/xm_internal.c | 4 +- src/xenconfig/xen_sxpr.c | 2 +- src/xenconfig/xen_xl.c | 2 +- src/xenconfig/xen_xm.c | 2 +- tests/cputest.c | 2 +- tests/openvzutilstest.c | 2 +- .../qemuxml2argv-numatune-memnode.xml | 2 +- .../qemuxml2xmlout-numatune-memnode.xml | 2 +- tests/securityselinuxtest.c | 2 +- tests/testutilsqemu.c | 4 - 34 files changed, 503 insertions(+), 424 deletions(-) rename src/conf/{numatune_conf.c => numa_conf.c} (60%) rename src/conf/{numatune_conf.h => numa_conf.h} (50%) -- 2.2.2

2 58

[libvirt] [PATCH v2] libvirt-guests: Allow time sync on guests resume
by Michal Privoznik 20 Feb '15

20 Feb '15

Well, imagine domains were running, and as the host went down, they were managesaved. Later, after some time, the host went up again and domains got restored. But without correct time. And depending on how long was the host shut off, it may take some time for ntp to sync the time too. But hey, wait a minute. We have an API just for that! So: 1) Introduce SYNC_TIME variable in libvirt-guests.sysconf to allow users control over the new functionality 2) Call 'virsh domtime --sync $dom' in the libvirt-guests script. Unfortunately, this is all-or-nothing approach (just like anything else with the script). Domains are required to have configured and running qemu-ga inside. Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com> --- diff to v1: -ignore the return value of domtime command tools/libvirt-guests.sh.in | 5 +++++ tools/libvirt-guests.sysconf | 7 +++++++ 2 files changed, 12 insertions(+) diff --git a/tools/libvirt-guests.sh.in b/tools/libvirt-guests.sh.in index 1b17bbe..9aa06fa 100644 --- a/tools/libvirt-guests.sh.in +++ b/tools/libvirt-guests.sh.in @@ -171,7 +171,9 @@ start() { isfirst=true bypass= + sync_time=false test "x$BYPASS_CACHE" = x0 || bypass=--bypass-cache + test "x$SYNC_TIME" = x0 || sync_time=true while read uri list; do configured=false set -f @@ -206,6 +208,9 @@ start() { retval run_virsh "$uri" start $bypass "$name" \ >/dev/null && \ gettext "done"; echo + if "$sync_time"; then + run_virsh "$uri" domtime --sync "$name" >/dev/null + fi fi fi done diff --git a/tools/libvirt-guests.sysconf b/tools/libvirt-guests.sysconf index d1f2051..03e732f 100644 --- a/tools/libvirt-guests.sysconf +++ b/tools/libvirt-guests.sysconf @@ -39,3 +39,10 @@ # restoring guests, even though this may give slower operation for # some file systems. #BYPASS_CACHE=0 + +# If non-zero, try to sync guest time on a domain resume. Be aware, that +# this requires guest agent, which, moreover, has to run under supported +# system. For instance, qemu-ga doesn't support guest time synchronization +# on Windows guests, but Linux ones. By default, this piece of +# functionality is turned off. +#SYNC_TIME=1 -- 2.0.5

2 1

[libvirt] [PATCH 0/3] Introduce machine type based capabilities filtering
by Michal Privoznik 20 Feb '15

20 Feb '15

Well, here's an alternative approach to my question here [1]. 1: https://www.redhat.com/archives/libvir-list/2015-February/msg00369.html Michal Privoznik (3): virQEMUCapsCacheLookupCopy: Pass machine type virQEMUCapsCacheLookupCopy: Filter qemuCaps based on machineType qemuCaps: Disable memdev for rhel6.5.0 machine type src/qemu/qemu_capabilities.c | 45 +++++++++++++++++++++- src/qemu/qemu_capabilities.h | 5 ++- src/qemu/qemu_migration.c | 3 +- src/qemu/qemu_process.c | 9 +++-- .../qemuxml2argv-numatune-memnode-rhel650.args | 7 ++++ .../qemuxml2argv-numatune-memnode-rhel650.xml | 31 +++++++++++++++ tests/qemuxml2argvtest.c | 6 +++ 7 files changed, 100 insertions(+), 6 deletions(-) create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-numatune-memnode-rhel650.args create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-numatune-memnode-rhel650.xml -- 2.0.5

3 10