- Devel - libvirt List Archives

[libvirt] [PATCH] Doc fix of /etc/libvirt/libvirt-guests
by Lily Zhu 21 Aug '17

21 Aug '17

As in previous doc, it may cause confusion about whether parallel_shutdown would take effect when guests are asked to suspend on host shutdown. Hence, changed the doc of parallel_shutdown, make it more clear. Signed-off-by: Lily Zhu <lizhu(a)redhat.com> --- tools/libvirt-guests.sysconf | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/tools/libvirt-guests.sysconf b/tools/libvirt-guests.sysconf index 2770056..785c3c0 100644 --- a/tools/libvirt-guests.sysconf +++ b/tools/libvirt-guests.sysconf @@ -24,8 +24,10 @@ # value suitable for your guests. #ON_SHUTDOWN=suspend -# If set to non-zero, shutdown will suspend guests concurrently. Number of -# guests on shutdown at any time will not exceed number set in this variable. +# Number of guests will be shutdown concurrently, taking effect when +# "ON_SHUTDOWN" is set to "shutdown". If Set to 0, guests will be +# shutdown one after another. Number of guests on shutdown at any +# time will not exceed number set in this variable. #PARALLEL_SHUTDOWN=0 # Number of seconds we're willing to wait for a guest to shut down. If parallel -- 1.8.3.1

2 1

[libvirt] [PATCH] docs: Improve PCI topology and hotplug guidelines
by Andrea Bolognani 21 Aug '17

21 Aug '17

Address some minor flaws in the original document that were pointed out during review. Signed-off-by: Andrea Bolognani <abologna(a)redhat.com> --- docs/pci-hotplug.html.in | 46 ++++++++++++++++++++++++++++++++-------------- 1 file changed, 32 insertions(+), 14 deletions(-) diff --git a/docs/pci-hotplug.html.in b/docs/pci-hotplug.html.in index 809e36f5d..a90d47bc3 100644 --- a/docs/pci-hotplug.html.in +++ b/docs/pci-hotplug.html.in @@ -13,10 +13,12 @@ <p> The reason for this apparent limitation is the fact that each hotplugged PCI device might require additional PCI controllers to - be added to the guest, and libvirt has no way of knowing in advance - how many devices will be hotplugged during the guest's lifetime, - thus making it impossible to automatically provide the right amount - of PCI controllers: any arbitrary number would end up being too big + be added to the guest. Since most PCI controllers can't be + hotplugged, they need to be added before the guest is started; + however, libvirt has no way of knowing in advance how many devices + will be hotplugged during the guest's lifetime, thus making it + impossible to automatically provide the right amount of PCI + controllers: any arbitrary number would end up being too big for some users, and too small for others. </p> <p> @@ -53,6 +55,14 @@ emulated or assigned from the host. </p> <p> + If you have a very specialized use case, such as the appliances + used by <a href="http://libguestfs.org/">libguestfs</a> behind + the scenes to access disk images, and this automatically-added + <code>pcie-root-port</code> controller ends up being a nuisance, + you can prevent libvirt from adding it by manually managing PCI + controllers and addresses according to you needs. + </p> + <p> Slots on the <code>pcie-root</code> controller do not support hotplug, so the device will be hotplugged into the <code>pcie-root-port</code> controller. If you plan to hotplug @@ -73,6 +83,12 @@ remaining details automatically. </p> <p> + Note that if you're adding PCI controllers to a guest at the + same time you're also adding PCI devices, some of the + controllers will be used for the newly-added devices and won't + be available for hotplug once the guest has been started. + </p> + <p> If you expect to hotplug legacy PCI devices, then you will need specialized controllers, since all those mentioned above are intended for PCI Express devices only: add @@ -84,7 +100,8 @@ <p> and you'll be able to hotplug up to 31 legacy PCI devices, - either emulated or assigned from the host. + either emulated or assigned from the host, in the slots + from 0x01 to 0x1f of the <code>pci-bridge</code> controller. </p> <h3><a name="x86_64-i440fx">i440fx (pc) machine type</a></h3> @@ -98,9 +115,10 @@ <controller type='pci' index='0' model='pci-root'/></pre> <p> - where each of the 31 slots on the <code>pci-root</code> - controller is hotplug capable and can accept a legacy PCI - device, either emulated or assigned from the guest. + where each of the 31 slots (from 0x01 to 0x1f) on the + <code>pci-root</code> controller is hotplug capable and + can accept a legacy PCI device, either emulated or + assigned from the guest. </p> <h2><a name="ppc64">ppc64 architecture</a></h2> @@ -119,12 +137,12 @@ </controller></pre> <p> - The 31 slots on a <code>pci-root</code> controller are all - hotplug capable and, despite the name suggesting otherwise, - starting with QEMU 2.9 all of them can accept PCI Express - devices in addition to legacy PCI devices; however, - libvirt will only place emulated devices on the default - <code>pci-root</code> controller. + The 31 slots, from 0x01 to 0x1f, on a <code>pci-root</code> + controller are all hotplug capable and, despite the name + suggesting otherwise, starting with QEMU 2.9 all of them + can accept PCI Express devices in addition to legacy PCI + devices; however, libvirt will only place emulated devices + on the default <code>pci-root</code> controller. </p> <p> In order to take advantage of improved error reporting and -- 2.13.5

2 2

[libvirt] [PATCH] virsh: Honour --readonly with cmdConnect and no name
by Martin Kletzander 21 Aug '17

21 Aug '17

Signed-off-by: Martin Kletzander <mkletzan(a)redhat.com> --- tools/virsh.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) diff --git a/tools/virsh.c b/tools/virsh.c index 94bb7ff6c93a..9883e87df2b5 100644 --- a/tools/virsh.c +++ b/tools/virsh.c @@ -218,7 +218,13 @@ virshReconnect(vshControl *ctl, const char *name, bool readonly, bool force) { bool connected = false; virshControlPtr priv = ctl->privData; - bool ro = name ? readonly : priv->readonly; + + /* If the flag was not specified, then it depends on whether we are + * reconnecting to the default URI (in which case we want to keep the + * readonly flag as it was) or to a specified URI in which case it + * should stay false */ + if (!readonly && !name) + readonly = priv->readonly; if (priv->conn) { int ret; @@ -233,7 +239,7 @@ virshReconnect(vshControl *ctl, const char *name, bool readonly, bool force) "disconnect from the hypervisor")); } - priv->conn = virshConnect(ctl, name ? name : ctl->connname, ro); + priv->conn = virshConnect(ctl, name ? name : ctl->connname, readonly); if (!priv->conn) { if (disconnected) -- 2.14.1

2 3

[libvirt] [PATCH] nodedev: add switchdev to NIC capabilities
by Edan David 21 Aug '17

21 Aug '17

Adding functionality to libvirt that will allow it query the interface for the availability of switchdev Offloading NIC capabilities --- configure.ac | 13 ++ docs/formatnode.html.in | 1 + src/util/virnetdev.c | 204 +++++++++++++++++++++- src/util/virnetdev.h | 1 + tests/nodedevschemadata/net_00_13_02_b9_f9_d3.xml | 1 + tests/nodedevschemadata/net_00_15_58_2f_e9_55.xml | 1 + 6 files changed, 220 insertions(+), 1 deletion(-) diff --git a/configure.ac b/configure.ac index afacf40..a050b99 100644 --- a/configure.ac +++ b/configure.ac @@ -627,6 +627,19 @@ if test "$with_linux" = "yes"; then AC_CHECK_HEADERS([linux/btrfs.h]) fi +dnl +dnl check for kernel headers required by devlink +dnl +if test "$with_linux" = "yes"; then + AC_CHECK_HEADERS([linux/devlink.h]) + AC_CHECK_DECLS([DEVLINK_GENL_VERSION, DEVLINK_GENL_NAME, DEVLINK_ATTR_MAX, DEVLINK_CMD_ESWITCH_GET, DEVLINK_ATTR_BUS_NAME, DEVLINK_ATTR_DEV_NAME, DEVLINK_ATTR_ESWITCH_MODE, DEVLINK_ESWITCH_MODE_SWITCHDEV], + [AC_DEFINE([HAVE_DECL_DEVLINK], + [1], + [whether devlink declarations is available])], + [], + [[#include <linux/devlink.h>]]) +fi + dnl Allow perl/python overrides AC_PATH_PROGS([PYTHON], [python2 python]) if test -z "$PYTHON"; then diff --git a/docs/formatnode.html.in b/docs/formatnode.html.in index 32451d5..e7b30ea 100644 --- a/docs/formatnode.html.in +++ b/docs/formatnode.html.in @@ -227,6 +227,7 @@ <dt><code>rxhash</code></dt><dd>receive-hashing</dd> <dt><code>rdma</code></dt><dd>remote-direct-memory-access</dd> <dt><code>txudptnl</code></dt><dd>tx-udp-tunnel-segmentation</dd> + <dt><code>switchdev</code></dt><dd>kernel-forward-plane-offload</dd> </dl> </dd> <dt><code>capability</code></dt> diff --git a/src/util/virnetdev.c b/src/util/virnetdev.c index 90b7bee..084fb41 100644 --- a/src/util/virnetdev.c +++ b/src/util/virnetdev.c @@ -59,6 +59,10 @@ # include <net/if_dl.h> #endif +#if HAVE_DECL_DEVLINK +# include <linux/devlink.h> +#endif + #ifndef IFNAMSIZ # define IFNAMSIZ 16 #endif @@ -95,6 +99,7 @@ VIR_LOG_INIT("util.netdev"); (FEATURE_WORD(blocks, index, field) & FEATURE_FIELD_FLAG(index)) #endif + typedef enum { VIR_MCAST_TYPE_INDEX_TOKEN, VIR_MCAST_TYPE_NAME_TOKEN, @@ -2396,7 +2401,8 @@ VIR_ENUM_IMPL(virNetDevFeature, "ntuple", "rxhash", "rdma", - "txudptnl") + "txudptnl", + "switchdev") #ifdef __linux__ int @@ -2851,6 +2857,199 @@ int virNetDevGetRxFilter(const char *ifname, return ret; } + + +#if HAVE_DECL_DEVLINK +/** + * virNetDevPutExtraHeader + * reserve and prepare room for an extra header + * This function sets to zero the room that is required to put the extra + * header after the initial Netlink header. This function also increases + * the nlmsg_len field. You have to invoke mnl_nlmsg_put_header() before + * you call this function. This function returns a pointer to the extra + * header. + * + * @nlh: pointer to Netlink header + * @size: size of the extra header that we want to put + * + * Returns pointer to the start of the extended header + */ +static void * +virNetDevPutExtraHeader(struct nlmsghdr *nlh, size_t size) +{ + char *ptr = (char *)nlh + nlh->nlmsg_len; + size_t len = NLMSG_ALIGN(size); + nlh->nlmsg_len += len; + memset(ptr, 0, len); + return ptr; +} + + +/** + * virNetDevGetFamilyId: + * This function supplies the devlink family id + * + * @family_name: the name of the family to query + * + * Returns family id or 0 on failure. + */ +static int +virNetDevGetFamilyId(const char *family_name) +{ + struct nl_msg *nl_msg = NULL; + struct nlmsghdr *resp = NULL; + struct genlmsghdr* gmsgh = NULL; + struct nlattr *tb[CTRL_ATTR_MAX + 1] = {NULL, }; + unsigned int recvbuflen; + uint32_t family_id = 0; + + if (!(nl_msg = nlmsg_alloc_simple(GENL_ID_CTRL, + NLM_F_REQUEST | NLM_F_ACK))) { + virReportOOMError(); + goto cleanup; + } + + gmsgh = virNetDevPutExtraHeader(nlmsg_hdr(nl_msg), sizeof(struct genlmsghdr)); + if (!gmsgh) + goto cleanup; + + gmsgh->cmd = CTRL_CMD_GETFAMILY; + gmsgh->version = DEVLINK_GENL_VERSION; + + if (nla_put_string(nl_msg, CTRL_ATTR_FAMILY_NAME, family_name) < 0) + goto buffer_too_small; + + if (virNetlinkCommand(nl_msg, &resp, &recvbuflen, 0, 0, NETLINK_GENERIC, 0) < 0) + goto cleanup; + + if (nlmsg_parse(resp, sizeof(struct nlmsghdr), tb, CTRL_CMD_MAX, NULL) < 0) + goto malformed_resp; + + if (tb[CTRL_ATTR_FAMILY_ID] == NULL) + goto cleanup; + + family_id = *(int *)RTA_DATA(tb[CTRL_ATTR_FAMILY_ID]); + + cleanup: + nlmsg_free(nl_msg); + VIR_FREE(resp); + return family_id; + + malformed_resp: + virReportError(VIR_ERR_INTERNAL_ERROR, "%s", + _("malformed netlink response message")); + goto cleanup; + + buffer_too_small: + virReportError(VIR_ERR_INTERNAL_ERROR, "%s", + _("allocated netlink buffer is too small")); + goto cleanup; +} + + +/** + * virNetDevSwitchdevFeature + * This function checks for the availability of Switchdev feature + * and add it to bitmap + * + * @ifname: name of the interface + * @out: add Switchdev feature if exist to bitmap + * + * Returns 0 on success, -1 on failure. + */ +static int +virNetDevSwitchdevFeature(const char *ifname, + virBitmapPtr *out) +{ + struct nl_msg *nl_msg = NULL; + struct nlmsghdr *resp = NULL; + unsigned int recvbuflen; + struct nlattr *tb[DEVLINK_ATTR_MAX + 1] = {NULL, }; + virPCIDevicePtr pci_device_ptr = NULL; + struct genlmsghdr* gmsgh = NULL; + const char *pci_name; + char *pfname = NULL; + int func_ret_val = -1; + int ret = -1; + + int family_id = virNetDevGetFamilyId(DEVLINK_GENL_NAME); + if (family_id == 0) + goto cleanup; + + func_ret_val = virNetDevIsVirtualFunction(ifname); + if (func_ret_val == 1) { + if (virNetDevGetPhysicalFunction(ifname, &pfname) < 0) + goto cleanup; + } + else if (func_ret_val < 0) + goto cleanup; + + if (!(nl_msg = nlmsg_alloc_simple(family_id, + NLM_F_REQUEST | NLM_F_ACK))) { + virReportOOMError(); + goto cleanup; + } + + pci_device_ptr = pfname ? virNetDevGetPCIDevice(pfname) : + virNetDevGetPCIDevice(ifname); + if (!pci_device_ptr) + goto cleanup; + + pci_name = virPCIDeviceGetName(pci_device_ptr); + + gmsgh = virNetDevPutExtraHeader(nlmsg_hdr(nl_msg), sizeof(struct genlmsghdr)); + if (!gmsgh) + goto cleanup; + + gmsgh->cmd = DEVLINK_CMD_ESWITCH_GET; + gmsgh->version = DEVLINK_GENL_VERSION; + + if (nla_put(nl_msg, DEVLINK_ATTR_BUS_NAME, strlen("pci")+1, "pci") < 0) + goto buffer_too_small; + + if (nla_put(nl_msg, DEVLINK_ATTR_DEV_NAME, strlen(pci_name)+1, pci_name) < 0) + goto buffer_too_small; + + if (virNetlinkCommand(nl_msg, &resp, &recvbuflen, 0, 0, + NETLINK_GENERIC, 0) < 0) + goto cleanup; + + if (nlmsg_parse(resp, sizeof(struct genlmsghdr), tb, DEVLINK_ATTR_MAX, NULL) < 0) + goto malformed_resp; + + if (tb[DEVLINK_ATTR_ESWITCH_MODE] && + *(int *)RTA_DATA(tb[DEVLINK_ATTR_ESWITCH_MODE]) == DEVLINK_ESWITCH_MODE_SWITCHDEV) { + ignore_value(virBitmapSetBit(*out, VIR_NET_DEV_FEAT_SWITCHDEV)); + } + + ret = 0; + + cleanup: + nlmsg_free(nl_msg); + virPCIDeviceFree(pci_device_ptr); + VIR_FREE(resp); + return ret; + + malformed_resp: + virReportError(VIR_ERR_INTERNAL_ERROR, "%s", + _("malformed netlink response message")); + goto cleanup; + + buffer_too_small: + virReportError(VIR_ERR_INTERNAL_ERROR, "%s", + _("allocated netlink buffer is too small")); + goto cleanup; +} +# else +static int +virNetDevSwitchdevFeature(const char *ifname, + virBitmapPtr *out) +{ + return 0; +} +# endif + + #if defined(SIOCETHTOOL) && defined(HAVE_STRUCT_IFREQ) /** @@ -3230,6 +3429,9 @@ virNetDevGetFeatures(const char *ifname, if (virNetDevRDMAFeature(ifname, out) < 0) goto cleanup; + if (virNetDevSwitchdevFeature(ifname, out) < 0) + goto cleanup; + ret = 0; cleanup: VIR_FORCE_CLOSE(fd); diff --git a/src/util/virnetdev.h b/src/util/virnetdev.h index 2e9a9c4..8fd6036 100644 --- a/src/util/virnetdev.h +++ b/src/util/virnetdev.h @@ -112,6 +112,7 @@ typedef enum { VIR_NET_DEV_FEAT_RXHASH, VIR_NET_DEV_FEAT_RDMA, VIR_NET_DEV_FEAT_TXUDPTNL, + VIR_NET_DEV_FEAT_SWITCHDEV, VIR_NET_DEV_FEAT_LAST } virNetDevFeature; diff --git a/tests/nodedevschemadata/net_00_13_02_b9_f9_d3.xml b/tests/nodedevschemadata/net_00_13_02_b9_f9_d3.xml index d4c96e8..88252e6 100644 --- a/tests/nodedevschemadata/net_00_13_02_b9_f9_d3.xml +++ b/tests/nodedevschemadata/net_00_13_02_b9_f9_d3.xml @@ -15,6 +15,7 @@ <feature name='rxhash'/> <feature name='rdma'/> <feature name='txudptnl'/> + <feature name='switchdev'/> <capability type='80211'/> </capability> </device> diff --git a/tests/nodedevschemadata/net_00_15_58_2f_e9_55.xml b/tests/nodedevschemadata/net_00_15_58_2f_e9_55.xml index 71bf90e..f77dfcc 100644 --- a/tests/nodedevschemadata/net_00_15_58_2f_e9_55.xml +++ b/tests/nodedevschemadata/net_00_15_58_2f_e9_55.xml @@ -15,6 +15,7 @@ <feature name='rxhash'/> <feature name='rdma'/> <feature name='txudptnl'/> + <feature name='switchdev'/> <capability type='80203'/> </capability> </device> -- 2.1.4

2 2

[libvirt] [PATCH 0/3] tests: Make hash table mocking arch independent
by Peter Krempa 21 Aug '17

21 Aug '17

Plus one fix. Peter Krempa (3): util: hash: Include stdbool.h in the header file util: hash: Make virHashCodeGen mockable tests: deterministichash: Make hash tables arch-independent src/libvirt_private.syms | 4 ++++ src/util/virhash.h | 1 + src/util/virhashcode.h | 3 ++- .../qemumonitorjson-nodename-relative.result | 24 +++++++++++----------- .../qemumonitorjson-nodename-same-backing.result | 24 +++++++++++----------- tests/virdeterministichashmock.c | 17 +++++++++++---- tests/virmacmaptestdata/simple2.json | 12 +++++------ 7 files changed, 50 insertions(+), 35 deletions(-) -- 2.13.2

6 18

[libvirt] [PATCH] qemu: don't check whether offline migration is safe
by Pavel Hrdina 18 Aug '17

18 Aug '17

Offline migration transfers only the domain definition. Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1449715 Signed-off-by: Pavel Hrdina <phrdina(a)redhat.com> --- src/qemu/qemu_migration.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/src/qemu/qemu_migration.c b/src/qemu/qemu_migration.c index 056c051b3e..ca1f67146b 100644 --- a/src/qemu/qemu_migration.c +++ b/src/qemu/qemu_migration.c @@ -1946,7 +1946,7 @@ qemuMigrationBeginPhase(virQEMUDriverPtr driver, if (!qemuMigrationIsAllowed(driver, vm, true, flags)) goto cleanup; - if (!(flags & VIR_MIGRATE_UNSAFE) && + if (!(flags & (VIR_MIGRATE_UNSAFE | VIR_MIGRATE_OFFLINE)) && !qemuMigrationIsSafe(vm->def, nmigrate_disks, migrate_disks, flags)) goto cleanup; @@ -4809,7 +4809,7 @@ qemuMigrationPerformJob(virQEMUDriverPtr driver, if (!qemuMigrationIsAllowed(driver, vm, true, flags)) goto endjob; - if (!(flags & VIR_MIGRATE_UNSAFE) && + if (!(flags & (VIR_MIGRATE_UNSAFE | VIR_MIGRATE_OFFLINE)) && !qemuMigrationIsSafe(vm->def, nmigrate_disks, migrate_disks, flags)) goto endjob; -- 2.13.5

3 3

[libvirt] [PATCH v2 0/3] conf: Use the correct limit for the number of PHBs
by Andrea Bolognani 18 Aug '17

18 Aug '17

Andrea Bolognani (3): tests: Improve target index validation coverage conf: Move target index validation conf: Use the correct limit for the number of PHBs src/conf/domain_conf.c | 40 +++++++++++++--------- ...ml2argv-pseries-phb-invalid-target-index-1.xml} | 4 --- ...ml2argv-pseries-phb-invalid-target-index-2.xml} | 6 +--- ...ml2argv-pseries-phb-invalid-target-index-3.xml} | 11 +++--- tests/qemuxml2argvtest.c | 4 ++- 5 files changed, 32 insertions(+), 33 deletions(-) copy tests/qemuxml2argvdata/{qemuxml2argv-pseries-phb-wrong-target-index.xml => qemuxml2argv-pseries-phb-invalid-target-index-1.xml} (79%) copy tests/qemuxml2argvdata/{qemuxml2argv-pseries-phb-wrong-target-index.xml => qemuxml2argv-pseries-phb-invalid-target-index-2.xml} (68%) rename tests/qemuxml2argvdata/{qemuxml2argv-pseries-phb-wrong-target-index.xml => qemuxml2argv-pseries-phb-invalid-target-index-3.xml} (57%) -- 2.13.5

2 4

[libvirt] [PATCH] util: string: Introduce virStringHasChars
by Peter Krempa 18 Aug '17

18 Aug '17

The helper returns true if a string contains any of the given chars. virStringHasControlChars can be reimplemented using that helper. --- Pavel was loudly thinking about such function. It turns out that I have it ready on one of my branches. src/util/virstring.c | 23 +++++++++++++++++++---- src/util/virstring.h | 2 ++ 2 files changed, 21 insertions(+), 4 deletions(-) diff --git a/src/util/virstring.c b/src/util/virstring.c index 58abf9dd6..0288d1e67 100644 --- a/src/util/virstring.c +++ b/src/util/virstring.c @@ -1195,6 +1195,24 @@ virStringStripIPv6Brackets(char *str) } +/** + * virStringHasChars: + * @str: string to look for chars in + * @chars: chars to find in string @str + * + * Returns true if @str contains any of the chars in @chars. + */ +bool +virStringHasChars(const char *str, + const char *chars) +{ + if (!str) + return false; + + return str[strcspn(str, chars)] != '\0'; +} + + static const char control_chars[] = "\x01\x02\x03\x04\x05\x06\x07" "\x08" /* \t \n */ "\x0B\x0C" /* \r */ "\x0E\x0F" @@ -1204,10 +1222,7 @@ static const char control_chars[] = bool virStringHasControlChars(const char *str) { - if (!str) - return false; - - return str[strcspn(str, control_chars)] != '\0'; + return virStringHasChars(str, control_chars); } diff --git a/src/util/virstring.h b/src/util/virstring.h index ff5f0148d..1290fcce1 100644 --- a/src/util/virstring.h +++ b/src/util/virstring.h @@ -286,6 +286,8 @@ char *virStringReplace(const char *haystack, ATTRIBUTE_NONNULL(1) ATTRIBUTE_NONNULL(2) ATTRIBUTE_NONNULL(3); void virStringStripIPv6Brackets(char *str); +bool virStringHasChars(const char *str, + const char *chars); bool virStringHasControlChars(const char *str); void virStringStripControlChars(char *str); -- 2.14.0

2 1

[libvirt] [PATCH 00/12] qemu: don't lose VMs if emulator is not installed
by Peter Krempa 18 Aug '17

18 Aug '17

The post-parse callback grew into an abomination which requires qemuCaps to succeed. That won't work out well if for some reasons qemu is uninstalled. Restarting of libvirtd would result in all VMs being lost untill qemu is reinstalled. Fix this by allowing qemuCaps to be missing and re-running the postparse callbacks when attempting a VM start. Peter Krempa (12): conf: domainlist: Explicitly report failure to load domain config conf: Add 'basic' post parse callback qemu: Move assignment of default emulator to the basic post parse callback conf: Add callbacks that allocate per-def private data qemu: domain: Don't re-allocate qemuCaps in post parse callbacks conf: Return any non-zero value from virDomainDeviceInfoIterateInternal callback conf: add infrastructure for tolerating certain post parse callback failures qemu: capabilities: Tolerate missing @qemuCaps in virQEMUCapsGetCanonicalMachine qemu: capabilities: Tolerate missing @qemuCaps in virQEMUCapsSupportsGICVersion qemu: domain: Don't return default NIC model if @qemuCaps are missing qemu: domain: Don't set default USB model if qemuCaps is missing qemu: Implement postParse callback skipping on config reload src/conf/domain_conf.c | 202 +++++++++++++++++++++++++++++-------------- src/conf/domain_conf.h | 35 +++++++- src/conf/virdomainobjlist.c | 8 +- src/qemu/qemu_capabilities.c | 20 +++-- src/qemu/qemu_domain.c | 111 ++++++++++++++++-------- src/qemu/qemu_process.c | 9 ++ 6 files changed, 277 insertions(+), 108 deletions(-) -- 2.14.0

2 16

Re: [libvirt] [PATCH v2] Add support for virtio-net.tx_queue_size
by Jan Scheurich 18 Aug '17

18 Aug '17

On Thu, Jul 27 at 14:05:09 +0200 Peter Krempa wrote: > On Thu, Jul 20, 2017 at 21:36:28 -0400, Laine Stump wrote: > > On 07/18/2017 07:12 AM, Michal Privoznik wrote: [...] > > >>> So only users that know how virtio works under the hood are > > >>> expected to also know what rx/tx queue size is and how to set it. > > >>> But frankly, I > > >> This statement is ridiculous by itself. > > > Why? There are more experienced users (for whom libvirt's just a > > > stable > > > API) and less experienced ones (for whom libvirt's and tools on the > > > top of it are great for their automatic chose of parameters, e.g. gnome > boxes). > > > > Beyond that, that same statement (or something nearly identical) is > > repeated in multiple places throughout the XML documentation. There > > are at least two classes of these attributes that I can think of: > > > > 1) attributes that are automatically set to a sane value by libvirt > > when not specified (and they usually *aren't* specified), and saved in > > the config XML in order to assure they are set the same every time the > > domain is started (to preserve guest ABI). These are intended to > > record whatever was the default setting for the attribute at the time > > the domain was created. For example, a pcie-root-port controller might > > have <model name='ioh3420'/> set, if that was the only type of > > pcie-root-port available at the time a domain was created; this comes > > in handy now that there is a generic pcie-root-port (which also has > > <model > > name='pcie-root-port'/>) - existing domains don't get their ABI > > screwed up when they're migrated to a newer host. > > > > 2) knobs that have been added in over the years at the request/demand > > from below (qemu) and above (ovirt / openstack), many of them obscure, > > difficult to explain with any amount of useful detail, and almost > > never worthy of being manually set (and if you "don't know what you're > > doing", you're just as likely to make things worse as to make them > better). > > > > tx_queue_size is one of the latter. > > > > For either of these types of attributes, they need to be documented so > > that someone encountering them (or actively searching them out) will > > at least have a basic idea what the attribute is named / used for, but > > we also need to warn the casual user to not mess with them. I don't > > see anything ridiculous about that. > We, Ericsson, have been driving the possibility to increase the Rx and Tx queue sizes for virtio interfaces in Qemu because our throughput-intensive and packet-drop sensitive virtual network functions (VNFs) were suffering frequent early packet drops. In the age of DPDK and with packet rates in the Mpps range the hard-coded default queue size of 256 in Qemu was no longer adequate to compensate short interrupts on the receiving side or microbursts on the sending size. Many measurements across the industry had proven that increasing the queue size to 1024 very effectively reduced the packet drop probability and consequently improved throughput significantly. The Qemu community could not change the virtio-net default queue size to 1024 to avoid compatibility issues and limitations in legacy virtio drivers and virtio backends. But many VNFs and the DPDK vhostuser backend in vSwitches like OVS and fd.io/VPP can handle 1K queue size and benefit from it. So they made the queue sizes configurable in Qemu to give NFVI operators the chance to tune the performance of their infrastructure. This change in Libvirt is about making this tuning option available for Qemu users like OpenStack. There is no other way than Libvirt to make this accessible as Qemu does not have the facility of global default parameters. > > >>> think users setting this are always gonna go with the highest > > >>> value avaliable (1024). Such detailed description is a copy of > > >>> rx_virtio_queue size description which is result of review. Reasons not to set to 1024 could include compatibility constraints (see above) or a trade-off between packet drop probability and maximum latency/jitter. > > >>> > > >>>> Is it really needed? How should it be configured? Can't we or > > >>>> qemu pick a sane value? > > >>>> > > >>> No. Some users need bigger virtio rings otherwise they see a > > >>> packet drop. So this is a fine tuning that heavily depends on the use > case. > > >>> Thus libvirt should not try to come up with some value. > > >> That's why I think it's wrong. What's the drawback of setting it > > >> to maximum? Which workloads will hit it? Why is the default not > sufficient? > > >> > > >> And most notably, how do the users figure out that they need it? In our case the "user" is the NFVI deployment tool chain with tuning input from NFV solution architects, guided by Installation/Dimensioning guides. > > > > > > I think it's a bit too much burden on libvirt to expect that much > > detail to be included in formatdomain.html. Heck, I don't know if > > anyone even > > *knows* that much detail right now. > > That proves my point kind of. If there isn't knowledge how to set up this, > then why even add the option. Is there really a point? I think we could improve the documentation a bit with reference to vhostuser interfaces as main target interface type and mentioning potential compatibility constraints for applications. We would actually like to see an additional global config knob in Libvirt to (optionally) set default Rx and Tx queue sizes for vhostuser interfaces, so that Libvirt consumers like OpenStack would not have to be modified to use 1K queues. BR, Jan

1 0