- Devel - libvirt List Archives

[libvirt] [RFC] Faster libvirtd restart with nwfilter rules
by Nikolay Shirokovskiy 12 Oct '18

12 Oct '18

Hi, all. On fat hosts which are capable to run hundreds of VMs restarting libvirtd makes it's services unavailable for a long time if VMs use network filters. In my tests each of 100 VMs has no-promisc [1] and no-mac-spoofing filters and executing virsh list right after daemon restart takes appoximately 140s if no firewalld is running (that is ebtables/iptables/ip6tables commands are used to configure kernel tables). The problem is daemon does not even start to read from client connections because state drivers are not initialized. Initialization is blocked in state drivers autostart which grabs VMs locks. And VMs locks are hold by VMs reconnection code. Each VM reloads network tables on reconnection and this reloading is serialized on updateMutex in gentech nwfilter driver. Workarounding autostart won't help much because even if state drivers will initialize listing VM won't be possible because listing VMs takes each VM lock one by one too. However managing VM that passed reconnection phase will be possible which takes same 140s in worst case. Note that this issue is only applicable if we use filters configuration that don't need ip learning. In the latter case situation is different because reconnection code spawns new thread that apply network rules only after ip is learned from traffic and this thread does not grab VM lock. As result VMs are managable but reloading filters in background takes appoximately those same 140s. I guess managing network filters during this period can have issues too. Anyway this situation does not look good so fixing the described issue by spawning threads even without ip learning does not look nice to me. What speed up is possible on conservative approach? First we can remove for test purpuses firewall ruleLock, gentech dirver updateMutex and filter object mutex which do not serve function in restart scenario. This gives 36s restart time. The speed up is archived because heavy fork/preexec steps are now run concurrently. Next we can try to reduce fork/preexec time. To estimate its contibution alone let's bring back the above locks. It turns out the most time takes fork itself and closing 8k (on my system) file descriptors in preexec. Using vfork gives 2x boost and so does dropping mass close. (I check this mass close contribution because I not quite understand the purpose of this step - libvirt typically set close-on-exec flag on it's descriptors). So this two optimizations alone can result in restart time of 30s. Unfortunately combining the above two approaches does not give boost multiple of them along. The reason is due to concurrency and high number of VMs (100) preexec boost does not have significant role and using vfork dininishes concurrency as it freezes all parent threads before execve. So dropping locks and closes gives 33s restart time and adding vfork to this gives 25s restart time. Another approach is to use --atomic-file option for ebtables (iptables/ip6tables unfortunately does not have one). The idea is to save table to file/edit file/commit table to kernel. I hoped this could give performance boost because we don't need to load/store kernel network table for a single rule update. In order to isolate approaches I also dropped all ip/ip6 updates which can not be done this way. In this approach we can not drop ruleLock in firewall because no other VM threads should change tables between save/commit. This approach gives restart time 25s. But this approach is broken anyway as we can not be sure another application doesn't change newtork table between save/commit in which case these changes will be lost. After all I think we need to move in a different direction. We can add API to all binaries and firewalld to execute many commands in one run. We can pass commands as arguments or wrote them into file which is then given to binary. Then libvirt itself can update for example bridge network table in couple of commands. The exact number depends on new API. For example if we add option to delete chains recursively and an option not to fail on NOENT error we can change table in one command (no listing current rules is required). [1] no-promisc filter <filter name='no-promisc' chain='root' priority='-750'> <uuid>6d055022-1192-4a3d-ae1f-576baa5564b6</uuid> <rule action='return' direction='in' priority='500'> <mac dstmacaddr='ff:ff:ff:ff:ff:ff'/> </rule> <rule action='return' direction='in' priority='500'> <mac dstmacaddr='$MAC'/> </rule> <rule action='return' direction='in' priority='500'> <mac dstmacaddr='33:33:00:00:00:00' dstmacmask='ff:ff:00:00:00:00'/> </rule> <rule action='drop' direction='in' priority='500'> <mac/> </rule> <rule action='return' direction='in' priority='500'> <mac dstmacaddr='01:00:5e:00:00:00' dstmacmask='ff:ff:ff:80:00:00'/> </rule> </filter>

3 6

[libvirt] [PATCH] qemu: Remove unused qemuProcessAutostartAll
by John Ferlan 12 Oct '18

12 Oct '18

The function was never defined in source, just the protoype. Signed-off-by: John Ferlan <jferlan(a)redhat.com> --- Pushed as trivial. src/qemu/qemu_process.h | 1 - 1 file changed, 1 deletion(-) diff --git a/src/qemu/qemu_process.h b/src/qemu/qemu_process.h index c2f7c2b5d2..2037467c94 100644 --- a/src/qemu/qemu_process.h +++ b/src/qemu/qemu_process.h @@ -47,7 +47,6 @@ int qemuProcessDestroyMemoryBackingPath(virQEMUDriverPtr driver, virDomainObjPtr vm, virDomainMemoryDefPtr mem); -void qemuProcessAutostartAll(virQEMUDriverPtr driver); void qemuProcessReconnectAll(virQEMUDriverPtr driver); typedef struct _qemuProcessIncomingDef qemuProcessIncomingDef; -- 2.17.1

1 0

[libvirt] [glib PATCH 0/3] Add ich9 sound device + _domain_os_get_machine()
by Fabiano Fidêncio 11 Oct '18

11 Oct '18

The first patch is needed to properly support Q35 machine type on GNOME Boxes (or any other app using libvirt-glib) as the ICH9 audio device should be used together with Q35 machine type, The second patch is mostly a convenience for the apps and the third one just adds a test case for the second. Fabiano Fidêncio (3): gconfig-domain-sound: Add ich9 sound device gconfig: add _domain_os_get_machine() tests: Add testes for _domain_os_{set,get}_machine() libvirt-gconfig/libvirt-gconfig-domain-os.c | 7 +++++++ libvirt-gconfig/libvirt-gconfig-domain-os.h | 1 + libvirt-gconfig/libvirt-gconfig-domain-sound.h | 3 ++- libvirt-gconfig/libvirt-gconfig.sym | 1 + tests/test-gconfig.c | 2 ++ tests/xml/gconfig-domain-os.xml | 2 +- 6 files changed, 14 insertions(+), 2 deletions(-) -- 2.19.1

2 4

[libvirt] [PATCH v2 1/2] add nodeset='all' for interleave mode
by Peng Hao 11 Oct '18

11 Oct '18

sometimes we hoped that the memory of vm can be evenly distributed in all nodes according to interleave mode. But different hosts has different node number. So we add nodeset='all' for interleave mode. Signed-off-by: Peng Hao <peng.hao2(a)zte.com.cn> --- src/conf/numa_conf.c | 77 ++++++++++++++++++++++++++++++++++++------- 1 files changed, 64 insertions(+), 13 deletions(-) diff --git a/src/conf/numa_conf.c b/src/conf/numa_conf.c index 97a3ca4..a336a62 100644 --- a/src/conf/numa_conf.c +++ b/src/conf/numa_conf.c @@ -29,6 +29,9 @@ #include "virnuma.h" #include "virstring.h" +#if WITH_NUMACTL +#include <numa.h> +#endif /* * Distance definitions defined Conform ACPI 2.0 SLIT. * See include/linux/topology.h @@ -66,6 +69,7 @@ typedef virDomainNumaNode *virDomainNumaNodePtr; struct _virDomainNuma { struct { bool specified; + bool allnode; virBitmapPtr nodeset; virDomainNumatuneMemMode mode; virDomainNumatunePlacement placement; @@ -259,13 +263,21 @@ virDomainNumatuneParseXML(virDomainNumaPtr numa, tmp = virXMLPropString(node, "nodeset"); if (tmp) { - if (virBitmapParse(tmp, &nodeset, VIR_DOMAIN_CPUMASK_LEN) < 0) - goto cleanup; - - if (virBitmapIsAllClear(nodeset)) { + if (STREQ(tmp, "all") && !virNumaIsAvailable()) { virReportError(VIR_ERR_CONFIG_UNSUPPORTED, - _("Invalid value of 'nodeset': %s"), tmp); + _("Invalid nodeset=%s when numactl is not supported"), tmp); goto cleanup; + } else if (STREQ(tmp, "all") && virNumaIsAvailable()) { + numa->memory.allnode = true; + } else { + if (virBitmapParse(tmp, &nodeset, VIR_DOMAIN_CPUMASK_LEN) < 0) + goto cleanup; + + if (virBitmapIsAllClear(nodeset)) { + virReportError(VIR_ERR_CONFIG_UNSUPPORTED, + _("Invalid value of 'nodeset': %s"), tmp); + goto cleanup; + } } VIR_FREE(tmp); @@ -319,10 +331,14 @@ virDomainNumatuneFormatXML(virBufferPtr buf, virBufferAsprintf(buf, "<memory mode='%s' ", tmp); if (numatune->memory.placement == VIR_DOMAIN_NUMATUNE_PLACEMENT_STATIC) { - if (!(nodeset = virBitmapFormat(numatune->memory.nodeset))) - return -1; - virBufferAsprintf(buf, "nodeset='%s'/>\n", nodeset); - VIR_FREE(nodeset); + if (numatune->memory.allnode == true) { + virBufferAddLit(buf, "nodeset='all'/>\n"); + } else { + if (!(nodeset = virBitmapFormat(numatune->memory.nodeset))) + return -1; + virBufferAsprintf(buf, "nodeset='%s'/>\n", nodeset); + VIR_FREE(nodeset); + } } else if (numatune->memory.placement) { tmp = virDomainNumatunePlacementTypeToString(numatune->memory.placement); virBufferAsprintf(buf, "placement='%s'/>\n", tmp); @@ -489,6 +505,37 @@ virDomainNumatuneMaybeFormatNodeset(virDomainNumaPtr numatune, return 0; } +#if WITH_NUMACTL +static int +makeAllnodeBitmap(virDomainNumaPtr numa) +{ + size_t i = 0, maxnode = 0; + virBitmapPtr bitmap = NULL; + + if ((bitmap = virBitmapNew(VIR_DOMAIN_CPUMASK_LEN)) == NULL) + return -1; + virBitmapClearAll(bitmap); + maxnode = numa_max_node(); + for (i = 0; i <= maxnode; i++) { + if (virBitmapSetBit(bitmap, i) < 0) { + virBitmapFree(bitmap); + return -1; + } + } + + virBitmapFree(numa->memory.nodeset); + numa->memory.nodeset = bitmap; + + return 0; +} +#else +static int +makeAllnodeBitmap(virDomainNumaPtr numa) +{ + return -1; +} +#endif + int virDomainNumatuneSet(virDomainNumaPtr numa, bool placement_static, @@ -538,20 +585,28 @@ virDomainNumatuneSet(virDomainNumaPtr numa, } if (placement == VIR_DOMAIN_NUMATUNE_PLACEMENT_DEFAULT) { - if (numa->memory.nodeset || placement_static) + if (numa->memory.nodeset || placement_static || numa->memory.allnode) placement = VIR_DOMAIN_NUMATUNE_PLACEMENT_STATIC; else placement = VIR_DOMAIN_NUMATUNE_PLACEMENT_AUTO; } if (placement == VIR_DOMAIN_NUMATUNE_PLACEMENT_STATIC && - !numa->memory.nodeset) { + !numa->memory.nodeset && !numa->memory.allnode) { virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s", _("nodeset for NUMA memory tuning must be set " "if 'placement' is 'static'")); goto cleanup; } + if (placement == VIR_DOMAIN_NUMATUNE_PLACEMENT_STATIC && + mode == VIR_DOMAIN_NUMATUNE_MEM_INTERLEAVE && + numa->memory.allnode && virNumaIsAvailable()) { + + if (makeAllnodeBitmap(numa) < 0) + goto cleanup; + } + /* setting nodeset when placement auto is invalid */ if (placement == VIR_DOMAIN_NUMATUNE_PLACEMENT_AUTO && numa->memory.nodeset) { -- 1.8.3.1

3 5

Re: [libvirt] [Qemu-devel] [PATCH v2 0/3] HMP/snapshot changes - do not use ID anymore
by Markus Armbruster 11 Oct '18

11 Oct '18

Cc: libvir-list for review of the compatibility argument below. Daniel Henrique Barboza <danielhb413(a)gmail.com> writes: > Hey David, > > On 9/21/18 9:29 AM, Dr. David Alan Gilbert wrote: >> * Daniel Henrique Barboza (danielhb413(a)gmail.com) wrote: >>> changes in v2: >>> - removed the "RFC" marker; >>> - added a new patch (patch 2) that removes >>> bdrv_snapshot_delete_by_id_or_name from the code; >>> - made changes in patch 1 as suggested by Murilo; >>> - previous patch set link: >>> https://lists.gnu.org/archive/html/qemu-devel/2018-08/msg04658.html >>> >>> >>> It is not uncommon to see bugs being opened by testers that attempt to >>> create VM snapshots using HMP. It turns out that "0" and "1" are quite >>> common snapshot names and they trigger a lot of bugs. I gave an example >>> in the commit message of patch 1, but to sum up here: QEMU treats the >>> input of savevm/loadvm/delvm sometimes as 'ID', sometimes as 'name'. It >>> is documented as such, but this can lead to strange situations. >>> >>> Given that it is strange for an API to consider a parameter to be 2 fields >>> at the same time, and inadvently treating them as one or the other, and >>> that removing the ID field is too drastic, my idea here is to keep the >>> ID field for internal control, but do not let the user set it. >>> >>> I guess there's room for discussion about considering this change an API >>> change or not. It doesn't affect users of HMP and it doesn't affect Libvirt, >>> but simplifying the meaning of the parameters of savevm/loadvm/delvm. >> Can you clarify a couple of things: >> a) What is it about libvirt's use that means it's OK ? Does it never >> use numeric tags? > > I am glad you asked because my understanding in how Libvirt was dealing > with snapshots was wrong, and I just looked into it further to answer your > question. Luckily, this series fixes the situation for Libvirt as well. > > I was misled by the fact that Libvirt does not show the same symptoms > we see in > QEMU of this problem, but the bug is there. Here's a quick test with > Libvirt with > "0" and "1" as snapshot names, considering a VM named with no snapshots, > using QEMU 2.12 and Libvirt 4.0.0: > > - create the "0" snapshot: > > $ sudo virsh snapshot-create-as --name 0 dhb > Domain snapshot 0 created > > $ sudo virsh snapshot-list dhb > Name Creation Time State > ------------------------------------------------------------ > 0 2018-09-24 15:47:56 -0400 running > > $ sudo virsh qemu-monitor-command dhb --hmp info snapshots > List of snapshots present on all disks: > ID TAG VM SIZE DATE VM CLOCK > -- 0 405M 2018-09-24 15:47:56 00:04:20.867 > > > - created the "1" snapshot. Here, Libvirt shows both snapshots with > snapshot-list, > but the snapshot was erased inside QEMU: > > $ sudo virsh snapshot-create-as --name 1 dhb > Domain snapshot 1 created > $ > $ sudo virsh snapshot-list dhb > Name Creation Time State > ------------------------------------------------------------ > 0 2018-09-24 15:47:56 -0400 running > 1 2018-09-24 15:50:09 -0400 running > > $ sudo virsh qemu-monitor-command dhb --hmp info snapshots > List of snapshots present on all disks: > ID TAG VM SIZE DATE VM CLOCK > -- 1 404M 2018-09-24 15:50:10 00:05:36.226 > > > This is where I stopped checking out Libvirt at first, concluding > wrongly that it > was immune to the bug. > > Libvirt does not throw an error when trying to apply snapshot 0. In > fact, it acts > like everything went fine: > > $ sudo virsh snapshot-revert dhb 0 > > $ echo $? > 0 Is that a libvirt bug? > Reverting back to snapshot "1" works as intended, restoring the VM > state when it > was created. > > > (perhaps this is something we should let Libvirt be aware of ...) > > > > This series fixes this behavior because the snapshot 0 isn't erased > from QEMU, allowing > Libvirt to successfully restore it. > > >> b) After this series are you always guaranteed to be able to fix >> any existing oddly named snapshots? > > The oddly named snapshots that already exists can be affected by the > semantic > change in loadvm and delvm, in a way that the user can't load/del > using the snap > ID, just the tag. Aside from that, I don't see any side effects with > existing > snapshots and this patch series. Do all snapshots have a tag that is unique within their image? Even snapshots created by old versions of QEMU?

3 3

[libvirt] [PATCH] virfiletest: Load mock on Linux only
by Michal Privoznik 11 Oct '18

11 Oct '18

The mock is built on Linux only. Therefore we should load it only on Linux too. This fixes the FreeBSD build. Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com> --- tests/virfiletest.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/tests/virfiletest.c b/tests/virfiletest.c index d5102b1cc4..1f2be74c8d 100644 --- a/tests/virfiletest.c +++ b/tests/virfiletest.c @@ -460,4 +460,8 @@ mymain(void) return ret != 0 ? EXIT_FAILURE : EXIT_SUCCESS; } +#ifdef __linux__ VIR_TEST_MAIN_PRELOAD(mymain, abs_builddir "/.libs/virfilemock.so") +#else +VIR_TEST_MAIN(mymain) +#endif -- 2.18.0

2 1

[libvirt] [PATCH 00/30] syntax: Remove spaces after casts
by Martin Kletzander 11 Oct '18

11 Oct '18

3 54

[libvirt] [PATCH v3] virfile: fix cast-align error
by Bjoern Walk 11 Oct '18

11 Oct '18

From: Marc Hartmayer <mhartmay(a)linux.ibm.com> On s390x the struct member f_type of statsfs is hard coded to 'unsigned int'. Change virFileIsSharedFixFUSE() to take a 'long long int' and use a temporary to avoid pointer-casting. This fixes the following error: ../../src/util/virfile.c:3578:38: error: cast increases required alignment of target type [-Werror=cast-align] virFileIsSharedFixFUSE(path, (long *) &sb.f_type); Signed-off-by: Marc Hartmayer <mhartmay(a)linux.ibm.com> Signed-off-by: Bjoern Walk <bwalk(a)linux.ibm.com> --- src/util/virfile.c | 23 +++++++++++++---------- 1 file changed, 13 insertions(+), 10 deletions(-) diff --git a/src/util/virfile.c b/src/util/virfile.c index 2a7e8710..067ca569 100644 --- a/src/util/virfile.c +++ b/src/util/virfile.c @@ -3466,7 +3466,7 @@ int virFilePrintf(FILE *fp, const char *msg, ...) static int virFileIsSharedFixFUSE(const char *path, - long *f_type) + long long *f_type) { char *dirpath = NULL; const char **mounts = NULL; @@ -3537,6 +3537,7 @@ virFileIsSharedFSType(const char *path, char *p; struct statfs sb; int statfs_ret; + long long f_type = 0; if (VIR_STRDUP(dirpath, path) < 0) return -1; @@ -3573,32 +3574,34 @@ virFileIsSharedFSType(const char *path, return -1; } - if (sb.f_type == FUSE_SUPER_MAGIC) { + f_type = sb.f_type; + + if (f_type == FUSE_SUPER_MAGIC) { VIR_DEBUG("Found FUSE mount for path=%s. Trying to fix it", path); - virFileIsSharedFixFUSE(path, (long *) &sb.f_type); + virFileIsSharedFixFUSE(path, &f_type); } VIR_DEBUG("Check if path %s with FS magic %lld is shared", - path, (long long int)sb.f_type); + path, f_type); if ((fstypes & VIR_FILE_SHFS_NFS) && - (sb.f_type == NFS_SUPER_MAGIC)) + (f_type == NFS_SUPER_MAGIC)) return 1; if ((fstypes & VIR_FILE_SHFS_GFS2) && - (sb.f_type == GFS2_MAGIC)) + (f_type == GFS2_MAGIC)) return 1; if ((fstypes & VIR_FILE_SHFS_OCFS) && - (sb.f_type == OCFS2_SUPER_MAGIC)) + (f_type == OCFS2_SUPER_MAGIC)) return 1; if ((fstypes & VIR_FILE_SHFS_AFS) && - (sb.f_type == AFS_FS_MAGIC)) + (f_type == AFS_FS_MAGIC)) return 1; if ((fstypes & VIR_FILE_SHFS_SMB) && - (sb.f_type == SMB_SUPER_MAGIC)) + (f_type == SMB_SUPER_MAGIC)) return 1; if ((fstypes & VIR_FILE_SHFS_CIFS) && - (sb.f_type == CIFS_SUPER_MAGIC)) + (f_type == CIFS_SUPER_MAGIC)) return 1; return 0; -- 2.17.0

2 2

[libvirt] [PATCH 0/4] Couple of virFileIsShared* fixes
by Michal Privoznik 10 Oct '18

10 Oct '18

This patch [1] got me looking into the code. I've found couple of issues which I'm fixing and also I'm introducing couple of new test cases. 1: https://www.redhat.com/archives/libvir-list/2018-October/msg00513.html Michal Prívozník (4): virfiletest: Fix test name prefix for virFileInData test virfiletst: Test virFileIsSharedFS virFileIsSharedFSType: Detect direct mount points virfile: Rework virFileIsSharedFixFUSE src/util/virfile.c | 66 ++++++--------- tests/Makefile.am | 7 +- tests/virfiledata/mounts3.txt | 35 ++++++++ tests/virfilemock.c | 154 ++++++++++++++++++++++++++++++++++ tests/virfiletest.c | 64 +++++++++++++- 5 files changed, 284 insertions(+), 42 deletions(-) create mode 100644 tests/virfiledata/mounts3.txt create mode 100644 tests/virfilemock.c -- 2.18.0

2 11

[libvirt] Recall: [PATCHv5 04/19] util: Add interface for adding PID to monitor
by Wang, Huaqiang 10 Oct '18

10 Oct '18

Wang, Huaqiang would like to recall the message, "[libvirt] [PATCHv5 04/19] util: Add interface for adding PID to monitor".

1 0