February 2019 - Devel - libvirt List Archives

[libvirt] Plans for next release
by Daniel Veillard 22 Feb '19

22 Feb '19

So it's getting time to think about 5.1.0, I'm thinking about entering freeze on Monday next week, push rc2 Wed and if everything goes well we can have the release out on Friday 1st, I hope that works for everyone, Daniel P.S. I will be on west coast so freeze and pushes likely to be later in the day than usual. -- Daniel Veillard | Red Hat Developers Tools http://developer.redhat.com/ veillard(a)redhat.com | libxml Gnome XML XSLT toolkit http://xmlsoft.org/ http://veillard.com/ | virtualization library http://libvirt.org/

1 0

[libvirt] [PATCH 0/1] update copyright notice
by David Kiarie 22 Feb '19

22 Feb '19

this updates the patch to use a full legal name on the copyright notice. David Kiarie (1): src/xenconfig: update copyright notice src/xenconfig/xen_xl.c | 1 + src/xenconfig/xen_xl.h | 1 + tests/xlconfigtest.c | 1 + 3 files changed, 3 insertions(+) -- 2.20.1

2 2

[libvirt] [PATCH 0/1] update copyright notice
by David Kiarie 22 Feb '19

22 Feb '19

these patche follows up a very co-operative review on a preview patch or lack of thereof. i have updated this to reflect the exact name i use in person. David Kiarie (1): src/xenconfig: update copyright notice src/xenconfig/xen_xl.c | 1 + src/xenconfig/xen_xl.h | 1 + tests/xlconfigtest.c | 1 + 3 files changed, 3 insertions(+) -- 2.20.1

2 3

[libvirt] [PATCH] qemu: fix memory leak in qemuBuildDiskDeviceStr
by Ján Tomko 22 Feb '19

22 Feb '19

Commit a1dce962 added the allocated scsiVPDDeviceId without freeing it. Signed-off-by: Ján Tomko <jtomko(a)redhat.com> --- src/qemu/qemu_command.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/qemu/qemu_command.c b/src/qemu/qemu_command.c index f511193fca..74f34af292 100644 --- a/src/qemu/qemu_command.c +++ b/src/qemu/qemu_command.c @@ -1890,7 +1890,7 @@ qemuBuildDiskDeviceStr(const virDomainDef *def, const char *bus = virDomainDiskQEMUBusTypeToString(disk->bus); const char *contAlias; char *backendAlias = NULL; - char *scsiVPDDeviceId = NULL; + VIR_AUTOFREE(char *) scsiVPDDeviceId = NULL; int controllerModel; if (qemuCheckDiskConfig(disk, qemuCaps) < 0) -- 2.19.2

2 1

[libvirt] [PATCH 0/7] Introduce automatic clearing for virBuffer
by Peter Krempa 22 Feb '19

22 Feb '19

First 4 patches clean up existing mess. Note that this series both fixes and then deletes virBufferEscapeN. The split is deliberate, so that if somebody needs virBufferEscapeN in the future they can revert the removal and get a working version. Peter Krempa (7): util: buffer: Remove misleading AUTOPTR func for 'virBuffer' util: buf: Fix memory leak in virBufferEscapeN tests: buf: Fix debug messages in 'testBufEscapeRegex' util: buf: Remove virBufferEscapeN util: alloc: Introduce 'VIR_AUTOCLEAN' macros for clearing stack'd structs util: buffer: Introduce VIR_AUTOCLEAN function for virBuffer qemu: domain: Use VIR_AUTOCLEAN for virBuffer src/libvirt_private.syms | 1 - src/qemu/qemu_domain.c | 27 +++------- src/util/viralloc.h | 32 +++++++++++- src/util/virbuffer.c | 107 --------------------------------------- src/util/virbuffer.h | 7 ++- tests/virbuftest.c | 57 +++++---------------- 6 files changed, 56 insertions(+), 175 deletions(-) -- 2.20.1

2 10

[libvirt] [PATCHv2 0/2] network: add netmask to dhcp range of dnsmasq conf file for IPv4
by Laine Stump 22 Feb '19

22 Feb '19

I sent V1 of this patch way back in October: https://www.redhat.com/archives/libvir-list/2018-October/msg00889.html but then self-NACKed it because I had written the patch to add the network prefix to dhcp-range, but IPv4 requires the netmask rather than range. Unfortunately, when I modified the patch accordingly the getnameinfo() function began failing, I couldn't immediately see why, and I had to pack everything up and go to KVM Forum. After that, I promptly forgot to look back at it. Today I finally revisited the problem (after the person who originally found it sent a reminder), and discovered the failure of getnameinfo() was due to a bug in a function that I added to the virSocketAddr library in 2010! Laine Stump (2): util: set missing data length in virSocketAddrPrefixToNetmask() network: add netmask to dhcp range of dnsmasq conf file for IPv4 src/network/bridge_driver.c | 27 +++++++++++++++---- src/util/virsocketaddr.c | 2 ++ .../dhcp6-nat-network.conf | 2 +- .../networkxml2confdata/isolated-network.conf | 2 +- .../nat-network-dns-srv-record-minimal.conf | 2 +- .../nat-network-dns-srv-record.conf | 2 +- .../nat-network-dns-txt-record.conf | 2 +- .../networkxml2confdata/nat-network-mtu.conf | 2 +- .../nat-network-name-with-quotes.conf | 2 +- tests/networkxml2confdata/nat-network.conf | 2 +- .../networkxml2confdata/netboot-network.conf | 2 +- .../netboot-proxy-network.conf | 2 +- .../networkxml2confdata/ptr-domains-auto.conf | 2 +- 13 files changed, 35 insertions(+), 16 deletions(-) -- 2.20.1

2 6

[libvirt] [PATCH 0/7] bulk snapshot list/redefine (incremental backup saga)
by Eric Blake 22 Feb '19

22 Feb '19

While looking at my work on incremental backups, Nir raised a good point: if we want to recreate a set of known checkpoints on one machine that will be taking over a domain from another machine, my initial proposal required making multiple API calls to list the XML for each checkpoint on the source, and then more API calls to redefine each checkpoint on the destination; it also had the drawback that the list has to be presented in topological order (the API won't let you define a child checkpoint if the parent is not defined first). He asked if I could instead add bulk APIs, for getting the XML for all checkpoints at once, and then for redefining all checkpoints at once. Since my checkpoint code borrows heavily from concepts in the snapshot code, I chose to tackle the problem by starting with this series, which does the same thing for snapshots as what I plan to do for checkpoints. That is, since this patch series adds virDomainGetXMLDesc(, VIR_DOMAIN_XML_SNAPSHOTS) and virDomainSnapshotCreateXML(, VIR_DOMAIN_SNAPSHOT_CREATE_REDEFINE_LIST), the checkpoint series will add virDOmainGetXMLDesc(, VIR_DOMAIN_XML_CHECKPOINTS) and virDomainCheckpointCreateXML(, VIR_DOMAIN_CHECKPOINT_CREATE_REDEFINE_LIST) with very similar code. Note that I'm posting the series now to start review; the bulk query is fully implemented (patch 6), while the bulk redefine still needs more work (patch 7 defines the API, but the actual implementation is trickier, so I'll be posting additional patches later once I get them working). Patch 1 is a revision of work posted previously in a different series where all other patches of that series are now on mainline. Also available at: https://repo.or.cz/libvirt/ericb.git/shortlog/refs/tags/snapshot-bulk-v1 Eric Blake (7): domain: Document VIR_DOMAIN_XML_MIGRATABLE snapshot: Give virDomainSnapshotDefFormat its own flags snapshot: Refactor virDomainSnapshotDefFormat snapshot: Add virDomainSnapshotObjListFormat domain: Expand virDomainDefFormatInternal with snapshots domain: Add VIR_DOMAIN_XML_SNAPSHOTS flag snapshot: Add VIR_DOMAIN_SNAPSHOT_CREATE_REDEFINE_LIST flag include/libvirt/libvirt-domain-snapshot.h | 3 + include/libvirt/libvirt-domain.h | 1 + src/conf/domain_conf.h | 8 ++ src/conf/snapshot_conf.h | 18 ++- src/conf/domain_conf.c | 71 ++++++++-- src/conf/snapshot_conf.c | 162 ++++++++++++++++------ src/esx/esx_driver.c | 1 - src/libvirt-domain-snapshot.c | 21 ++- src/libvirt-domain.c | 14 ++ src/libvirt_private.syms | 2 + src/network/bridge_driver.c | 3 +- src/qemu/qemu_domain.c | 24 +++- src/qemu/qemu_driver.c | 7 +- src/test/test_driver.c | 10 +- src/vbox/vbox_common.c | 5 +- src/vz/vz_driver.c | 3 +- tests/domainsnapshotxml2xmltest.c | 16 ++- tools/virsh-domain.c | 7 + tools/virsh-snapshot.c | 13 ++ 19 files changed, 293 insertions(+), 96 deletions(-) -- 2.20.1

1 10

[libvirt] [PATCH] snapshot: Saner use of uuid
by Eric Blake 22 Feb '19

22 Feb '19

Most of the code base is fairly consistent about using the name 'uuidstr' when dealing with a formatted human-readable form, and 'uuid' when dealing with the smaller raw bytes form. Fix snapshot_conf to comply, as well as reducing the scope of a human string to only the error message that needs it. Signed-off-by: Eric Blake <eblake(a)redhat.com> --- Pushing this one under the trivial rule. src/conf/snapshot_conf.h | 2 +- src/conf/snapshot_conf.c | 12 ++++++------ 2 files changed, 7 insertions(+), 7 deletions(-) diff --git a/src/conf/snapshot_conf.h b/src/conf/snapshot_conf.h index cc25164261..7a175dfc96 100644 --- a/src/conf/snapshot_conf.h +++ b/src/conf/snapshot_conf.h @@ -111,7 +111,7 @@ virDomainSnapshotDefPtr virDomainSnapshotDefParseNode(xmlDocPtr xml, virDomainXMLOptionPtr xmlopt, unsigned int flags); void virDomainSnapshotDefFree(virDomainSnapshotDefPtr def); -char *virDomainSnapshotDefFormat(const char *domain_uuid, +char *virDomainSnapshotDefFormat(const char *uuidstr, virDomainSnapshotDefPtr def, virCapsPtr caps, virDomainXMLOptionPtr xmlopt, diff --git a/src/conf/snapshot_conf.c b/src/conf/snapshot_conf.c index 1afc7de30c..489c25f511 100644 --- a/src/conf/snapshot_conf.c +++ b/src/conf/snapshot_conf.c @@ -688,7 +688,7 @@ virDomainSnapshotDiskDefFormat(virBufferPtr buf, char * -virDomainSnapshotDefFormat(const char *domain_uuid, +virDomainSnapshotDefFormat(const char *uuidstr, virDomainSnapshotDefPtr def, virCapsPtr caps, virDomainXMLOptionPtr xmlopt, @@ -744,10 +744,10 @@ virDomainSnapshotDefFormat(const char *domain_uuid, if (def->dom) { if (virDomainDefFormatInternal(def->dom, caps, flags, &buf, xmlopt) < 0) goto error; - } else if (domain_uuid) { + } else if (uuidstr) { virBufferAddLit(&buf, "<domain>\n"); virBufferAdjustIndent(&buf, 2); - virBufferAsprintf(&buf, "<uuid>%s</uuid>\n", domain_uuid); + virBufferAsprintf(&buf, "<uuid>%s</uuid>\n", uuidstr); virBufferAdjustIndent(&buf, -2); virBufferAddLit(&buf, "</domain>\n"); } @@ -1224,11 +1224,8 @@ virDomainSnapshotRedefinePrep(virDomainPtr domain, int ret = -1; int align_location = VIR_DOMAIN_SNAPSHOT_LOCATION_INTERNAL; bool align_match = true; - char uuidstr[VIR_UUID_STRING_BUFLEN]; virDomainSnapshotObjPtr other; - virUUIDFormat(domain->uuid, uuidstr); - /* Prevent circular chains */ if (def->parent) { if (STREQ(def->name, def->parent)) { @@ -1274,6 +1271,9 @@ virDomainSnapshotRedefinePrep(virDomainPtr domain, if (def->dom && memcmp(def->dom->uuid, domain->uuid, VIR_UUID_BUFLEN)) { + char uuidstr[VIR_UUID_STRING_BUFLEN]; + + virUUIDFormat(domain->uuid, uuidstr); virReportError(VIR_ERR_INVALID_ARG, _("definition for snapshot %s must use uuid %s"), def->name, uuidstr); -- 2.20.1

1 0

[libvirt] [jenkins-ci PATCH] lcitool: Don't import crypt
by Andrea Bolognani 21 Feb '19

21 Feb '19

It hasn't been needed since dbe55924f90d, and flake8 rightfully complains about that. Signed-off-by: Andrea Bolognani <abologna(a)redhat.com> --- guests/lcitool | 1 - 1 file changed, 1 deletion(-) diff --git a/guests/lcitool b/guests/lcitool index 759eff6..e8263da 100755 --- a/guests/lcitool +++ b/guests/lcitool @@ -17,7 +17,6 @@ # with this program. If not, see <https://www.gnu.org/licenses/>. import argparse -import crypt import fnmatch import json import os -- 2.20.1

2 1

[libvirt] libvirt 'make cov' vs gnulib 'make coverage'
by Cole Robinson 21 Feb '19

21 Feb '19

Recently I've played with libvirt coverage/gcov/lcov support. The libvirt specific way to enable it is: ./configure --enable-test-coverage make check make cov firefox coverage/index.html However it looks like gnulib provides an automagic way to get similar output. The process is: make coverage firefox doc/coverage/index.html 'make coverage' is composed of these 3 additional commands: make init-coverage # clear lcov counters make build-coverage # 'make check' with coverage flags, generate # lcov metadata make gen-coverage # generate the html report It seems like we can just drop our custom coverage implementation. * Less code for us * The gnulib implementation is more flexible with the split commands * libvirt impl needs an explicit directory list. right now gnulib output reports info for the tools/ dir but libvirt's doesn't * gnulib impl doesn't require a reconfigure * seems like libvirt impl never calls lcov --zerocounters which AFAICT means that every 'make check' run adds to the coverage stats. so if you wanted to check for coverage changes across commits we require a make clean in between to zero out the stats for testing. gnulib init-coverage takes care of that. Anyone know a reason not to rip this out? If not I'll send a patch Thanks, Cole

1 0