Re: [libvirt] [PATCH 1/5] rpc: push ref acquisition into RPC dispatch function

On 03/06/2018 12:58 PM, Daniel P. Berrangé wrote: > There's no reason why the virNetServerClientDispatchRead method needs to > acquire an extra reference on the "client" object. An extra reference is > only needed if the registered dispatch callback is going to keep hold of > the "client" for

Re: [libvirt] [PATCH 2/5] rpc: simplify calling convention of virNetServerClientDispatchFunc

On 03/06/2018 12:58 PM, Daniel P. Berrangé wrote: > Currently virNetServerClientDispatchFunc implementations are only > responsible for free'ing the "msg" parameter upon success. Simplify the > calling convention by making it their unconditional responsibility to > free the "msg", and close the

Re: [libvirt] [PATCH 3/5] rpc: invoke the message dispatch callback with client unlocked

On 03/06/2018 12:58 PM, Daniel P. Berrangé wrote: > Currently if the virNetServer instance is created with max_workers==0 to > request a non-threaded dispatch process, we deadlock during dispatch > > #0 0x7fb845f6f42d in __lll_lock_wait () from /lib64/libpthread.so.0 > #1

Re: [libvirt] [PATCH v2 05/12] qemu: Store pr runtime data in status XML

On 03/06/2018 12:31 PM, Michal Privoznik wrote: > On 03/02/2018 01:50 PM, John Ferlan wrote: >> >> >> On 02/21/2018 01:11 PM, Michal Privoznik wrote: >>> Now that we generate pr-manager alias and socket path store them >>> in status XML so that they are preserved across daemon restarts. >>> >>>

Re: [libvirt] [PATCH v2 10/12] qemu_hotplug: Hotplug of reservations

On 03/06/2018 12:31 PM, Michal Privoznik wrote: > On 03/02/2018 06:55 PM, John Ferlan wrote: >> >> >> On 02/21/2018 01:11 PM, Michal Privoznik wrote: >>> Surprisingly, nothing special is happening here. If we are the >>> first to use the managed helper then spawn it. If not, we're >>> almost

Re: [libvirt] [PATCH v2 03/12] qemu: Introduce pr-manager-helper capability

On 03/06/2018 12:31 PM, Michal Privoznik wrote: > On 03/02/2018 02:58 AM, John Ferlan wrote: >> >> >> On 02/21/2018 01:11 PM, Michal Privoznik wrote: >>> The capability tracks if qemu has pr-manager-helper object. At >>> this time don't actually detect if qemu has the capability. Not >>> just

Re: [libvirt] [PATCH v2 06/12] qemu: Generate cmd line at startup

On 03/06/2018 12:31 PM, Michal Privoznik wrote: > On 03/02/2018 02:22 PM, John Ferlan wrote: >> >> >> On 02/21/2018 01:11 PM, Michal Privoznik wrote: >>> This is the easier part. All we need to do here is put -object >>> pr-manger-helper,id=$alias,path=$socketPath and then just >>> reference the

[libvirt] RFC regarding libvirtd deadlock

Doing something foolish like # while true; do killall -HUP libvirtd; sleep 1; done # virsh create vm.xml can deadlock libvirtd. Threads of interest are Thread 4 (Thread 0x7fc13b53e700 (LWP 64084)): #0 0x7fc13fba10bf in pthread_cond_wait@@GLIBC_2.3.2 () from /lib64/libpthread.so.0 #1

Re: [libvirt] [PATCH 5/5] rpc: switch virtlockd and virtlogd to use single-threaded dispatch

On 03/06/2018 12:58 PM, Daniel P. Berrangé wrote: > Currently both virtlogd and virtlockd use a single worker thread for > dispatching RPC messages. Even this is overkill and their RPC message > handling callbacks all run in short, finite time and so blocking the > main loop is not an issue like

Re: [libvirt] [PATCH 4/5] rpc: avoid crashing in pre-exec if no workers are present

On 03/06/2018 12:58 PM, Daniel P. Berrangé wrote: > If max_workers is set to zero, then the worker thread pool won't be > created, so when serializing state for pre-exec we must set various > parameters to zero. > > Signed-off-by: Daniel P. Berrangé > --- >

Re: [libvirt] [PATCH v2 02/12] qemuDomainDiskChangeSupported: Deny changing reservations

On 03/06/2018 12:31 PM, Michal Privoznik wrote: > On 03/02/2018 02:58 AM, John Ferlan wrote: >> >> >> On 02/21/2018 01:11 PM, Michal Privoznik wrote: >>> Couple of reasons for that: >>> >>> a) there's no monitor command to change path where the pr-helper >>> connects to, or >>> b) there's no

Re: [libvirt] [PATCH v2 09/12] qemu: Start PR daemons on domain startup

On 03/06/2018 12:31 PM, Michal Privoznik wrote: > On 03/02/2018 04:54 PM, John Ferlan wrote: >> >> >> On 02/21/2018 01:11 PM, Michal Privoznik wrote: >>> Before we exec() qemu we have to spawn pr-helper processes for >>> all managed reservations (well, technically there can only one). >> >> "can

[libvirt] [PATCH rebase v4 3/5] qemu: introduce qemuARPGetInterfaces to get IP from host's arp table

From: Chen Hanxiao introduce VIR_DOMAIN_INTERFACE_ADDRESSES_SRC_ARP to get ip address of VM from the message of netlink RTM_GETNEIGH Signed-off-by: Chen Hanxiao --- v4： remove dummy entry use VIR_APPEND_ELEMENT v3: add docs in

[libvirt] [PATCH rebase v4 0/5] qemu: use arp table of host to get the

introduce VIR_DOMAIN_INTERFACE_ADDRESSES_SRC_ARP to get ip address of VM from the output of /proc/net/arp Chen Hanxiao (5): util: introduce virNetlinkGetNeighbor to get neighbor table entry util: introduce helper to parse message from RTM_GETNEIGH query qemu: introduce qemuARPGetInterfaces

[libvirt] [PATCH rebase v4 1/5] util: introduce virNetlinkGetNeighbor to get neighbor table entry

From: Chen Hanxiao use RTM_GETNEIGH to query arp table entry by netlink socket Signed-off-by: Chen Hanxiao --- v4: use netlink to get arp table entry src/libvirt_private.syms | 1 + src/util/virnetlink.c| 82

[libvirt] [PATCH rebase v4 5/5] news: qemu: use arp table of host to get the IP address of guests

From: Chen Hanxiao Signed-off-by: Chen Hanxiao --- v4: rebase on 4.2 v3: more verbose description docs/news.xml | 9 + 1 file changed, 9 insertions(+) diff --git a/docs/news.xml b/docs/news.xml index a51ca973e..6d729d508 100644 ---

[libvirt] [PATCH rebase v4 2/5] util: introduce helper to parse message from RTM_GETNEIGH query

From: Chen Hanxiao introduce helper to parse RTM_GETNEIGH query message and store it in struct virArpTable. Signed-off-by: Chen Hanxiao --- v4-rebase: fit split Makefile.am fit new virMacAddr fields v4: use netlink query instead of parsing

[libvirt] [PATCH rebase v4 4/5] virsh: add --source arp to domifaddr

From: Chen Hanxiao We can use: domifaddr f26-cloud --source arp to get the address. Acked-by: Michal Privoznik Signed-off-by: Chen Hanxiao --- tools/virsh-domain-monitor.c | 2 ++ tools/virsh.pod | 7 --- 2

Re: [libvirt] [RFC] Defining firmware (OVMF, et al) metadata format & file

> Suggested approach > -- > > Based on an upstream discussion on 'virt-tools'[1] mailing list and some > Bugzillas, Gerd Hoffmann, Laszlo Ersek and Dan Berrangé had a suggestion > to define a firmware metadata format and file (example in [1]): > > - For each firmware file we

Re: [libvirt] [RFC] Defining firmware (OVMF, et al) metadata format & file

Hi, > > [*] Open question: Who, between QEMU and libvirt, should define the said > > firmware metadata format and file? > > IMHO QEMU should be defining the format, because the file will contain > info about certain QEMU features associated with the firmware (eg smm). > Also there are

[libvirt] [PATCH] apparmor: add ro rule for sasl GSSAPI plugin on /etc/gss/mech.d/

If a system has sasl GSSAPI plugin available qemu with sasl support will try to read /etc/gss/mech.d/. It is required to allow that to let the modules fully work and it should be safe to do so as it only registers/configures plugins but has no secrets. Signed-off-by: Christian Ehrhardt

Re: [libvirt] [PATCH 0/3] Drop old Policy-Kit support

On Wed, Mar 07, 2018 at 10:29:29AM +0100, Ján Tomko wrote: > The build WITH_POLKIT0 and -Wunused-label was broken for almost two > years. Last time we discussed this, SUSE folks said they still had polkit0 on an actively supported distro. I'd like confirmation they don't still need it before

Re: [libvirt] [perl PATCH 1/3] docs: replace $vmm with $conn

On Tue, Mar 06, 2018 at 01:33:00PM -0500, Laine Stump wrote: > This is more in line with the naming in C language examples, and > within libvirt's own code. > > Signed-off-by: Laine Stump > --- > lib/Sys/Virt.pm | 188 >

Re: [libvirt] [perl PATCH 2/3] docs: replace VMM with hypervisor

On Tue, Mar 06, 2018 at 01:33:01PM -0500, Laine Stump wrote: > I *think* this makes the most sense in the context of the > descriptions, but am open to other ideas (I also considered > "virtualization host", which might be better; it just seemed a bit > long...) > > Signed-off-by: Laine Stump

Re: [libvirt] [perl PATCH 3/3] docs: fix other assorted example variable names

On Tue, Mar 06, 2018 at 01:33:02PM -0500, Laine Stump wrote: > There were quite a few instances of apparent copy/paste from domain > examples over to examples of the same action on a different type of > object (e.g. calling define_network() and assigning the result to a > variable called $dom).

Re: [libvirt] [PATCH] m4: fix pkgcheck detection

On Tue, Mar 06, 2018 at 07:53:03PM +0100, Olaf Hering wrote: AC_PATH_PROG searches either for a binary [pkcheck], or it uses the user provided absolute path in the variable PKCHECK_PATH. Since virt-polkit.m4 wipes the variable PKCHECK_PATH from environment the user decision is never used. All

Re: [libvirt] [PATCH 5/5] rpc: switch virtlockd and virtlogd to use single-threaded dispatch

On Tue, Mar 06, 2018 at 04:46:05PM -0700, Jim Fehlig wrote: > On 03/06/2018 10:58 AM, Daniel P. Berrangé wrote: > > Currently both virtlogd and virtlockd use a single worker thread for > > dispatching RPC messages. Even this is overkill and their RPC message > > handling callbacks all run in

Re: [libvirt] [PATCH] apparmor: add ro rule for sasl GSSAPI plugin on /etc/gss/mech.d/

On Wed, Mar 07, 2018 at 11:16:52AM +0100, Christian Ehrhardt wrote: > If a system has sasl GSSAPI plugin available qemu with sasl support will > try to read /etc/gss/mech.d/. > > It is required to allow that to let the modules fully work and it should > be safe to do so as it only

Re: [libvirt] [PATCH 1/3] Extract stats functions from the qemu driver

$SUBJ, prefix with "conf:" would be nice. On 03/05/2018 12:21 PM, Cédric Bosdonnat wrote: > Some of the qemu functions getting statistics can easily be reused in > other drivers. Create a conf/domain_stats.[ch] pair to host some of > them. > --- > src/Makefile.am | 1 + >

Re: [libvirt] [PATCH 3/3] lxc: implement connectGetAllDomainStats

On 03/05/2018 12:21 PM, Cédric Bosdonnat wrote: > LXC containers can also provide some statistics. Allow users to fetch > them using the existing API. I think you're fetching more than "some" ;-)! > --- > src/lxc/lxc_driver.c | 203 > +++ > 1

Re: [libvirt] [PATCH 2/3] lxc: factorize code for block stats

s/factorize/Refactor/ On 03/05/2018 12:21 PM, Cédric Bosdonnat wrote: > lxcDomainBlockStats and lxcDomainBlockStatsFlags were both using very > similar code. This commit factorizes them into a s/factorsizes/refactors > lxcDomainBlockStatsInternal. > --- > src/lxc/lxc_driver.c | 131 >

[libvirt] [PATCH v2 03/10] tests: qemuxml2xml: Remove testing with allowed format detection

Nobody should use format detection due to security implications. The result of the change is that 'raw' format will be printed unless specified explicitly. Signed-off-by: Peter Krempa --- tests/qemuxml2argvdata/disk-drive-detect-zeroes.xml| 2 +-

[libvirt] [PATCH v2 06/10] tests: qemuxml2xml: Add synthetic test for in status XML

Add a random lockstate string to the status XML2XML test. Signed-off-by: Peter Krempa --- tests/qemustatusxml2xmldata/blockjob-mirror-in.xml | 1 + 1 file changed, 1 insertion(+) diff --git a/tests/qemustatusxml2xmldata/blockjob-mirror-in.xml

[libvirt] [PATCH v2 07/10] tests: qemuxml2xml: Remove fake status XML testing

Now that the better approach is in place we can remove the old functions doing the fake formatting. Signed-off-by: Peter Krempa --- tests/qemuxml2xmltest.c | 218 +--- 1 file changed, 1 insertion(+), 217 deletions(-) diff --git

[libvirt] [PATCH v2 05/10] tests: qemuxml2xml: Add proper domain status XML testing

Add new approach to properly test status XML files by supplying a full XML file rather than generating synthetic test cases by prepending the status header. The two tests introduced here are copies of existing cases using the synthetic header so that current level of testing is kept. The files are

Re: [libvirt] [PATCH 0/3] Drop old Policy-Kit support

On Wed, Mar 07, 2018 at 10:15:35 +, Daniel Berrange wrote: > On Wed, Mar 07, 2018 at 10:29:29AM +0100, Ján Tomko wrote: > > The build WITH_POLKIT0 and -Wunused-label was broken for almost two > > years. > > Last time we discussed this, SUSE folks said they still had polkit0 on > an actively

Re: [libvirt] [PATCH] qemuDomainObjBeginJob: Don't account DESTROY job to maxQueuedJobs

On 03/06/2018 05:20 PM, John Ferlan wrote: > > > On 03/05/2018 04:36 AM, Michal Privoznik wrote: >> When trying to destroy a domain (e.g. because we've seen EOF on >> the monitor) we try to acquire QEMU_JOB_DESTROY. However, if >> max_queued is set in qemu.conf this may fail and since our code

Re: [libvirt] [PATCH] virt-aa-helper: resolve yet to be created paths

On 03/01/2018 03:53 PM, Christian Ehrhardt wrote: > In certain cases a xml contains paths that do not yet exist, but > are valid as qemu will create them later on - for example > vhostuser mode=server sockets. > > In any such cases so far the check to virFileExists failed and due to > that the

Re: [libvirt] [PATCH 5/5] rpc: switch virtlockd and virtlogd to use single-threaded dispatch

On Wed, Mar 07, 2018 at 10:10:29AM +, Daniel P. Berrangé wrote: > On Tue, Mar 06, 2018 at 04:46:05PM -0700, Jim Fehlig wrote: > > On 03/06/2018 10:58 AM, Daniel P. Berrangé wrote: > > > Currently both virtlogd and virtlockd use a single worker thread for > > > dispatching RPC messages. Even

Re: [libvirt] [PATCH 0/3] Drop old Policy-Kit support

On Wed, Mar 07, 2018 at 12:09:15PM +0100, Peter Krempa wrote: > On Wed, Mar 07, 2018 at 10:15:35 +, Daniel Berrange wrote: > > On Wed, Mar 07, 2018 at 10:29:29AM +0100, Ján Tomko wrote: > > > The build WITH_POLKIT0 and -Wunused-label was broken for almost two > > > years. > > > > Last time we

[libvirt] [RFC] Defining firmware (OVMF, et al) metadata format & file

Problem background -- The various OVMF binary file names and paths are slightly different[+] for each Linux distribution. And each high-level management tool (libguestfs, oVirt, `virt-manager` and OpenStack Nova) is inventing its own approach to detect and configure the said OVMF

Re: [libvirt] [PATCH] qemu: Fix memory leak in qemuConnectGetAllDomainStats error path

On Wed, Mar 07, 2018 at 09:46:53 -0500, John Ferlan wrote: > If we return -1 on VIR_ALLOC_N failure, we leaked @vms, so > goto cleanup instead. > > Signed-off-by: John Ferlan > --- > src/qemu/qemu_driver.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) ACK

Re: [libvirt] [PATCH] util: mdev: Treat the 'name' sysfs attribute as optional

On 03/07/2018 02:46 AM, Erik Skultety wrote: > On Tue, Mar 06, 2018 at 11:47:51AM -0500, John Ferlan wrote: >> >> >> On 03/05/2018 09:43 AM, Erik Skultety wrote: >>> When commit 3545cbef moved the sysfs attribute reading logic from >>> _udev.c module to virmdev.c, it had to replace our udev read

[libvirt] [PATCH v2 00/10] Add proper testing of qemu status XML files

Changes to v1: - added tests specifically for testing format probing (also formating of argv) - removed format probing from snapshot tests - added a synthetic test for - added to the 'modern' test (note that allowReboot is tested in one of

[libvirt] [PATCH v2 08/10] tests: util: Remove callback from testCompareDomXML2XMLFiles

The testCompareDomXML2XMLPreFormatCallback is no longer used and thus can be removed. Signed-off-by: Peter Krempa --- tests/genericxml2xmltest.c | 2 +- tests/lxcxml2xmltest.c | 2 +- tests/qemuxml2xmltest.c| 5 ++--- tests/testutils.c | 8 +---

[libvirt] [PATCH v2 01/10] tests: qemuxml2xml: Rename testInfoFree to testInfoClear

Signed-off-by: Peter Krempa --- tests/qemuxml2xmltest.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/tests/qemuxml2xmltest.c b/tests/qemuxml2xmltest.c index 0eb9e6c77a..2cbe0c82a7 100644 --- a/tests/qemuxml2xmltest.c +++ b/tests/qemuxml2xmltest.c

[libvirt] [PATCH v2 02/10] tests: qemu: Explicitly add tests with format probing allowed

Add a single testcase for the case where format probing is allowed. Signed-off-by: Peter Krempa --- .../qemuxml2argvdata/disk-many-format-probing.args | 29 .../qemuxml2argvdata/disk-many-format-probing.xml | 48 tests/qemuxml2argvtest.c

[libvirt] [PATCH v2 04/10] tests: domainsnapshotxml2xml: Remove testing with allowed format detection

We have a test of the domain XML that allows format detection, so there's no need to do it for snapshot XMLs where the parameter would influence the domain XML portion. --- tests/domainsnapshotxml2xmlout/disk_snapshot_redefine.xml | 6 ++

[libvirt] [PATCH] qemu: Fix memory leak in qemuConnectGetAllDomainStats error path

If we return -1 on VIR_ALLOC_N failure, we leaked @vms, so goto cleanup instead. Signed-off-by: John Ferlan --- src/qemu/qemu_driver.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index

Re: [libvirt] [RFC] Defining firmware (OVMF, et al) metadata format & file

On Wed, Mar 07, 2018 at 03:49:51PM +0100, Kashyap Chamarthy wrote: > Problem background > -- > > The various OVMF binary file names and paths are slightly different[+] > for each Linux distribution. And each high-level management tool > (libguestfs, oVirt, `virt-manager` and

[libvirt] [PATCH v2 09/10] tests: qemuxml2xml: Add modern example of status XML to the test

Signed-off-by: Peter Krempa --- tests/qemustatusxml2xmldata/modern-in.xml | 448 + tests/qemustatusxml2xmldata/modern-out.xml | 1 + tests/qemuxml2xmltest.c| 1 + 3 files changed, 450 insertions(+) create mode 100644

[libvirt] [PATCH v2 10/10] tests: qemuxml2xml: Add status XML with outgoing migration with NBD

Signed-off-by: Peter Krempa --- .../qemustatusxml2xmldata/migration-out-nbd-in.xml | 449 + .../migration-out-nbd-out.xml | 1 + tests/qemuxml2xmltest.c| 1 + 3 files changed, 451 insertions(+) create

Re: [libvirt] [PATCH 0/3] Drop old Policy-Kit support

On 03/07/2018 03:15 AM, Daniel P. Berrangé wrote: On Wed, Mar 07, 2018 at 10:29:29AM +0100, Ján Tomko wrote: The build WITH_POLKIT0 and -Wunused-label was broken for almost two years. Last time we discussed this, SUSE folks said they still had polkit0 on an actively supported distro. I'd like

Re: [libvirt] [PATCH 5/5] rpc: switch virtlockd and virtlogd to use single-threaded dispatch

On 03/07/2018 06:07 AM, Daniel P. Berrangé wrote: On Wed, Mar 07, 2018 at 10:10:29AM +, Daniel P. Berrangé wrote: On Tue, Mar 06, 2018 at 04:46:05PM -0700, Jim Fehlig wrote: On 03/06/2018 10:58 AM, Daniel P. Berrangé wrote: Currently both virtlogd and virtlockd use a single worker thread

[libvirt] [PATCH 0/3] Drop old Policy-Kit support

The build WITH_POLKIT0 and -Wunused-label was broken for almost two years. Ján Tomko (3): Remove Policy-Kit support Merge WITH_POLKIT1 and WITH_POLKIT Do not check for pkcheck .gitignore | 1 - m4/virt-polkit.m4 |

[libvirt] [PATCH 1/3] Remove Policy-Kit support

Policy-Kit has been replaced by polkit (referred to as POLKIT0 and POLKIT1 in our Makefiles). The last build fix with old Policy-Kit was in May 2013: commit <442eb2ba> and build with -Wunused-label was broken since April 2016: commit <8437130> This includes a partial revert of commit , which

[libvirt] [PATCH 3/3] Do not check for pkcheck

All we need is DBus. Signed-off-by: Ján Tomko --- m4/virt-polkit.m4 | 30 +++--- 1 file changed, 11 insertions(+), 19 deletions(-) diff --git a/m4/virt-polkit.m4 b/m4/virt-polkit.m4 index 5c2a3c1e3..1016df4b3 100644 --- a/m4/virt-polkit.m4 +++

[libvirt] [PATCH 2/3] Merge WITH_POLKIT1 and WITH_POLKIT

There is just one polkit now. Signed-off-by: Ján Tomko --- m4/virt-polkit.m4 | 6 -- src/access/Makefile.inc.am| 6 +++--- src/access/viraccessmanager.c | 4 ++-- src/util/virpolkit.c | 6 +++--- tests/Makefile.am | 4 ++-- 5 files

Re: [libvirt] [PATCH] virt-aa-helper: resolve yet to be created paths

On Wed, Mar 7, 2018 at 12:45 PM, Michal Privoznik wrote: > On 03/01/2018 03:53 PM, Christian Ehrhardt wrote: > > In certain cases a xml contains paths that do not yet exist, but > > are valid as qemu will create them later on - for example > > vhostuser mode=server sockets.

Re: [libvirt] [PATCH 5/5] rpc: switch virtlockd and virtlogd to use single-threaded dispatch

On 03/06/2018 10:58 AM, Daniel P. Berrangé wrote: Currently both virtlogd and virtlockd use a single worker thread for dispatching RPC messages. Even this is overkill and their RPC message handling callbacks all run in short, finite time and so blocking the main loop is not an issue like you'd

Re: [libvirt] [PATCH v5 11/12] qemu: Validate PCI controllers (QEMU capabilities)

On Mon, 2018-03-05 at 16:58 +0100, Andrea Bolognani wrote: > I'm actually fairly confident we could drop that hunk with no ill > effect, because the spapr-pci-host-bridge must have existed even > before we started checking for it, even if it might possibly have > been, at some point, not

Re: [libvirt] [PATCH] util: mdev: Treat the 'name' sysfs attribute as optional

On Wed, Mar 07, 2018 at 10:31:27AM -0500, John Ferlan wrote: > > > On 03/07/2018 02:46 AM, Erik Skultety wrote: > > On Tue, Mar 06, 2018 at 11:47:51AM -0500, John Ferlan wrote: > >> > >> > >> On 03/05/2018 09:43 AM, Erik Skultety wrote: > >>> When commit 3545cbef moved the sysfs attribute reading